New upstream version 3.4.14upstream/3.4.14

682 files changed, 16256 insertions, 5608 deletions

diff --git a/README b/README
index 858f8a1d2c8..5a772b66e7d 100644
--- a/README
+++ b/README
@@ -4,23 +4,26 @@ Welcome to MongoDB 3.4!
 

 COMPONENTS

 

-  mongod - The database process.

-  mongos - Sharding controller.

+  mongod - The database server.

+  mongos - Sharding router.

   mongo  - The database shell (uses interactive javascript).

 

 UTILITIES

 

-  mongodump         - MongoDB dump tool - for backups, snapshots, etc.

-  mongorestore      - MongoDB restore a dump

-  mongoexport       - Export a single collection to JSON or CSV

-  mongoimport       - Import from JSON or CSV

-  mongofiles        - Utility for putting and getting files from MongoDB GridFS

-  mongostat         - Show performance statistics

-  mongoreplay       - Workload capture and analysis tool

+  mongodump         - Create a binary dump of the contents of a database.

+  mongorestore      - Restore data from the output created by mongodump.

+  mongoexport       - Export the contents of a collection to JSON or CSV.

+  mongoimport       - Import data from JSON, CSV or TSV.

+  mongofiles        - Put, get and delete files from GridFS.

+  mongostat         - Show the status of a running mongod/mongos.

+  bsondump          - Convert BSON files into human-readable formats.

+  mongooplog        - Poll the oplog and apply to a local server.

+  mongoreplay       - Traffic capture and replay tool.

+  mongotop          - Track time spent reading and writing data.

 

 BUILDING

 

-  See docs/building.md, also www.mongodb.org search for "Building".

+  See docs/building.md.

 

 RUNNING

 

@@ -30,7 +33,7 @@ RUNNING
 

   To run a single server database:

 

-    $ mkdir /data/db

+    $ sudo mkdir -p /data/db

     $ ./mongod

     $

     $ # The mongo javascript shell connects to localhost and test database by default:

@@ -40,7 +43,12 @@ RUNNING
 DRIVERS

 

   Client drivers for most programming languages are available at

-  mongodb.org.  Use the shell ("mongo") for administrative tasks.

+  https://docs.mongodb.com/manual/applications/drivers/. Use the shell

+  ("mongo") for administrative tasks.

+

+BUG REPORTS

+

+  See https://github.com/mongodb/mongo/wiki/Submit-Bug-Reports.

 

 PACKAGING

 

@@ -49,26 +57,28 @@ PACKAGING
 

 DOCUMENTATION

 

-  http://www.mongodb.org/

+  https://docs.mongodb.com/manual/

+

+CLOUD HOSTED MONGODB

 

-CLOUD MANAGED MONGODB

+  https://www.mongodb.com/cloud/atlas

 

-  http://cloud.mongodb.com/

+MAIL LISTS

 

-MAIL LISTS AND IRC

+  https://groups.google.com/forum/#!forum/mongodb-user

 

-  http://dochub.mongodb.org/core/community

+    A forum for technical questions about using MongoDB.

+

+  https://groups.google.com/forum/#!forum/mongodb-dev

+

+    A forum for technical questions about building and developing MongoDB.

 

 LEARN MONGODB

 

-  http://university.mongodb.com/

+  https://university.mongodb.com/

 

 LICENSE

 

   Most MongoDB source files (src/mongo folder and below) are made available

-  under the terms of the GNU Affero General Public License (AGPL).  See

+  under the terms of the GNU Affero General Public License (GNU AGPLv3). See

   individual files for details.

-

-  As an exception, the files in the client/, debian/, rpm/,

-  utils/mongoutils, and all subdirectories thereof are made available under

-  the terms of the Apache License, version 2.0.

diff --git a/buildscripts/burn_in_tests.py b/buildscripts/burn_in_tests.py
index d171a5a6bc1..6e39685ef8b 100644
--- a/buildscripts/burn_in_tests.py
+++ b/buildscripts/burn_in_tests.py
@@ -17,7 +17,7 @@ import sys
 import urlparse
 import yaml
 
-API_SERVER_DEFAULT = "http://evergreen-api.mongodb.com:8080"
+API_SERVER_DEFAULT = "https://evergreen.mongodb.com"
 
 # Get relative imports to work when the package is not installed on the PYTHONPATH.
 if __name__ == "__main__" and __package__ is None:
@@ -418,6 +418,8 @@ def main():
         tests_by_task = _load_tests_file(values.test_list_file)
         # If there are no tests to run, carry on.
         if tests_by_task is None:
+            test_results = {"failures": 0, "results": []}
+            _write_report_file(test_results, values.report_file)
             sys.exit(0)
 
     # Run the executor finder.
@@ -438,6 +440,7 @@ def main():
         # If there are no changed tests, exit cleanly.
         if not changed_tests:
             print "No new or modified tests found."
+            _write_report_file({}, values.test_list_outfile)
             sys.exit(0)
         suites = resmokelib.parser.get_suites(values, changed_tests)
         tests_by_executor = create_executor_list(suites, exclude_suites)
diff --git a/buildscripts/hang_analyzer.py b/buildscripts/hang_analyzer.py
index 73e1cbe6d41..09245904a7f 100755
--- a/buildscripts/hang_analyzer.py
+++ b/buildscripts/hang_analyzer.py
@@ -24,6 +24,7 @@ import signal
 import subprocess
 import sys
 import tempfile
+import traceback
 import time
 from distutils import spawn
 from optparse import OptionParser
@@ -53,7 +54,7 @@ def call(a, logger):
 
     if ret != 0:
         logger.error("Bad exit code %d" % (ret))
-        raise Exception()
+        raise Exception("Bad exit code %d from %s" % (ret, " ".join(a)))
 
 
 # Copied from python 2.7 version of subprocess.py
@@ -375,14 +376,14 @@ class GDBDumper(object):
             base, ext = os.path.splitext(logger.mongo_process_filename)
             raw_stacks_filename = base + '_raw_stacks' + ext
             raw_stacks_commands = [
-                    'echo \\nWriting raw stacks to %s.\\n' % raw_stacks_filename,
-                    # This sends output to log file rather than stdout until we turn logging off.
-                    'set logging redirect on',
-                    'set logging file ' + raw_stacks_filename,
-                    'set logging on',
-                    'thread apply all bt',
-                    'set logging off',
-                    ]
+                'echo \\nWriting raw stacks to %s.\\n' % raw_stacks_filename,
+                # This sends output to log file rather than stdout until we turn logging off.
+                'set logging redirect on',
+                'set logging file ' + raw_stacks_filename,
+                'set logging on',
+                'thread apply all bt',
+                'set logging off',
+                ]
 
         cmds = [
             "set interactive-mode off",
@@ -732,28 +733,38 @@ def main():
         # a signal handler to wait for the signal since it supports POSIX signals.
         if _is_windows:
             root_logger.info("Calling SetEvent to signal python process %s with PID %d" %
-                (process_name, pid))
+                             (process_name, pid))
             signal_event_object(root_logger, pid)
         else:
             root_logger.info("Sending signal SIGUSR1 to python process %s with PID %d" %
-                (process_name, pid))
+                             (process_name, pid))
             signal_process(root_logger, pid, signal.SIGUSR1)
 
+    trapped_exceptions = []
+
     # Dump all processes, except python & java.
     for (pid, process_name) in [(p, pn) for (p, pn) in processes
                                 if not re.match("^(java|python)", pn)]:
         process_logger = get_process_logger(options.debugger_output, pid, process_name)
-        dbg.dump_info(
-            root_logger,
-            process_logger,
-            pid,
-            process_name,
-            options.dump_core and check_dump_quota(max_dump_size_bytes, dbg.get_dump_ext()))
+        try:
+            dbg.dump_info(
+                root_logger,
+                process_logger,
+                pid,
+                process_name,
+                options.dump_core and check_dump_quota(max_dump_size_bytes, dbg.get_dump_ext()))
+        except Exception as err:
+            root_logger.info("Error encountered when invoking debugger %s" % err)
+            trapped_exceptions.append(traceback.format_exc())
 
     # Dump java processes using jstack.
     for (pid, process_name) in [(p, pn) for (p, pn) in processes if pn.startswith("java")]:
         process_logger = get_process_logger(options.debugger_output, pid, process_name)
-        jstack.dump_info(root_logger, process_logger, pid, process_name)
+        try:
+            jstack.dump_info(root_logger, process_logger, pid, process_name)
+        except Exception as err:
+            root_logger.info("Error encountered when invoking debugger %s" % err)
+            trapped_exceptions.append(traceback.format_exc())
 
     # Signal go processes to ensure they print out stack traces, and die on POSIX OSes.
     # On Windows, this will simply kill the process since python emulates SIGABRT as
@@ -761,10 +772,15 @@ def main():
     # Note: The stacktrace output may be captured elsewhere (i.e. resmoke).
     for (pid, process_name) in [(p, pn) for (p, pn) in processes if pn in go_processes]:
         root_logger.info("Sending signal SIGABRT to go process %s with PID %d" %
-            (process_name, pid))
+                         (process_name, pid))
         signal_process(root_logger, pid, signal.SIGABRT)
 
     root_logger.info("Done analyzing all processes for hangs")
 
+    for exception in trapped_exceptions:
+        root_logger.info(exception)
+    if trapped_exceptions:
+        sys.exit(1)
+
 if __name__ == "__main__":
     main()
diff --git a/buildscripts/package_test/.kitchen.yml b/buildscripts/package_test/.kitchen.yml
index e9b6989d221..3b4a476551c 100644
--- a/buildscripts/package_test/.kitchen.yml
+++ b/buildscripts/package_test/.kitchen.yml
@@ -2,11 +2,13 @@
 driver:
   name: ec2
   region: us-east-1
+  vpc_mode: true
+  vpc_id: <%= ENV['KITCHEN_VPC'] %>
   subnet_id: <%= ENV['KITCHEN_SUBNET'] %>
   security_group_ids:
     - <%= ENV['KITCHEN_SECURITY_GROUP'] %>
   aws_ssh_key_id: <%= ENV['KITCHEN_SSH_KEY_ID'] %>
-  interface: dns
+  interface: private
   associate_public_ip: true
 
 verifier:
diff --git a/buildscripts/package_test/recipes/install_mongodb.rb b/buildscripts/package_test/recipes/install_mongodb.rb
index dcf7499289e..9b4ea703b2b 100644
--- a/buildscripts/package_test/recipes/install_mongodb.rb
+++ b/buildscripts/package_test/recipes/install_mongodb.rb
@@ -84,6 +84,16 @@ if platform_family? 'suse'
   EOD
   end
 
+  %w(
+     SLES12-Pool
+     SLES12-Updates
+  ).each do |repo|
+    execute "add #{repo}" do
+      command "zypper addrepo --check --refresh --name \"#{repo}\" http://smt-ec2.susecloud.net/repo/SUSE/Products/SLE-SERVER/12/x86_64/product?credentials=SMT-http_smt-ec2_susecloud_net 'SMT-http_smt-ec2_susecloud_net:#{repo}'"
+      not_if "zypper lr | grep #{repo}"
+    end
+  end
+
   execute 'install mongod' do
     command 'zypper -n install `find . -name "*server*.rpm"`'
     cwd homedir
diff --git a/buildscripts/resmokeconfig/suites/aggregation_read_concern_majority_passthrough.yml b/buildscripts/resmokeconfig/suites/aggregation_read_concern_majority_passthrough.yml
index bc227b09279..72b838b3c07 100644
--- a/buildscripts/resmokeconfig/suites/aggregation_read_concern_majority_passthrough.yml
+++ b/buildscripts/resmokeconfig/suites/aggregation_read_concern_majority_passthrough.yml
@@ -6,11 +6,28 @@ selector:
     - jstests/aggregation/expressions/*.js
     - jstests/aggregation/sources/*/*.js
     exclude_files:
-    - jstests/aggregation/bugs/server18198.js # Uses a mocked mongo client to test read preference.
-    - jstests/aggregation/mongos_slaveok.js # Majority read on secondary requires afterOpTime.
-    - jstests/aggregation/sources/facet/use_cases.js # Cannot specify write concern when
-                                                     # secondaryThrottle is not set.
-    - jstests/aggregation/testSlave.js # Majority read on secondary requires afterOpTime.
+    # These tests fail due to the inability to specify a writeConcern when secondaryThrottle is not
+    # set as part of the moveChunk command.
+    - jstests/aggregation/sources/facet/use_cases.js
+    # These test fail because afterOpTime is required to guarantee a secondary has advanced its
+    # majority-committed snapshot.
+    - jstests/aggregation/mongos_slaveok.js
+    - jstests/aggregation/testSlave.js
+    exclude_with_any_tags:
+    ##
+    # The next three tags correspond to the special errors thrown by the
+    # set_read_and_write_concerns.js override when it refuses to replace the readConcern or
+    # writeConcern of a particular command. Above each tag are the message(s) that cause the tag to
+    # be warranted.
+    ##
+    # "Cowardly refusing to override read concern of command: ..."
+    - assumes_read_concern_unchanged
+    # "Cowardly refusing to override write concern of command: ..."
+    - assumes_write_concern_unchanged
+    # "Cowardly refusing to run test with overridden write concern when it uses a command that can
+    #  only perform w=1 writes: ..."
+    - requires_collmod_command
+    - requires_eval_command
 
 executor:
   js_test:
@@ -18,13 +35,19 @@ executor:
       shell_options:
         global_vars:
           TestData:
+            defaultReadConcernLevel: majority
             enableMajorityReadConcern: ''
-        eval: "var testingReplication = true; load('jstests/libs/override_methods/set_majority_read_and_write_concerns.js');"
+        eval: >-
+          var testingReplication = true;
+          load('jstests/libs/override_methods/set_read_and_write_concerns.js');
         readMode: commands
     hooks:
-    - class: ValidateCollections
+    # The CheckReplDBHash hook waits until all operations have replicated to and have been applied
+    # on the secondaries, so we run the ValidateCollections hook after it to ensure we're
+    # validating the entire contents of the collection.
     - class: CheckReplOplogs
     - class: CheckReplDBHash
+    - class: ValidateCollections
     fixture:
       class: ReplicaSetFixture
       mongod_options:
@@ -33,5 +56,3 @@ executor:
           enableTestCommands: 1
           numInitialSyncAttempts: 1
       num_nodes: 2
-      # Needs to be set for any ephemeral or no-journaling storage engine
-      write_concern_majority_journal_default: false
diff --git a/buildscripts/resmokeconfig/suites/read_concern_majority_passthrough.yml b/buildscripts/resmokeconfig/suites/read_concern_majority_passthrough.yml
index f47e424e1cf..22e84bc9c63 100644
--- a/buildscripts/resmokeconfig/suites/read_concern_majority_passthrough.yml
+++ b/buildscripts/resmokeconfig/suites/read_concern_majority_passthrough.yml
@@ -3,79 +3,47 @@ selector:
     roots:
     - jstests/core/**/*.js
     exclude_files:
-    # Tests that won't work with an injected 'majority' readConcern
-    # and/or an injected 'majority' writeConcern. Where a function is
-    # listed the reason is we don't have a reliable solution to override
-    # the write concern for that function.
-    - jstests/core/batch_write_command*.js # these tests use various write concerns
-    - jstests/core/bench_test*.js # benchRun() used for writes
-    - jstests/core/capped_update.js # uses godinsert and can't run under replication.
-    - jstests/core/crud_api.js # has specific w:0 tests
-    - jstests/core/error2.js # db.eval() used
-    - jstests/core/eval0.js # db.eval() used
-    - jstests/core/eval1.js # db.eval() used
-    - jstests/core/eval3.js # db.eval() used
-    - jstests/core/eval4.js # db.eval() used
-    - jstests/core/eval5.js # db.eval() used
-    - jstests/core/eval6.js # db.eval() used
-    - jstests/core/eval7.js # db.eval() used
-    - jstests/core/eval9.js # db.eval() used
-    - jstests/core/evala.js # db.eval() used
-    - jstests/core/evalb.js # db.eval() used
-    - jstests/core/evald.js # db.eval() used
-    - jstests/core/evale.js # db.eval() used
-    - jstests/core/evalg.js # db.eval() used
-    - jstests/core/eval_mr.js # db.eval() used
-    - jstests/core/eval_nolock.js # db.eval() used
-    - jstests/core/geo_s2cursorlimitskip.js # drops system.profile collection and counts ops.
-    - jstests/core/js3.js # db.dbEval() used
-    - jstests/core/js7.js # db.eval() used
-    - jstests/core/js9.js # db.eval() used
-    - jstests/core/mr_merge.js # mr temp tables aren't replicated
-    - jstests/core/mr_merge2.js # mr temp tables aren't replicated
-    - jstests/core/mr_outreduce.js # mr temp tables aren't replicated
-    - jstests/core/mr_outreduce2.js # mr temp tables aren't replicated
-    - jstests/core/opcounters_active.js  # off by n problem with opcounters
-    - jstests/core/opcounters_write_cmd.js # off by n problem with opcounters
-    - jstests/core/profile_agg.js # system.profile not replicated
-    - jstests/core/profile_count.js # system.profile not replicated
-    - jstests/core/profile_delete.js # system.profile not replicated
-    - jstests/core/profile_distinct.js # system.profile not replicated
-    - jstests/core/profile_find.js # system.profile not replicated
-    - jstests/core/profile_findandmodify.js # system.profile not replicated
-    - jstests/core/profile_geonear.js # system.profile not replicated
-    - jstests/core/profile_getmore.js # system.profile not replicated
-    - jstests/core/profile_group.js # system.profile not replicated
-    - jstests/core/profile_insert.js # system.profile not replicated
-    - jstests/core/profile_mapreduce.js # system.profile not replicated
-    - jstests/core/profile_update.js # system.profile not replicated
-    - jstests/core/profile1.js # system.profile not replicated
-    - jstests/core/profile2.js # system.profile not replicated
-    - jstests/core/profile3.js # system.profile not replicated
-    - jstests/core/read_after_optime.js # verifies read after optime fails on standalone
-    - jstests/core/remove8.js # db.eval() used
-    - jstests/core/rename4.js # db.eval() used
-    - jstests/core/shell1.js # tests setSlaveOk() variations on standalone mongod
-    - jstests/core/shellkillop.js # db.eval() used
-    - jstests/core/shell_writeconcern.js # checks write concern shell helpers
-    - jstests/core/storefunc.js # db.eval() used
-    - jstests/core/write_result.js # Tests invalid writeConcern, we shouldn't override.
-    # Tests that need triaging & remediation | blacklist decision
-    # Comments list possible problem point under review.
-    - jstests/core/capped6.js # Uses captrunc test command.
-    - jstests/core/convert_to_capped_nonexistant.js # Uses convertToCapped and captrunc command.
-    - jstests/core/stages_delete.js # Uses stageDebug command for deletes.
+    # These tests are not expected to pass with replica-sets:
+    - jstests/core/dbadmin.js
+    - jstests/core/opcounters_write_cmd.js
+    - jstests/core/read_after_optime.js
+    - jstests/core/capped_update.js
+    # These tests use benchRun(), which isn't configured to use the overridden writeConcern.
+    - jstests/core/bench_test*.js
+    exclude_with_any_tags:
+    ##
+    # The next three tags correspond to the special errors thrown by the
+    # set_read_and_write_concerns.js override when it refuses to replace the readConcern or
+    # writeConcern of a particular command. Above each tag are the message(s) that cause the tag to
+    # be warranted.
+    ##
+    # "Cowardly refusing to override read concern of command: ..."
+    - assumes_read_concern_unchanged
+    # "Cowardly refusing to override write concern of command: ..."
+    - assumes_write_concern_unchanged
+    # "Cowardly refusing to run test with overridden write concern when it uses a command that can
+    #  only perform w=1 writes: ..."
+    - requires_collmod_command
+    - requires_eval_command
 
 executor:
   js_test:
     config:
       shell_options:
-        eval: "var testingReplication = true; load('jstests/libs/override_methods/set_majority_read_and_write_concerns.js');"
+        global_vars:
+          TestData:
+            defaultReadConcernLevel: majority
+        eval: >-
+          var testingReplication = true;
+          load('jstests/libs/override_methods/set_read_and_write_concerns.js');
         readMode: commands
     hooks:
-    - class: ValidateCollections
+    # The CheckReplDBHash hook waits until all operations have replicated to and have been applied
+    # on the secondaries, so we run the ValidateCollections hook after it to ensure we're
+    # validating the entire contents of the collection.
     - class: CheckReplOplogs
     - class: CheckReplDBHash
+    - class: ValidateCollections
     - class: CleanEveryN
       n: 20
     fixture:
@@ -84,7 +52,6 @@ executor:
         set_parameters:
           enableTestCommands: 1
           numInitialSyncAttempts: 1
+          writePeriodicNoops: 1
         enableMajorityReadConcern: ''
       num_nodes: 2
-      # Needs to be set for any ephemeral or no-journaling storage engine
-      write_concern_majority_journal_default: false
diff --git a/buildscripts/resmokeconfig/suites/replica_sets_initsync_jscore_passthrough.yml b/buildscripts/resmokeconfig/suites/replica_sets_initsync_jscore_passthrough.yml
index 832e0ab9bfc..114ebcab372 100644
--- a/buildscripts/resmokeconfig/suites/replica_sets_initsync_jscore_passthrough.yml
+++ b/buildscripts/resmokeconfig/suites/replica_sets_initsync_jscore_passthrough.yml
@@ -14,6 +14,7 @@ selector:
     - jstests/core/commands_that_do_not_write_do_not_accept_wc.js
     - jstests/core/constructors.js
     - jstests/core/eval_mr.js
+    - jstests/core/function_string_representations.js
     - jstests/core/geo_big_polygon3.js
     - jstests/core/geo_mapreduce.js
     - jstests/core/geo_mapreduce2.js
diff --git a/buildscripts/resmokeconfig/suites/replica_sets_resync_jscore_passthrough.yml b/buildscripts/resmokeconfig/suites/replica_sets_resync_jscore_passthrough.yml
index 3b48456c92e..1e51713a475 100644
--- a/buildscripts/resmokeconfig/suites/replica_sets_resync_jscore_passthrough.yml
+++ b/buildscripts/resmokeconfig/suites/replica_sets_resync_jscore_passthrough.yml
@@ -14,6 +14,7 @@ selector:
     - jstests/core/commands_that_do_not_write_do_not_accept_wc.js
     - jstests/core/constructors.js
     - jstests/core/eval_mr.js
+    - jstests/core/function_string_representations.js
     - jstests/core/geo_big_polygon3.js
     - jstests/core/geo_mapreduce.js
     - jstests/core/geo_mapreduce2.js
diff --git a/buildscripts/resmokeconfig/suites/sharding_continuous_config_stepdown.yml b/buildscripts/resmokeconfig/suites/sharding_continuous_config_stepdown.yml
index 0bafa32a1e5..acf8ae53e96 100644
--- a/buildscripts/resmokeconfig/suites/sharding_continuous_config_stepdown.yml
+++ b/buildscripts/resmokeconfig/suites/sharding_continuous_config_stepdown.yml
@@ -50,6 +50,7 @@ selector:
     - jstests/sharding/shard2.js
     - jstests/sharding/shard3.js
     - jstests/sharding/shard_collection_basic.js
+    - jstests/sharding/shard_existing_coll_chunk_count.js
     - jstests/sharding/sharding_balance1.js
     - jstests/sharding/sharding_balance2.js
     - jstests/sharding/sharding_balance3.js
diff --git a/buildscripts/resmokeconfig/suites/sharding_last_stable_mongos_and_mixed_shards.yml b/buildscripts/resmokeconfig/suites/sharding_last_stable_mongos_and_mixed_shards.yml
index f4c741badd4..c7ff1b24395 100644
--- a/buildscripts/resmokeconfig/suites/sharding_last_stable_mongos_and_mixed_shards.yml
+++ b/buildscripts/resmokeconfig/suites/sharding_last_stable_mongos_and_mixed_shards.yml
@@ -95,6 +95,8 @@ selector:
     # TODO (SERVER-27379): Cannot create views when featureCompatibilityVersion is 3.2.
     # Unblacklist when the featureCompatibilityVersion is 3.4 or higher.
     - jstests/sharding/views.js
+    # SERVER-20392: Reenable when backported to 3.4 and released as last-stable.
+    - jstests/sharding/shard_existing_coll_chunk_count.js
 
 executor:
   js_test:
diff --git a/buildscripts/resmokeconfig/suites/write_concern_majority_passthrough.yml b/buildscripts/resmokeconfig/suites/write_concern_majority_passthrough.yml
new file mode 100644
index 00000000000..388046d2782
--- /dev/null
+++ b/buildscripts/resmokeconfig/suites/write_concern_majority_passthrough.yml
@@ -0,0 +1,78 @@
+test_kind: js_test
+
+selector:
+  js_test:
+    roots:
+    - jstests/core/**/*.js
+    exclude_files:
+    # These tests are not expected to pass with replica-sets:
+    - jstests/core/dbadmin.js
+    - jstests/core/opcounters_write_cmd.js
+    - jstests/core/read_after_optime.js
+    - jstests/core/capped_update.js
+    # The connection_string_validation.js test does not expect the mongo shell to be using a replica
+    # set connection string.
+    - jstests/core/connection_string_validation.js
+    # These tests attempt to read from the "system.profile" collection, which may be missing entries
+    # if a write was performed on the primary of the replica set instead.
+    - jstests/core/*profile*.js
+    # The shellkillop.js test spawns a parallel shell without using startParallelShell() and
+    # therefore doesn't inherit the w="majority" write concern when performing its writes.
+    - jstests/core/shellkillop.js
+    exclude_with_any_tags:
+    ##
+    # The next three tags correspond to the special errors thrown by the
+    # set_read_and_write_concerns.js override when it refuses to replace the readConcern or
+    # writeConcern of a particular command. Above each tag are the message(s) that cause the tag to
+    # be warranted.
+    ##
+    # "Cowardly refusing to override read concern of command: ..."
+    - assumes_read_concern_unchanged
+    # "Cowardly refusing to override write concern of command: ..."
+    - assumes_write_concern_unchanged
+    # "Cowardly refusing to run test with overridden write concern when it uses a command that can
+    #  only perform w=1 writes: ..."
+    - requires_collmod_command
+    - requires_eval_command
+    ##
+    # The next tag corresponds to the special error thrown by the set_read_preference_secondary.js
+    # override when it refuses to replace the readPreference of a particular command. Above each tag
+    # are the message(s) that cause the tag to be warranted.
+    ##
+    # "Cowardly refusing to override read preference of command: ..."
+    # "Cowardly refusing to run test with overridden read preference when it reads from a
+    #  non-replicated collection: ..."
+    - assumes_read_preference_unchanged
+
+executor:
+  js_test:
+    config:
+      shell_options:
+        global_vars:
+          TestData:
+            defaultReadConcernLevel: local
+        eval: >-
+          testingReplication = true;
+          load('jstests/libs/override_methods/set_read_and_write_concerns.js');
+          load('jstests/libs/override_methods/set_read_preference_secondary.js');
+        readMode: commands
+    hooks:
+    # The CheckReplDBHash hook waits until all operations have replicated to and have been applied
+    # on the secondaries, so we run the ValidateCollections hook after it to ensure we're
+    # validating the entire contents of the collection.
+    - class: CheckReplOplogs
+    - class: CheckReplDBHash
+    - class: ValidateCollections
+    - class: CleanEveryN
+      n: 20
+    fixture:
+      class: ReplicaSetFixture
+      mongod_options:
+        set_parameters:
+          enableTestCommands: 1
+          numInitialSyncAttempts: 1
+      # This suite requires w="majority" writes to be applied on all secondaries. By using a 2-node
+      # replica set and having secondaries vote, the majority of the replica set is all nodes.
+      num_nodes: 2
+      voting_secondaries: true
+      use_replica_set_connection_string: true
diff --git a/buildscripts/resmokelib/core/programs.py b/buildscripts/resmokelib/core/programs.py
index e4cadc8e006..d94cd438ee0 100644
--- a/buildscripts/resmokelib/core/programs.py
+++ b/buildscripts/resmokelib/core/programs.py
@@ -115,8 +115,8 @@ def mongos_program(logger, executable=None, process_kwargs=None, **kwargs):
     return _process.Process(logger, args, **process_kwargs)
 
 
-def mongo_shell_program(logger, executable=None, filename=None, process_kwargs=None,
-                        isMainTest=True, **kwargs):
+def mongo_shell_program(logger, executable=None, connection_string=None, filename=None,
+                        process_kwargs=None, isMainTest=True, **kwargs):
     """
     Returns a Process instance that starts a mongo shell with arguments
     constructed from 'kwargs'.
@@ -184,9 +184,22 @@ def mongo_shell_program(logger, executable=None, filename=None, process_kwargs=N
     if config.SHELL_WRITE_MODE is not None:
         kwargs["writeMode"] = config.SHELL_WRITE_MODE
 
+    if connection_string is not None:
+        # The --host and --port options are ignored by the mongo shell when an explicit connection
+        # string is specified. We remove these options to avoid any ambiguity with what server the
+        # logged mongo shell invocation will connect to.
+        if "port" in kwargs:
+            kwargs.pop("port")
+
+        if "host" in kwargs:
+            kwargs.pop("host")
+
     # Apply the rest of the command line arguments.
     _apply_kwargs(args, kwargs)
 
+    if connection_string is not None:
+        args.append(connection_string)
+
     # Have the mongos shell run the specified file.
     args.append(filename)
 
diff --git a/buildscripts/resmokelib/logging/buildlogger.py b/buildscripts/resmokelib/logging/buildlogger.py
index f058372df8c..e8e37d2c004 100644
--- a/buildscripts/resmokelib/logging/buildlogger.py
+++ b/buildscripts/resmokelib/logging/buildlogger.py
@@ -5,7 +5,9 @@ Defines handlers for communicating with a buildlogger server.
 from __future__ import absolute_import
 
 import functools
+import httplib
 import urllib2
+import json
 
 from . import handlers
 from . import loggers
@@ -131,6 +133,47 @@ def new_test_id(build_id, build_config, test_filename, test_command):
     return response["id"]
 
 
+class _LogsSplitter(object):
+    """Class with static methods used to split list of log lines into smaller batches."""
+
+    @staticmethod
+    def split_logs(log_lines, max_size):
+        """
+        Splits the log lines into batches of size less than or equal to max_size.
+
+        Args:
+            log_lines: A list of log lines.
+            max_size: The maximum size in bytes a batch of log lines can have in JSON.
+        Returns:
+            A list of list of log lines. Each item is a list is a list of log lines
+            satisfying the size requirement.
+        """
+        if not max_size:
+            return [log_lines]
+
+        def line_size(line):
+            """
+            Computes the encoded JSON size of a log line as part of an array.
+            2 is added to each string size to account for the array representation of the logs,
+            as each line is preceded by a '[' or a space and followed by a ',' or a ']'.
+            """
+            return len(json.dumps(line, encoding="utf-8")) + 2
+
+        curr_logs = []
+        curr_logs_size = 0
+        split_logs = []
+        for line in log_lines:
+            size = line_size(line)
+            if curr_logs_size + size > max_size:
+                split_logs.append(curr_logs)
+                curr_logs = []
+                curr_logs_size = 0
+            curr_logs.append(line)
+            curr_logs_size += size
+        split_logs.append(curr_logs)
+        return split_logs
+
+
 class _BaseBuildloggerHandler(handlers.BufferedHandler):
     """
     Base class of the buildlogger handler for the global logs and the
@@ -138,8 +181,8 @@ class _BaseBuildloggerHandler(handlers.BufferedHandler):
     """
 
     def __init__(self,
-                 build_id,
                  build_config,
+                 endpoint,
                  capacity=_SEND_AFTER_LINES,
                  interval_secs=_SEND_AFTER_SECS):
         """
@@ -157,8 +200,9 @@ class _BaseBuildloggerHandler(handlers.BufferedHandler):
                                                  username,
                                                  password)
 
-        self.build_id = build_id
+        self.endpoint = endpoint
         self.retry_buffer = []
+        self.max_size = None
 
     def process_record(self, record):
         """
@@ -177,12 +221,55 @@ class _BaseBuildloggerHandler(handlers.BufferedHandler):
         """
         Convenience method for subclasses to use when making POST requests.
         """
-
         return self.http_handler.post(*args, **kwargs)
 
     def _append_logs(self, log_lines):
-        raise NotImplementedError("_append_logs must be implemented by _BaseBuildloggerHandler"
-                                  " subclasses")
+        """
+        Sends a POST request to the handlers endpoint with the logs that have been captured.
+
+        Returns:
+            The number of log lines that have been successfully sent.
+        """
+        lines_sent = 0
+        for chunk in _LogsSplitter.split_logs(log_lines, self.max_size):
+            chunk_lines_sent = self.__append_logs_chunk(chunk)
+            lines_sent += chunk_lines_sent
+            if chunk_lines_sent < len(chunk):
+                # Not all lines have been sent. We stop here.
+                break
+        return lines_sent
+
+    def __append_logs_chunk(self, log_lines_chunk):
+        """
+        Sends a log lines chunk, handles 413 Request Entity Too Large errors and retries
+        if necessary.
+
+        Returns:
+            The number of log lines that have been successfully sent.
+        """
+        try:
+            self.post(self.endpoint, data=log_lines_chunk)
+            return len(log_lines_chunk)
+        except urllib2.HTTPError as err:
+            # Handle the "Request Entity Too Large" error, set the max size and retry.
+            if err.code == httplib.REQUEST_ENTITY_TOO_LARGE:
+                response_data = json.load(err, "utf-8")
+                if isinstance(response_data, dict) and "max_size" in response_data:
+                    new_max_size = response_data["max_size"]
+                    if self.max_size and new_max_size >= self.max_size:
+                        loggers._BUILDLOGGER_FALLBACK.exception(
+                            "Received an HTTP 413 code, but already had max_size set")
+                        return 0
+                    loggers._BUILDLOGGER_FALLBACK.warning(
+                        "Received an HTTP 413 code, updating the request max_size to %s",
+                        new_max_size)
+                    self.max_size = new_max_size
+                    return self._append_logs(log_lines_chunk)
+            loggers._BUILDLOGGER_FALLBACK.exception("Encountered an error.")
+            return 0
+        except:
+            loggers._BUILDLOGGER_FALLBACK.exception("Encountered an error.")
+            return 0
 
     def _flush_buffer_with_lock(self, buf, close_called):
         """
@@ -196,9 +283,10 @@ class _BaseBuildloggerHandler(handlers.BufferedHandler):
 
         self.retry_buffer.extend(buf)
 
-        if self._append_logs(self.retry_buffer):
-            self.retry_buffer = []
-        elif close_called:
+        nb_sent = self._append_logs(self.retry_buffer)
+        if nb_sent:
+            self.retry_buffer = self.retry_buffer[nb_sent:]
+        if close_called and self.retry_buffer:
             # Request to the buildlogger server returned an error, so use the fallback logger to
             # avoid losing the log messages entirely.
             for (_, message) in self.retry_buffer:
@@ -215,29 +303,14 @@ class BuildloggerTestHandler(_BaseBuildloggerHandler):
     Buildlogger handler for the test logs.
     """
 
-    def __init__(self, build_id, build_config, test_id, **kwargs):
-        """
-        Initializes the buildlogger handler with the build id, test id,
-        and credentials.
-        """
-
-        _BaseBuildloggerHandler.__init__(self, build_id, build_config, **kwargs)
-
-        self.test_id = test_id
-
-    @_log_on_error
-    def _append_logs(self, log_lines):
-        """
-        Sends a POST request to the APPEND_TEST_LOGS_ENDPOINT with the
-        logs that have been captured.
-        """
+    def __init__(self, build_config, build_id, test_id,
+                 capacity=_SEND_AFTER_LINES, interval_secs=_SEND_AFTER_SECS):
+        """Initializes the buildlogger handler with the credentials, build id, and test id."""
         endpoint = APPEND_TEST_LOGS_ENDPOINT % {
-            "build_id": self.build_id,
-            "test_id": self.test_id,
+            "build_id": build_id,
+            "test_id": test_id,
         }
-
-        response = self.post(endpoint, data=log_lines)
-        return response is not None
+        _BaseBuildloggerHandler.__init__(self, build_config, endpoint, capacity, interval_secs)
 
     @_log_on_error
     def _finish_test(self, failed=False):
@@ -245,12 +318,7 @@ class BuildloggerTestHandler(_BaseBuildloggerHandler):
         Sends a POST request to the APPEND_TEST_LOGS_ENDPOINT with the
         test status.
         """
-        endpoint = APPEND_TEST_LOGS_ENDPOINT % {
-            "build_id": self.build_id,
-            "test_id": self.test_id,
-        }
-
-        self.post(endpoint, headers={
+        self.post(self.endpoint, headers={
             "X-Sendlogs-Test-Done": "true",
             "X-Sendlogs-Test-Failed": "true" if failed else "false",
         })
@@ -271,12 +339,8 @@ class BuildloggerGlobalHandler(_BaseBuildloggerHandler):
     Buildlogger handler for the global logs.
     """
 
-    @_log_on_error
-    def _append_logs(self, log_lines):
-        """
-        Sends a POST request to the APPEND_GLOBAL_LOGS_ENDPOINT with
-        the logs that have been captured.
-        """
-        endpoint = APPEND_GLOBAL_LOGS_ENDPOINT % {"build_id": self.build_id}
-        response = self.post(endpoint, data=log_lines)
-        return response is not None
+    def __init__(self, build_config, build_id,
+                 capacity=_SEND_AFTER_LINES, interval_secs=_SEND_AFTER_SECS):
+        """Initializes the buildlogger handler with the credentials and build id."""
+        endpoint = APPEND_GLOBAL_LOGS_ENDPOINT % {"build_id": build_id}
+        _BaseBuildloggerHandler.__init__(self, build_config, endpoint, capacity, interval_secs)
diff --git a/buildscripts/resmokelib/logging/config.py b/buildscripts/resmokelib/logging/config.py
index c3960bbafd3..c3216b2d879 100644
--- a/buildscripts/resmokelib/logging/config.py
+++ b/buildscripts/resmokelib/logging/config.py
@@ -67,8 +67,8 @@ def apply_buildlogger_global_handler(logger, logging_config, build_id=None, buil
         log_format = logger_info.get("format", _DEFAULT_FORMAT)
         formatter = formatters.ISO8601Formatter(fmt=log_format)
 
-        handler = buildlogger.BuildloggerGlobalHandler(build_id,
-                                                       build_config,
+        handler = buildlogger.BuildloggerGlobalHandler(build_config,
+                                                       build_id,
                                                        **handler_info)
         handler.setFormatter(formatter)
     else:
@@ -98,8 +98,8 @@ def apply_buildlogger_test_handler(logger,
         log_format = logger_info.get("format", _DEFAULT_FORMAT)
         formatter = formatters.ISO8601Formatter(fmt=log_format)
 
-        handler = buildlogger.BuildloggerTestHandler(build_id,
-                                                     build_config,
+        handler = buildlogger.BuildloggerTestHandler(build_config,
+                                                     build_id,
                                                      test_id,
                                                      **handler_info)
         handler.setFormatter(formatter)
diff --git a/buildscripts/resmokelib/testing/fixtures/interface.py b/buildscripts/resmokelib/testing/fixtures/interface.py
index b4b0066a5aa..83418502ead 100644
--- a/buildscripts/resmokelib/testing/fixtures/interface.py
+++ b/buildscripts/resmokelib/testing/fixtures/interface.py
@@ -79,13 +79,22 @@ class Fixture(object):
         """
         return True
 
-    def get_connection_string(self):
+    def get_internal_connection_string(self):
         """
         Returns the connection string for this fixture. This is NOT a
         driver connection string, but a connection string of the format
         expected by the mongo::ConnectionString class.
         """
-        raise NotImplementedError("get_connection_string must be implemented by Fixture subclasses")
+        raise NotImplementedError(
+            "get_internal_connection_string must be implemented by Fixture subclasses")
+
+    def get_driver_connection_url(self):
+        """
+        Return the mongodb connection string as defined here:
+        https://docs.mongodb.com/manual/reference/connection-string/
+        """
+        raise NotImplementedError(
+            "get_driver_connection_url must be implemented by Fixture subclasses")
 
     def __str__(self):
         return "%s (Job #%d)" % (self.__class__.__name__, self.job_num)
diff --git a/buildscripts/resmokelib/testing/fixtures/masterslave.py b/buildscripts/resmokelib/testing/fixtures/masterslave.py
index 469c7ac0816..fb444cfe097 100644
--- a/buildscripts/resmokelib/testing/fixtures/masterslave.py
+++ b/buildscripts/resmokelib/testing/fixtures/masterslave.py
@@ -5,7 +5,6 @@ Master/slave fixture for executing JSTests against.
 from __future__ import absolute_import
 
 import os.path
-import socket
 
 import pymongo
 
@@ -160,6 +159,12 @@ class MasterSlaveFixture(interface.ReplFixture):
         mongod_options = self.mongod_options.copy()
         mongod_options.update(self.slave_options)
         mongod_options["slave"] = ""
-        mongod_options["source"] = "%s:%d" % (socket.gethostname(), self.port)
+        mongod_options["source"] = self.master.get_internal_connection_string()
         mongod_options["dbpath"] = os.path.join(self._dbpath_prefix, "slave")
         return self._new_mongod(mongod_logger, mongod_options)
+
+    def get_internal_connection_string(self):
+        return self.master.get_internal_connection_string()
+
+    def get_driver_connection_url(self):
+        return self.master.get_driver_connection_url()
diff --git a/buildscripts/resmokelib/testing/fixtures/replicaset.py b/buildscripts/resmokelib/testing/fixtures/replicaset.py
index 141bbd60ea1..0ec3de30280 100644
--- a/buildscripts/resmokelib/testing/fixtures/replicaset.py
+++ b/buildscripts/resmokelib/testing/fixtures/replicaset.py
@@ -37,7 +37,8 @@ class ReplicaSetFixture(interface.ReplFixture):
                  write_concern_majority_journal_default=None,
                  auth_options=None,
                  replset_config_options=None,
-                 voting_secondaries=True):
+                 voting_secondaries=False,
+                 use_replica_set_connection_string=False):
 
         interface.ReplFixture.__init__(self, logger, job_num)
 
@@ -50,6 +51,7 @@ class ReplicaSetFixture(interface.ReplFixture):
         self.auth_options = auth_options
         self.replset_config_options = utils.default_if_none(replset_config_options, {})
         self.voting_secondaries = voting_secondaries
+        self.use_replica_set_connection_string = use_replica_set_connection_string
 
         # The dbpath in mongod_options is used as the dbpath prefix for replica set members and
         # takes precedence over other settings. The ShardedClusterFixture uses this parameter to
@@ -97,7 +99,7 @@ class ReplicaSetFixture(interface.ReplFixture):
         # Initiate the replica set.
         members = []
         for (i, node) in enumerate(self.nodes):
-            member_info = {"_id": i, "host": node.get_connection_string()}
+            member_info = {"_id": i, "host": node.get_internal_connection_string()}
             if i > 0:
                 member_info["priority"] = 0
                 if i >= 7 or not self.voting_secondaries:
@@ -107,7 +109,7 @@ class ReplicaSetFixture(interface.ReplFixture):
             members.append(member_info)
         if self.initial_sync_node:
             members.append({"_id": self.initial_sync_node_idx,
-                            "host": self.initial_sync_node.get_connection_string(),
+                            "host": self.initial_sync_node.get_internal_connection_string(),
                             "priority": 0,
                             "hidden": 1,
                             "votes": 0})
@@ -291,11 +293,27 @@ class ReplicaSetFixture(interface.ReplFixture):
 
         return logging.loggers.new_logger(logger_name, parent=self.logger)
 
-    def get_connection_string(self):
+    def get_internal_connection_string(self):
         if self.replset_name is None:
-            raise ValueError("Must call setup() before calling get_connection_string()")
+            raise ValueError("Must call setup() before calling get_internal_connection_string()")
 
-        conn_strs = [node.get_connection_string() for node in self.nodes]
+        conn_strs = [node.get_internal_connection_string() for node in self.nodes]
         if self.initial_sync_node:
-            conn_strs.append(self.initial_sync_node.get_connection_string())
+            conn_strs.append(self.initial_sync_node.get_internal_connection_string())
         return self.replset_name + "/" + ",".join(conn_strs)
+
+    def get_driver_connection_url(self):
+        if self.replset_name is None:
+            raise ValueError("Must call setup() before calling get_driver_connection_url()")
+
+        if self.use_replica_set_connection_string:
+            # We use a replica set connection string when all nodes are electable because we
+            # anticipate the client will want to gracefully handle any failovers.
+            conn_strs = [node.get_internal_connection_string() for node in self.nodes]
+            if self.initial_sync_node:
+                conn_strs.append(self.initial_sync_node.get_internal_connection_string())
+            return "mongodb://" + ",".join(conn_strs) + "/?replicaSet=" + self.replset_name
+        else:
+            # We return a direct connection to the expected pimary when only the first node is
+            # electable because we want the client to error out if a stepdown occurs.
+            return self.nodes[0].get_driver_connection_url()
diff --git a/buildscripts/resmokelib/testing/fixtures/shardedcluster.py b/buildscripts/resmokelib/testing/fixtures/shardedcluster.py
index ac7e597f24b..2e2db535d6d 100644
--- a/buildscripts/resmokelib/testing/fixtures/shardedcluster.py
+++ b/buildscripts/resmokelib/testing/fixtures/shardedcluster.py
@@ -170,11 +170,14 @@ class ShardedClusterFixture(interface.Fixture):
                 all(shard.is_running() for shard in self.shards) and
                 self.mongos is not None and self.mongos.is_running())
 
-    def get_connection_string(self):
+    def get_internal_connection_string(self):
         if self.mongos is None:
-            raise ValueError("Must call setup() before calling get_connection_string()")
+            raise ValueError("Must call setup() before calling get_internal_connection_string()")
 
-        return "%s:%d" % (socket.gethostname(), self.mongos.port)
+        return self.mongos.get_internal_connection_string()
+
+    def get_driver_connection_url(self):
+        return "mongodb://" + self.get_internal_connection_string()
 
     def _new_configsvr(self):
         """
@@ -229,16 +232,11 @@ class ShardedClusterFixture(interface.Fixture):
         mongos_logger = logging.loggers.new_logger(logger_name, parent=self.logger)
 
         mongos_options = copy.deepcopy(self.mongos_options)
-        configdb_hostname = socket.gethostname()
 
         if self.separate_configsvr:
-            configdb_replset = ShardedClusterFixture._CONFIGSVR_REPLSET_NAME
-            configdb_port = self.configsvr.port
-            mongos_options["configdb"] = "%s/%s:%d" % (configdb_replset,
-                                                       configdb_hostname,
-                                                       configdb_port)
+            mongos_options["configdb"] = self.configsvr.get_internal_connection_string()
         else:
-            mongos_options["configdb"] = "%s:%d" % (configdb_hostname, self.shards[0].port)
+            mongos_options["configdb"] = "localhost:%d" % (self.shards[0].port)
 
         return _MongoSFixture(mongos_logger,
                               self.job_num,
@@ -254,9 +252,9 @@ class ShardedClusterFixture(interface.Fixture):
         for more details.
         """
 
-        hostname = socket.gethostname()
-        self.logger.info("Adding %s:%d as a shard..." % (hostname, shard.port))
-        client.admin.command({"addShard": "%s:%d" % (hostname, shard.port)})
+        connection_string = shard.get_internal_connection_string()
+        self.logger.info("Adding %s as a shard...", connection_string)
+        client.admin.command({"addShard": connection_string})
 
 
 class _MongoSFixture(interface.Fixture):
@@ -356,3 +354,12 @@ class _MongoSFixture(interface.Fixture):
 
     def is_running(self):
         return self.mongos is not None and self.mongos.poll() is None
+
+    def get_internal_connection_string(self):
+        if self.mongos is None:
+            raise ValueError("Must call setup() before calling get_internal_connection_string()")
+
+        return "localhost:%d" % self.port
+
+    def get_driver_connection_url(self):
+        return "mongodb://" + self.get_internal_connection_string()
diff --git a/buildscripts/resmokelib/testing/fixtures/standalone.py b/buildscripts/resmokelib/testing/fixtures/standalone.py
index ba62b3d2b8c..bc69775c285 100644
--- a/buildscripts/resmokelib/testing/fixtures/standalone.py
+++ b/buildscripts/resmokelib/testing/fixtures/standalone.py
@@ -7,7 +7,6 @@ from __future__ import absolute_import
 import os
 import os.path
 import shutil
-import socket
 import time
 
 import pymongo
@@ -146,8 +145,11 @@ class MongoDFixture(interface.Fixture):
     def is_running(self):
         return self.mongod is not None and self.mongod.poll() is None
 
-    def get_connection_string(self):
+    def get_internal_connection_string(self):
         if self.mongod is None:
-            raise ValueError("Must call setup() before calling get_connection_string()")
+            raise ValueError("Must call setup() before calling get_internal_connection_string()")
 
-        return "%s:%d" % (socket.gethostname(), self.port)
+        return "localhost:%d" % self.port
+
+    def get_driver_connection_url(self):
+        return "mongodb://" + self.get_internal_connection_string()
diff --git a/buildscripts/resmokelib/testing/testcases.py b/buildscripts/resmokelib/testing/testcases.py
index b4029fc6ea8..21d35215a29 100644
--- a/buildscripts/resmokelib/testing/testcases.py
+++ b/buildscripts/resmokelib/testing/testcases.py
@@ -187,7 +187,7 @@ class CPPIntegrationTestCase(TestCase):
     def configure(self, fixture, *args, **kwargs):
         TestCase.configure(self, fixture, *args, **kwargs)
 
-        self.program_options["connectionString"] = self.fixture.get_connection_string()
+        self.program_options["connectionString"] = self.fixture.get_internal_connection_string()
 
     def run_test(self):
         try:
@@ -362,6 +362,19 @@ class JSTestCase(TestCase):
             # Directory already exists.
             pass
 
+        process_kwargs = self.shell_options.get("process_kwargs", {}).copy()
+
+        if "KRB5_CONFIG" in process_kwargs and "KRB5CCNAME" not in process_kwargs:
+            # Use a job-specific credential cache for JavaScript tests involving Kerberos.
+            krb5_dir = os.path.join(data_dir, "krb5")
+            try:
+                os.makedirs(krb5_dir)
+            except os.error:
+                pass
+            process_kwargs["KRB5CCNAME"] = "DIR:" + os.path.join(krb5_dir, ".")
+
+        self.shell_options["process_kwargs"] = process_kwargs
+
     def _get_data_dir(self, global_vars):
         """
         Returns the value that the mongo shell should set for the
@@ -408,11 +421,13 @@ class JSTestCase(TestCase):
         is_main_test = True
         if thread_id > 0:
             is_main_test = False
-        return core.programs.mongo_shell_program(logger,
-                                                 executable=self.shell_executable,
-                                                 filename=self.js_filename,
-                                                 isMainTest=is_main_test,
-                                                 **self.shell_options)
+        return core.programs.mongo_shell_program(
+            logger,
+            executable=self.shell_executable,
+            filename=self.js_filename,
+            connection_string=self.fixture.get_driver_connection_url(),
+            isMainTest=is_main_test,
+            **self.shell_options)
 
     def _run_test_in_thread(self, thread_id):
         # Make a logger for each thread.
diff --git a/buildscripts/scons.py b/buildscripts/scons.py
index e7860162eb6..b0b9cfa834e 100755
--- a/buildscripts/scons.py
+++ b/buildscripts/scons.py
@@ -1,16 +1,26 @@
-#!/usr/bin/python
+#!/usr/bin/env python2
+
+from __future__ import print_function
 
 import os
-import subprocess
 import sys
 
 SCONS_VERSION = os.environ.get('SCONS_VERSION', "2.5.0")
 
-mongodb_root = os.path.dirname(os.path.dirname(__file__))
-scons_dir = os.path.join(mongodb_root, 'src', 'third_party', 'scons-' + SCONS_VERSION)
+mongodb_root = os.path.abspath(os.path.dirname(os.path.dirname(__file__)))
+scons_dir = os.path.join(mongodb_root, 'src', 'third_party','scons-' + SCONS_VERSION,
+                         'scons-local-' + SCONS_VERSION)
+
+if not os.path.exists(scons_dir):
+    print("Could not find SCons in '%s'" % (scons_dir))
+    sys.exit(1)
+
+sys.path = [scons_dir] + sys.path
+
+try:
+    import SCons.Script
+except ImportError:
+    print("Could not find SCons in '%s'" % (scons_dir))
+    sys.exit(1)
 
-if sys.platform == 'win32':
-    args = [sys.executable, os.path.join(scons_dir, 'scons.py')] + sys.argv[1:]
-    sys.exit(subprocess.call(args))
-else:
-    os.execv(os.path.join(scons_dir, 'scons.py'), sys.argv)
+SCons.Script.main()
diff --git a/buildscripts/tests/resmokelib/logging/test_buildlogger.py b/buildscripts/tests/resmokelib/logging/test_buildlogger.py
new file mode 100644
index 00000000000..7bc705948db
--- /dev/null
+++ b/buildscripts/tests/resmokelib/logging/test_buildlogger.py
@@ -0,0 +1,67 @@
+"""Unit tests for the buildscripts.resmokelib.logging.buildlogger module."""
+
+from __future__ import absolute_import
+
+import json
+import unittest
+
+from buildscripts.resmokelib.logging import buildlogger
+
+
+class TestLogsSplitter(unittest.TestCase):
+    """Unit tests for the _LogsSplitter class."""
+
+    def test_split_no_logs(self):
+        logs = []
+        max_size = 10
+        self.assertEqual([[]], buildlogger._LogsSplitter.split_logs(logs, max_size))
+
+    def test_split_no_max_size(self):
+        logs = self.__generate_logs(size=30)
+        max_size = None
+        self.assertEqual([logs], buildlogger._LogsSplitter.split_logs(logs, max_size))
+
+    def test_split_max_size_smaller(self):
+        logs = self.__generate_logs(size=20)
+        max_size = 30
+        self.assertEqual([logs], buildlogger._LogsSplitter.split_logs(logs, max_size))
+
+    def test_split_max_size_equal(self):
+        logs = self.__generate_logs(size=30)
+        max_size = 30
+        self.assertEqual([logs], buildlogger._LogsSplitter.split_logs(logs, max_size))
+
+    def test_split_max_size_larger(self):
+        logs = self.__generate_logs(size=31)
+        max_size = 30
+        self.assertEqual(
+            [logs[0:-1], logs[-1:]],
+            buildlogger._LogsSplitter.split_logs(logs, max_size))
+
+        logs = self.__generate_logs(size=149)
+        max_size = 19
+        self.assertEqual(
+            [logs[0:3], logs[3:6], logs[6:9], logs[9:12], logs[12:15],
+             logs[15:18], logs[18:21], logs[21:24], logs[24:27], logs[27:]],
+            buildlogger._LogsSplitter.split_logs(logs, max_size))
+
+    def check_split_sizes(self, splits, max_size):
+        for split in splits:
+            self.assertTrue(TestLogsSplitter.size(split) <= max_size)
+
+    def __generate_logs(self, size):
+        # The size of [ "x" ] is 5. This is the minimum size we generate.
+        self.assertTrue(size >= 5)
+        # Each new "x" adds 5 to the size.
+        nb_lines = size / 5
+        # Each additional "x" on a line adds 1 to the size.
+        last_line_extra = size % 5
+        logs = ["x"] * nb_lines
+        logs[-1] += "x" * last_line_extra
+        self.assertEqual(size, TestLogsSplitter.size(logs))
+        return logs
+
+    @staticmethod
+    def size(logs):
+        """Returns the size of the log lines when represented in JSON."""
+        return len(json.dumps(logs, encoding="utf-8"))
diff --git a/distsrc/README b/distsrc/README
index 8cc83742981..5a772b66e7d 100644
--- a/distsrc/README
+++ b/distsrc/README
@@ -1,22 +1,30 @@
 MongoDB README

 

-Welcome to MongoDB!
-
+Welcome to MongoDB 3.4!

+

 COMPONENTS

 

-  bin/mongod - The database process.

-  bin/mongos - Sharding controller.

-  bin/mongo  - The database shell (uses interactive javascript).

-
-UTILITIES
-
-  bin/mongodump         - MongoDB dump tool - for backups, snapshots, etc..
-  bin/mongorestore      - MongoDB restore a dump
-  bin/mongoexport       - Export a single collection to test (JSON, CSV)
-  bin/mongoimport       - Import from JSON or CSV
-  bin/mongofiles        - Utility for putting and getting files from MongoDB GridFS
-  bin/mongostat         - Show performance statistics
-
+  mongod - The database server.

+  mongos - Sharding router.

+  mongo  - The database shell (uses interactive javascript).

+

+UTILITIES

+

+  mongodump         - Create a binary dump of the contents of a database.

+  mongorestore      - Restore data from the output created by mongodump.

+  mongoexport       - Export the contents of a collection to JSON or CSV.

+  mongoimport       - Import data from JSON, CSV or TSV.

+  mongofiles        - Put, get and delete files from GridFS.

+  mongostat         - Show the status of a running mongod/mongos.

+  bsondump          - Convert BSON files into human-readable formats.

+  mongooplog        - Poll the oplog and apply to a local server.

+  mongoreplay       - Traffic capture and replay tool.

+  mongotop          - Track time spent reading and writing data.

+

+BUILDING

+

+  See docs/building.md.

+

 RUNNING

 

   For command line options invoke:

@@ -25,28 +33,52 @@ RUNNING
 

   To run a single server database:

 

-    $ mkdir /data/db

+    $ sudo mkdir -p /data/db

     $ ./mongod

     $

     $ # The mongo javascript shell connects to localhost and test database by default:

-    $ ./mongo 

+    $ ./mongo

     > help

 

 DRIVERS

 

-  Client drivers for most programming languages are available at mongodb.org.  Use the 
-  shell ("mongo") for administrative tasks.

+  Client drivers for most programming languages are available at

+  https://docs.mongodb.com/manual/applications/drivers/. Use the shell

+  ("mongo") for administrative tasks.

+

+BUG REPORTS

+

+  See https://github.com/mongodb/mongo/wiki/Submit-Bug-Reports.

+

+PACKAGING

+

+  Packages are created dynamically by the package.py script located in the

+  buildscripts directory. This will generate RPM and Debian packages.

 

 DOCUMENTATION

 

-  http://www.mongodb.org/
-
-MAIL LISTS AND IRC
-
-  http://dochub.mongodb.org/core/community

+  https://docs.mongodb.com/manual/

+

+CLOUD HOSTED MONGODB

+

+  https://www.mongodb.com/cloud/atlas

+

+MAIL LISTS

+

+  https://groups.google.com/forum/#!forum/mongodb-user

+

+    A forum for technical questions about using MongoDB.

+

+  https://groups.google.com/forum/#!forum/mongodb-dev

+

+    A forum for technical questions about building and developing MongoDB.

+

+LEARN MONGODB

+

+  https://university.mongodb.com/

 

-32 BIT BUILD NOTES

+LICENSE

 

-  MongoDB uses memory mapped files.  If built as a 32 bit executable, you will

-  not be able to work with large (multi-gigabyte) databases.  However, 32 bit

-  builds work fine with small development databases.

+  Most MongoDB source files (src/mongo folder and below) are made available

+  under the terms of the GNU Affero General Public License (GNU AGPLv3). See

+  individual files for details.

diff --git a/docs/building.md b/docs/building.md
index 6008a9b606f..48f1a188396 100644
--- a/docs/building.md
+++ b/docs/building.md
@@ -4,7 +4,7 @@ Building MongoDB
 To build MongoDB, you will need:
 
 * A modern C++ compiler. One of the following is required.
-    * GCC 4.8.2 or newer
+    * GCC 5.3.0 or newer
     * Clang 3.4 (or Apple XCode 5.1.1 Clang) or newer
     * Visual Studio 2013 Update 2 or newer
 * Python 2.7
diff --git a/etc/drivers_nightly.yml b/etc/drivers_nightly.yml
index f039112ea85..464f844ac57 100644
--- a/etc/drivers_nightly.yml
+++ b/etc/drivers_nightly.yml
@@ -3,7 +3,7 @@
 #####################################################
 
 ### Instructions for Adding a Driver ###
-# The Major sections are:
+## The Major sections are:
 
 # VARIABLES
 # FUNCTIONS (MongoDB and Drivers)
@@ -218,8 +218,9 @@ functions:
     - command: shell.exec
       type: test
       params:
-        working_dir: "src/${DRIVER_WORKING_DIRECTORY}"
+        working_dir: src
         script: |
+          cd ${DRIVER_WORKING_DIRECTORY}
           ${PREPARE_SHELL}
           echo $DRIVERS_TOOLS
           echo $MONGO_ORCHESTRATION_HOME
@@ -318,8 +319,9 @@ functions:
   "setup drivers environment":
     - command: shell.exec
       params:
-        working_dir: "src/${DRIVER_WORKING_DIRECTORY}"
+        working_dir: src
         script: |
+           cd "${DRIVER_WORKING_DIRECTORY}"
            # Get the current unique version of this checkout
            if [ "${is_patch}" = "true" ]; then
               CURRENT_VERSION=$(git describe)-patch-${version_id}
@@ -433,10 +435,7 @@ tasks:
           if [ `which strip` ]; then
             echo "found strip"
             find . -maxdepth 1 -type f -iname "mongo*" -exec strip {} \;
-            strip mongo/mongo
           fi
-          # On RHEL, the mongo shell binary is in a mongo directory
-          mv mongo/mongo ./mongodb-binaries/bin/ || true
           find . -maxdepth 1 -type f -iname "mongo*" -exec mv {} ./mongodb-binaries/bin/ \;
           ${compress} mongodb-binaries.${ext|tgz} mongodb-binaries/
 
@@ -584,7 +583,6 @@ axes:
         display_name: "Python 2.6"
         variables:
            ### There is no python toolchain on RHEL 6.2, using system python
-           ### PYTHON_BINARY: "/opt/python/2.6/bin/python"
            PYTHON_BINARY: "/usr/bin/python"
            DRIVER_WORKING_DIRECTORY: "mongo-python-driver"
            DRIVER_TEST_UPLOAD_DIRECTORY: "src/mongo-python-driver/xunit-results/TEST-*.xml"
@@ -629,7 +627,7 @@ modules:
 
 - name: mongo-python-driver
   repo: git@github.com:mongodb/mongo-python-driver.git
-  branch: evergreen
+  branch: master
 
 #######################################
 #            BUILDVARIANTS            #
diff --git a/etc/evergreen.yml b/etc/evergreen.yml
index fae0563fbde..520a49b77f5 100644
--- a/etc/evergreen.yml
+++ b/etc/evergreen.yml
@@ -113,7 +113,7 @@ variables:
   - enterprise-debian71-64
   - enterprise-debian81-64
   - enterprise-linux-64-amazon-ami
-  - enterprise-osx-107
+  - enterprise-osx-108
   - enterprise-rhel-62-64-bit
   - enterprise-rhel-62-64-bit-coverage
   - enterprise-rhel-62-64-bit-inmem
@@ -145,17 +145,15 @@ variables:
   - linux-64-ephemeralForTest
   - linux-64-lsm
   - linux-64-repeated-execution
-  - osx-107
-  - osx-107-debug
-  - osx-107-ssl
+  - osx-108
+  - osx-108-debug
+  - osx-108-ssl
   - rhel62
   - rhel70
-  - solaris-64-bit
   - suse11
   - suse12
   - ubuntu1204
   - ubuntu1404
-  - ubuntu1404-rocksdb
   - ubuntu1604
   - ubuntu1604-arm64
   - ubuntu1604-asan
@@ -238,9 +236,6 @@ functions:
         if [ "Windows_NT" = "$OS" ]; then
             typeperf -qx PhysicalDisk | grep Disk | grep -v _Total > disk_counters.txt
             typeperf -cf disk_counters.txt -si 5 -o mongo-diskstats
-        # Solaris: iostat -T d option for timestamp.
-        elif iostat -T d -Mx > /dev/null 2>&1; then
-            iostat -T d -Mx 5 > mongo-diskstats
         # Linux: iostat -t option for timestamp.
         elif iostat -tdmx > /dev/null 2>&1; then
             iostat -tdmx 5 > mongo-diskstats
@@ -285,11 +280,8 @@ functions:
           get_pids() { proc_pids=$(pgrep $1); }
           get_process_info() {
             proc_name=$(ps -p $1 -o comm=);
-            # prstat is available on Solaris
-            if [ ! -z $(which prstat 2> /dev/null) ]; then
-              proc_threads=$(prstat -p $1 1 1 | grep $1 | cut -f2 -d "/");
             # /proc is available on Linux platforms
-            elif [ -f /proc/$1/status ]; then
+            if [ -f /proc/$1/status ]; then
               ${set_sudo}
               proc_threads=$($sudo grep Threads /proc/$1/status | sed "s/\s//g" | cut -f2 -d ":");
             else
@@ -365,19 +357,6 @@ functions:
         mv mongodb*/mongo.{debug,dSYM,pdb} mongodb*/mongod.{debug,dSYM,pdb} mongodb*/mongos.{debug,dSYM,pdb} . 2>/dev/null || true
         rm -r mongodb*/*.{debug,dSYM,pdb} mongo-debugsymbols.tgz 2>/dev/null || true
 
-  "build rocksdb" :
-    command: shell.exec
-    params:
-      script: |
-        set -o errexit
-        set -o verbose
-        if [ "${build_rocksdb|}" = "true" ]; then
-        rm -rf rocksdb
-        git clone https://github.com/facebook/rocksdb.git
-        cd rocksdb
-        make CC=/opt/mongodbtoolchain/v2/bin/gcc CXX=/opt/mongodbtoolchain/v2/bin/g++ static_lib
-        fi
-
   "build new tools" :
     command: shell.exec
     params:
@@ -441,6 +420,13 @@ functions:
         set -o errexit
         set -o verbose
 
+        # Set the TMPDIR environment variable to be a directory in the task's working
+        # directory so that temporary files created by processes spawned by resmoke.py get
+        # cleaned up after the task completes. This also ensures the spawned processes
+        # aren't impacted by limited space in the mount point for the /tmp directory.
+        export TMPDIR="${workdir}/tmp"
+        mkdir -p $TMPDIR
+
         # check if virtualenv is set up
         if [ -d "venv" ]; then
           if [ "Windows_NT" = "$OS" ]; then
@@ -524,7 +510,7 @@ functions:
           path_value="$path_value:${task_path_suffix}"
         fi
 
-        ${resmoke_wrapper} PATH="$path_value" ${san_symbolizer} ${san_options} ${rlp_environment} ${python|/opt/mongodbtoolchain/v2/bin/python2} buildscripts/resmoke.py ${resmoke_args} $extra_args ${test_flags} --log=buildlogger --reportFile=report.json
+        ${resmoke_wrapper} PATH="$path_value" ${san_symbolizer} ${lang_environment} ${san_options} ${rlp_environment} ${python|/opt/mongodbtoolchain/v2/bin/python2} buildscripts/resmoke.py ${resmoke_args} $extra_args ${test_flags} --log=buildlogger --reportFile=report.json
 
 
   "do jepsen setup" :
@@ -852,10 +838,18 @@ functions:
       script: |
         set -o errexit
 
+        # Override the aws credentials with the kitchen specific credentials
+        cat <<EOF > ~/.aws/credentials
+        [default]
+        aws_access_key_id = ${kitchen_aws_key}
+        aws_secret_access_key = ${kitchen_aws_secret}
+        EOF
+
         export KITCHEN_ARTIFACTS_URL="https://s3.amazonaws.com/mciuploads/${project}/${build_variant}/${revision}/artifacts/${build_id}.tgz"
         export KITCHEN_SECURITY_GROUP="${kitchen_security_group}"
         export KITCHEN_SSH_KEY_ID="${kitchen_ssh_key_id}"
         export KITCHEN_SUBNET="${kitchen_subnet}"
+        export KITCHEN_VPC="${kitchen_vpc}"
 
         for i in {1..3}
         do
@@ -1125,6 +1119,12 @@ post:
             if [ ! -f $new_binary_file ]; then
               mv $binary_file_location $new_binary_file
             fi
+            # On Windows if a .pdb symbol file exists, include it in the archive.
+            pdb_file=$(echo $binary_file_location | sed "s/\.exe/.pdb/")
+            if [ -f $pdb_file ]; then
+              new_pdb_file=unittest_binaries/$(echo $pdb_file | sed "s/.*\///")
+              mv $pdb_file $new_pdb_file
+            fi
           done
         done
   - command: archive.targz_pack
@@ -1239,14 +1239,12 @@ tasks:
         directory: src
         revisions: # for each module include revision as <module_name> : ${<module_name>_rev}
           enterprise: ${enterprise_rev}
-          rocksdb: ${rocksdb_rev}
     - command: git.apply_patch
       params:
         directory: src
     - func: "get buildnumber"
     - func: "setup credentials"
     - func: "build new tools" # noop if ${newtools} is not "true"
-    - func: "build rocksdb" # noop if ${build_rocksdb} is not "true"
     - command: shell.exec
       params:
         working_dir: src
@@ -1295,14 +1293,6 @@ tasks:
             ${strip_command|/usr/bin/strip} mongobridge
           fi
 
-          # If this is a scheduled build, we check for changes against the last scheduled commit.
-          if [ "${is_patch}" != "true" ]; then
-            burn_in_args="--checkEvergreen"
-          fi
-
-          # Capture a list of new and modified tests.
-          ${python|/opt/mongodbtoolchain/v2/bin/python2} buildscripts/burn_in_tests.py --branch=${branch_name} --buildVariant=${build_variant} --testListOutfile=jstests/new_tests.json --noExec $burn_in_args
-
           # On windows we need to make sure the paths in unittests.txt are compatible with cygwin tar
           sed 's|\\|/|g' build/unittests.txt > build/unittests-tarlist.txt
           ${tar|tar} -czvf mongodb-unittests.tgz -T build/unittests-tarlist.txt build/unittests.txt
@@ -1421,7 +1411,6 @@ tasks:
         directory: src
         revisions: # for each module include revision as <module_name> : ${<module_name>_rev}
           enterprise: ${enterprise_rev}
-          rocksdb: ${rocksdb_rev}
     - command: git.apply_patch
       params:
         directory: src
@@ -1440,7 +1429,38 @@ tasks:
   depends_on:
   - name: compile
   commands:
+  - command: git.get_project
+    # The repository is cloned in a directory distinct from src for the modified test detection
+    # because the extraction of the artifacts performed in the 'do setup' causes
+    # 'git diff --name-only' to see all tests as modified on Windows (git 1.9.5). See SERVER-30634.
+    params:
+      directory: burn_in_tests_clonedir
+      revisions: # for each module include revision as <module_name> : ${<module_name>_rev}
+        enterprise: ${enterprise_rev}
+  - command: shell.exec
   - func: "do setup"
+  - func: "set up virtualenv"
+  - command: shell.exec
+    params:
+      working_dir: burn_in_tests_clonedir
+      script: |
+        set -o errexit
+        # Create a symbolic link to the venv in the src directory so activate_virtualenv can use it
+        # if it exists.
+        ln -s ../src/venv venv
+        ${activate_virtualenv}
+        set -o verbose
+        # If this is a scheduled build, we check for changes against the last scheduled commit.
+        if [ "${is_patch}" != "true" ]; then
+          burn_in_args="--checkEvergreen"
+        fi
+        # Copy the dbtest executable from the src dir because burn_in_tests.py calls it to get the
+        # list of dbtest suites.
+        cp ../src/dbtest${exe} .
+        # Capture a list of new and modified tests.
+        ${python|/opt/mongodbtoolchain/v2/bin/python2} buildscripts/burn_in_tests.py --branch=${branch_name} --buildVariant=${build_variant} --testListOutfile=jstests/new_tests.json --noExec $burn_in_args
+        # Copy the results to the src dir.
+        cp jstests/new_tests.json ../src/jstests/new_tests.json
   - func: "do multiversion setup"
   - func: "run tests"
     vars:
@@ -2717,6 +2737,17 @@ tasks:
       run_multiple_jobs: true
 
 - <<: *task_template
+  name: write_concern_majority_passthrough_WT
+  depends_on:
+  - name: jsCore_WT
+  commands:
+  - func: "do setup"
+  - func: "run tests"
+    vars:
+      resmoke_args: --suites=write_concern_majority_passthrough --storageEngine=wiredTiger
+      run_multiple_jobs: true
+
+- <<: *task_template
   name: replica_sets
   commands:
   - func: "do setup"
@@ -3097,7 +3128,7 @@ tasks:
           /usr/local/bin/notary-client.py --key-name "server-3.4" --auth-token-file ${workdir}/src/signing_auth_token --comment "Evergreen Automatic Signing ${revision} - ${build_variant} - ${branch_name}" --notary-url http://notary-service.build.10gen.cc:5000 --skip-missing mongodb-${push_name}-${push_arch}-${suffix}.${ext|tgz} mongodb-shell-${push_name}-${push_arch}-${suffix}.${ext|tgz} mongodb-${push_name}-${push_arch}-debugsymbols-${suffix}.${ext|tgz} mongodb-win32-${push_arch}-${suffix}.msi mongodb-src-${src_suffix}.${ext|tar.gz}
 
           if [ "${has_packages|}" = "true" ]; then
-             CURATOR_RELEASE="88f34a9f1c79db7ea9597b6e85eb5995d03714ba"
+             CURATOR_RELEASE="ea8d75dcc1a587111e7418e2428fb67e267af9fe"
              curl -L -O http://boxes.10gen.com/build/curator/curator-dist-rhel70-$CURATOR_RELEASE.tar.gz
              tar -zxvf curator-dist-rhel70-$CURATOR_RELEASE.tar.gz
              ./curator repo --config ./etc/repo_config.yaml --distro ${packager_distro} --edition ${repo_edition} --version ${version} --arch ${packager_arch} --packages repo
@@ -3513,11 +3544,6 @@ modules:
   prefix: src/mongo/db/modules
   branch: v3.4
 
-- name: rocksdb
-  repo: git@github.com:mongodb-partners/mongo-rocks.git
-  prefix: src/mongo/db/modules
-  branch: v3.4
-
 #######################################
 #            Buildvariants            #
 #######################################
@@ -3650,6 +3676,7 @@ buildvariants:
   - name: tool
   - name: tool_WT
   - name: unittests
+  - name: write_concern_majority_passthrough_WT
   - name: push
 
 - name: linux-64-repeated-execution
@@ -3767,6 +3794,7 @@ buildvariants:
   - name: tool
   - name: tool_WT
   - name: unittests
+  - name: write_concern_majority_passthrough_WT
 
 - name: linux-64-duroff
   display_name: Linux (No Journal)
@@ -4159,6 +4187,7 @@ buildvariants:
     push_bucket: downloads.mongodb.org
     push_name: linux
     push_arch: x86_64-ubuntu1604
+    lang_environment: LANG=C
     compile_flags: --ssl MONGO_DISTMOD=ubuntu1604 -j$(grep -c ^processor /proc/cpuinfo) --release CC=/opt/mongodbtoolchain/v2/bin/gcc CXX=/opt/mongodbtoolchain/v2/bin/g++ OBJCOPY=/opt/mongodbtoolchain/v2/bin/objcopy
     multiversion_platform_arch: "ubuntu1604"
     multiversion_edition: "targeted"
@@ -4283,6 +4312,7 @@ buildvariants:
     push_arch: arm64-enterprise-ubuntu1604
     compile_flags: --ssl MONGO_DISTMOD=ubuntu1604 -j$(grep -c ^processor /proc/cpuinfo) CCFLAGS="-march=armv8-a+crc -mtune=generic" --release CC=/opt/mongodbtoolchain/v2/bin/gcc CXX=/opt/mongodbtoolchain/v2/bin/g++ OBJCOPY=/opt/mongodbtoolchain/v2/bin/objcopy
     num_jobs_available: $(grep -c ^processor /proc/cpuinfo)
+    max_jobs: 8 # Avoid starting too many mongod's on ARM test servers
     variant_excluded_flags: requires_mmapv1
     has_packages: true
     packager_script: packager-enterprise.py
@@ -4370,6 +4400,7 @@ buildvariants:
     push_arch: arm64-ubuntu1604
     compile_flags: --ssl MONGO_DISTMOD=ubuntu1604 -j$(grep -c ^processor /proc/cpuinfo) --release CCFLAGS="-march=armv8-a+crc -mtune=generic" CC=/opt/mongodbtoolchain/v2/bin/gcc CXX=/opt/mongodbtoolchain/v2/bin/g++ OBJCOPY=/opt/mongodbtoolchain/v2/bin/objcopy
     num_jobs_available: $(grep -c ^processor /proc/cpuinfo)
+    max_jobs: 8 # Avoid starting too many mongod's on ARM test servers
     variant_excluded_flags: requires_mmapv1
     has_packages: true
     packager_script: packager.py
@@ -4381,62 +4412,8 @@ buildvariants:
   - name: compile
     distros:
     - ubuntu1604-arm64-large
-  - name: aggregation_WT
-  - name: aggregation_auth
-  - name: auth_WT
   - name: dbtest_WT
-  - name: failpoints
-  - name: failpoints_auth
-  - name: gle_auth_WT
-  - name: gle_auth_write_cmd_WT
-  - name: gle_auth_basics_passthrough_WT
-  - name: gle_auth_basics_passthrough_write_cmd_WT
-  - name: sharding_gle_auth_basics_passthrough_WT
-  - name: sharding_gle_auth_basics_passthrough_write_cmd_WT
   - name: jsCore_WT
-  - name: jsCore_compatibility_WT
-  - name: jsCore_decimal_WT
-  - name: jstestfuzz_WT
-  - name: jstestfuzz_concurrent_WT
-  - name: jstestfuzz_concurrent_replication_WT
-  - name: jstestfuzz_concurrent_sharded_WT
-  - name: jstestfuzz_replication_WT
-  - name: jstestfuzz_sharded_WT
-  - name: mongosTest
-  - name: noPassthrough_WT
-  - name: noPassthroughWithMongod_WT
-  - name: bulk_gle_passthrough_WT
-  - name: parallel_WT
-  - name: parallel_compatibility_WT
-  - name: concurrency_WT
-    distros:
-    # There is a known performance issue with the ubuntu1604-arm64-1.linaro.build.10gen.cc host
-    # that's part of the ubuntu1604-arm64-large pool. We run the concurrency tests only on the
-    # hosts from the ubuntu1604-arm64-small pool to work around this issue.
-    - ubuntu1604-arm64-small
-  - name: concurrency_replication_WT
-    distros:
-    - ubuntu1604-arm64-small
-  - name: concurrency_sharded_WT
-    distros:
-    - ubuntu1604-arm64-small
-  - name: concurrency_simultaneous_WT
-    distros:
-    - ubuntu1604-arm64-small
-  - name: replica_sets_WT
-  - name: replica_sets_auth
-  - name: replica_sets_jscore_passthrough_WT
-  - name: master_slave_WT
-  - name: master_slave_auth
-  - name: master_slave_jscore_passthrough_WT
-  - name: sharding_WT
-  - name: sharding_auth
-  - name: slow1_WT
-  - name: serial_run_WT
-  - name: sharding_jscore_passthrough_WT
-  - name: ssl
-  - name: sslSpecial
-  - name: tool_WT
   - name: unittests
   - name: push
     distros:
@@ -4529,6 +4506,7 @@ buildvariants:
   - ubuntu1604-zseries-large
   - ubuntu1604-zseries-small
   batchtime: 1440 # 1 day
+  stepback: false
   expansions:
     gorootvars: PATH=/opt/mongodbtoolchain/v2/bin:$PATH
     tooltags: -gccgoflags "$(pkg-config --libs --cflags libssl libcrypto libsasl2)" -tags 'sasl ssl'
@@ -4536,7 +4514,7 @@ buildvariants:
     push_bucket: downloads.10gen.com
     push_name: linux
     push_arch: s390x-enterprise-ubuntu1604
-    compile_flags: --ssl MONGO_DISTMOD=ubuntu1604 --release -j$(grep -c ^processor /proc/cpuinfo) CCFLAGS="-march=z196 -mtune=zEC12" CC=/opt/mongodbtoolchain/v2/bin/gcc CXX=/opt/mongodbtoolchain/v2/bin/g++ OBJCOPY=/opt/mongodbtoolchain/v2/bin/objcopy
+    compile_flags: --ssl MONGO_DISTMOD=ubuntu1604 --release -j3 CCFLAGS="-march=z196 -mtune=zEC12" CC=/opt/mongodbtoolchain/v2/bin/gcc CXX=/opt/mongodbtoolchain/v2/bin/g++ OBJCOPY=/opt/mongodbtoolchain/v2/bin/objcopy
     num_jobs_available: 2
     variant_excluded_flags: requires_mmapv1
     has_packages: true
@@ -4621,6 +4599,7 @@ buildvariants:
   - name: sslSpecial
   - name: tool_WT
   - name: unittests
+  - name: write_concern_majority_passthrough_WT
   - name: push
     distros:
     - ubuntu1604-test
@@ -5055,6 +5034,7 @@ buildvariants:
   - name: tool
   - name: tool_WT
   - name: unittests
+  - name: write_concern_majority_passthrough_WT
   - name: push
     distros:
     - rhel70-small
@@ -5194,6 +5174,9 @@ buildvariants:
   - name: tool
   - name: tool_WT
   - name: unittests
+  - name: write_concern_majority_passthrough_WT
+    distros:
+    - windows-64-vs2015-large
 
 - name: enterprise-windows-64
   display_name: "* Enterprise Windows"
@@ -5447,8 +5430,8 @@ buildvariants:
 #             OSX buildvariants           #
 ###########################################
 
-- name: osx-107
-  display_name: OS X 10.7
+- name: osx-108
+  display_name: OS X 10.8
   run_on:
   - macos-1012
   batchtime: 1440 # 1 day
@@ -5457,8 +5440,8 @@ buildvariants:
     push_bucket: downloads.mongodb.org
     push_name: osx
     push_arch: x86_64
-    gorootvars: CGO_CFLAGS=-mmacosx-version-min=10.7 CGO_LDFLAGS=-mmacosx-version-min=10.7
-    compile_flags: --allocator=system -j$(sysctl -n hw.logicalcpu) --release --osx-version-min=10.7 --libc++
+    gorootvars: CGO_CFLAGS=-mmacosx-version-min=10.8 CGO_LDFLAGS=-mmacosx-version-min=10.8
+    compile_flags: --allocator=system -j$(sysctl -n hw.logicalcpu) --release --osx-version-min=10.8 --libc++
     python: python2.7
     num_jobs_available: 1
     build_mongoreplay: true
@@ -5547,8 +5530,8 @@ buildvariants:
     distros:
     - rhel70-small
 
-- name: osx-107-ssl
-  display_name: SSL OS X 10.7
+- name: osx-108-ssl
+  display_name: SSL OS X 10.8
   run_on:
   - macos-1012
   batchtime: 1440 # 1 day
@@ -5558,8 +5541,8 @@ buildvariants:
     push_name: osx-ssl
     push_arch: x86_64
     tooltags: "-tags ssl"
-    gorootvars: CGO_CPPFLAGS=-I/opt/mongodbtoolchain/v2/include CGO_CFLAGS=-mmacosx-version-min=10.7 CGO_LDFLAGS=-mmacosx-version-min=10.7
-    compile_flags: --ssl --allocator=system -j$(sysctl -n hw.logicalcpu) --release --osx-version-min=10.7 --libc++ CPPPATH=/opt/mongodbtoolchain/v2/include
+    gorootvars: CGO_CPPFLAGS=-I/opt/mongodbtoolchain/v2/include CGO_CFLAGS=-mmacosx-version-min=10.8 CGO_LDFLAGS=-mmacosx-version-min=10.8
+    compile_flags: --ssl --allocator=system -j$(sysctl -n hw.logicalcpu) --release --osx-version-min=10.8 --libc++ CPPPATH=/opt/mongodbtoolchain/v2/include
     python: python2.7
     num_jobs_available: 1
     build_mongoreplay: true
@@ -5644,12 +5627,13 @@ buildvariants:
   - name: tool
   - name: tool_WT
   - name: unittests
+  - name: write_concern_majority_passthrough_WT
   - name: push
     distros:
     - rhel70-small
 
-- name: osx-107-debug
-  display_name: "* OS X 10.7 DEBUG"
+- name: osx-108-debug
+  display_name: "* OS X 10.8 DEBUG"
   run_on:
   - macos-1012
   expansions:
@@ -5658,8 +5642,8 @@ buildvariants:
     push_name: osx-debug
     push_arch: x86_64
     num_jobs_available: 1
-    gorootvars: CGO_CFLAGS=-mmacosx-version-min=10.7 CGO_LDFLAGS=-mmacosx-version-min=10.7
-    compile_flags: --dbg=on --opt=on --allocator=system -j$(sysctl -n hw.logicalcpu) --osx-version-min=10.7 --libc++
+    gorootvars: CGO_CFLAGS=-mmacosx-version-min=10.8 CGO_LDFLAGS=-mmacosx-version-min=10.8
+    compile_flags: --dbg=on --opt=on --allocator=system -j$(sysctl -n hw.logicalcpu) --osx-version-min=10.8 --libc++
     python: python2.7
     build_mongoreplay: true
   tasks:
@@ -5689,8 +5673,8 @@ buildvariants:
   - name: tool_WT
   - name: unittests
 
-- name: enterprise-osx-107
-  display_name: Enterprise OS X 10.7
+- name: enterprise-osx-108
+  display_name: Enterprise OS X 10.8
   modules:
   - enterprise
   run_on:
@@ -5702,8 +5686,8 @@ buildvariants:
     push_name: osx
     push_arch: x86_64-enterprise
     tooltags: "-tags 'ssl sasl'"
-    gorootvars: CGO_CPPFLAGS=-I/opt/mongodbtoolchain/v2/include CGO_CFLAGS=-mmacosx-version-min=10.7 CGO_LDFLAGS=-mmacosx-version-min=10.7
-    compile_flags: --ssl --allocator=system -j$(sysctl -n hw.logicalcpu) --release --osx-version-min=10.7 --libc++ CPPPATH=/opt/mongodbtoolchain/v2/include
+    gorootvars: CGO_CPPFLAGS=-I/opt/mongodbtoolchain/v2/include CGO_CFLAGS=-mmacosx-version-min=10.8 CGO_LDFLAGS=-mmacosx-version-min=10.8
+    compile_flags: --ssl --allocator=system -j$(sysctl -n hw.logicalcpu) --release --osx-version-min=10.8 --libc++ CPPPATH=/opt/mongodbtoolchain/v2/include
     python: python2.7
     num_jobs_available: 1
     build_mongoreplay: true
@@ -5961,6 +5945,9 @@ buildvariants:
   - name: tool
   - name: tool_WT
   - name: unittests
+  - name: write_concern_majority_passthrough_WT
+    distros:
+    - rhel62-large
   - name: package
     distros:
     - ubuntu1604-packer
@@ -6113,6 +6100,7 @@ buildvariants:
   - name: ssl
   - name: sslSpecial
   - name: unittests
+  - name: write_concern_majority_passthrough_WT
 
 - name: enterprise-rhel-70-64-bit
   display_name: Enterprise RHEL 7.0
@@ -6526,6 +6514,7 @@ buildvariants:
   - name: sslSpecial
   - name: tool_WT
   - name: unittests
+  - name: write_concern_majority_passthrough_WT
   - name: push
     distros:
     - rhel70-small
@@ -6628,6 +6617,7 @@ buildvariants:
   - name: sslSpecial
   - name: tool_WT
   - name: unittests
+  - name: write_concern_majority_passthrough_WT
   - name: push
     distros:
     - rhel70-small
@@ -6763,6 +6753,7 @@ buildvariants:
   # - name: tool
   - name: tool_WT
   - name: unittests
+  - name: write_concern_majority_passthrough_WT
   - name: push
     distros:
     - rhel62-small
@@ -6921,6 +6912,7 @@ buildvariants:
     push_path: linux
     push_bucket: downloads.10gen.com
     push_name: linux
+    lang_environment: LANG=C
     push_arch: x86_64-enterprise-ubuntu1604
     compile_flags: --ssl MONGO_DISTMOD=ubuntu1604 --release -j$(grep -c ^processor /proc/cpuinfo) CC=/opt/mongodbtoolchain/v2/bin/gcc CXX=/opt/mongodbtoolchain/v2/bin/g++ OBJCOPY=/opt/mongodbtoolchain/v2/bin/objcopy
     num_jobs_available: $(grep -c ^processor /proc/cpuinfo)
@@ -7053,6 +7045,7 @@ buildvariants:
   - suse12-zseries-build
   - suse12-zseries-test
   batchtime: 1440 # 1 day
+  stepback: false
   expansions:
     gorootvars: PATH=/opt/mongodbtoolchain/v2/bin:$PATH
     tooltags: -gccgoflags "$(pkg-config --libs --cflags libssl libsasl2)" -tags 'sasl ssl'
@@ -7060,7 +7053,7 @@ buildvariants:
     push_bucket: downloads.10gen.com
     push_name: linux
     push_arch: s390x-enterprise-suse12
-    compile_flags: --ssl MONGO_DISTMOD=suse12 --release -j$(( $(grep -c ^processor /proc/cpuinfo) / 2 )) CCFLAGS="-march=z196 -mtune=zEC12" CC=/opt/mongodbtoolchain/v2/bin/gcc CXX=/opt/mongodbtoolchain/v2/bin/g++ OBJCOPY=/opt/mongodbtoolchain/v2/bin/objcopy
+    compile_flags: --ssl MONGO_DISTMOD=suse12 --release -j3 CCFLAGS="-march=z196 -mtune=zEC12" CC=/opt/mongodbtoolchain/v2/bin/gcc CXX=/opt/mongodbtoolchain/v2/bin/g++ OBJCOPY=/opt/mongodbtoolchain/v2/bin/objcopy
     num_jobs_available: $(grep -c ^processor /proc/cpuinfo)
     variant_excluded_flags: requires_mmapv1
     has_packages: true
@@ -7142,6 +7135,7 @@ buildvariants:
   - name: sslSpecial
   - name: tool_WT
   - name: unittests
+  - name: write_concern_majority_passthrough_WT
   - name: push
     distros:
     - suse12-test
@@ -7327,6 +7321,7 @@ buildvariants:
   run_on:
   - suse11-zseries-build
   batchtime: 1440 # 1 day
+  stepback: false
   expansions:
     gorootvars: PATH=/opt/mongodbtoolchain/v2/bin:$PATH
     tooltags: -gccgoflags "$(pkg-config --libs --cflags libssl libsasl2)" -tags 'sasl ssl'
@@ -7334,7 +7329,7 @@ buildvariants:
     push_bucket: downloads.10gen.com
     push_name: linux
     push_arch: s390x-enterprise-suse11
-    compile_flags: --ssl MONGO_DISTMOD=suse11 --release -j$(grep -c ^processor /proc/cpuinfo) CCFLAGS="-march=z9-109 -mtune=z10" CC=/opt/mongodbtoolchain/v2/bin/gcc CXX=/opt/mongodbtoolchain/v2/bin/g++ OBJCOPY=/opt/mongodbtoolchain/v2/bin/objcopy --use-s390x-crc32=off
+    compile_flags: --ssl MONGO_DISTMOD=suse11 --release -j3 CCFLAGS="-march=z9-109 -mtune=z10" CC=/opt/mongodbtoolchain/v2/bin/gcc CXX=/opt/mongodbtoolchain/v2/bin/g++ OBJCOPY=/opt/mongodbtoolchain/v2/bin/objcopy --use-s390x-crc32=off
     num_jobs_available: $(grep -c ^processor /proc/cpuinfo)
     variant_excluded_flags: requires_mmapv1
     has_packages: true
@@ -7413,6 +7408,7 @@ buildvariants:
   - name: sslSpecial
   - name: tool_WT
   - name: unittests
+  - name: write_concern_majority_passthrough_WT
   - name: push
     distros:
     - suse11-test
@@ -7535,94 +7531,6 @@ buildvariants:
 
 
 ###########################################
-#          Solaris buildvariants          #
-###########################################
-
-- name: solaris-64-bit
-  display_name: "* Solaris"
-  run_on:
-  - solaris
-  expansions:
-    push_path: sunos5
-    push_bucket: downloads.mongodb.org
-    push_name: sunos5
-    push_arch: x86_64
-    gorootvars: PATH=/opt/mongodbtoolchain/v2/bin:$PATH
-    tooltags: -gccgoflags "-lsocket -lnsl"
-    compile_flags: CC=/opt/mongodbtoolchain/v2/bin/gcc CXX=/opt/mongodbtoolchain/v2/bin/g++ OBJCOPY=/opt/mongodbtoolchain/v2/bin/objcopy -j$(kstat cpu | sort -u | grep -c "^module") --release CCFLAGS="-m64" LINKFLAGS="-m64 -static-libstdc++ -static-libgcc" OBJCOPY=/opt/mongodbtoolchain/bin/objcopy
-    num_jobs_available: $(( $(kstat cpu | sort -u | grep -c "^module") / 2 ))
-    use_scons_cache: true
-  tasks:
-  - name: compile
-  - name: aggregation
-  - name: aggregation_WT
-  - name: auth
-  - name: auth_WT
-  - name: bulk_gle_passthrough
-  - name: bulk_gle_passthrough_WT
-  - name: concurrency
-  - name: concurrency_WT
-  - name: concurrency_replication
-  - name: concurrency_replication_WT
-  - name: concurrency_sharded
-  - name: concurrency_sharded_WT
-  - name: concurrency_simultaneous
-  - name: concurrency_simultaneous_WT
-  - name: dbtest
-  - name: dbtest_WT
-  - name: disk
-  - name: durability
-  - name: failpoints
-  - name: httpinterface
-  - name: jsCore
-  - name: jsCore_compatibility
-  - name: jsCore_compatibility_WT
-  - name: jsCore_decimal
-  - name: jsCore_decimal_WT
-  - name: jsCore_WT
-  - name: jstestfuzz
-  - name: jstestfuzz_WT
-  - name: jstestfuzz_replication
-  - name: jstestfuzz_replication_WT
-  - name: jstestfuzz_sharded
-  - name: jstestfuzz_sharded_WT
-  - name: mmap
-  - name: mongosTest
-  - name: noPassthrough
-  - name: noPassthroughWithMongod
-  - name: noPassthroughWithMongod_WT
-  - name: noPassthrough_WT
-  - name: parallel
-  - name: parallel_compatibility
-  - name: parallel_compatibility_WT
-  - name: parallel_WT
-  - name: replica_sets
-  - name: replica_sets_WT
-  - name: replica_sets_jscore_passthrough
-  - name: replica_sets_jscore_passthrough_WT
-  - name: master_slave
-  - name: master_slave_WT
-  - name: master_slave_jscore_passthrough
-  - name: master_slave_jscore_passthrough_WT
-  - name: sharding
-  - name: sharded_collections_jscore_passthrough
-  - name: sharded_collections_jscore_passthrough_WT
-  - name: sharding_jscore_passthrough
-  - name: sharding_jscore_passthrough_WT
-  - name: sharding_jscore_passthrough_wire_ops_WT
-  - name: sharding_WT
-  - name: slow1
-  - name: slow1_WT
-  - name: serial_run
-  - name: serial_run_WT
-  - name: tool
-  - name: tool_WT
-  - name: unittests
-  - name: push
-    distros:
-    - rhel70-small
-
-###########################################
 #          Debian buildvariants           #
 ###########################################
 
@@ -8075,6 +7983,7 @@ buildvariants:
   - name: sslSpecial
   - name: tool
   - name: unittests
+  - name: write_concern_majority_passthrough_WT
 
 - name: linux-64-ephemeralForTest
   display_name: Linux (ephemeralForTest)
@@ -8232,6 +8141,7 @@ buildvariants:
   - name: sslSpecial
   - name: tool
   - name: unittests
+  - name: write_concern_majority_passthrough_WT
 
 - name: enterprise-rhel-72-s390x-inmem
   display_name: Enterprise RHEL 7.2 s390x (inMemory) DEBUG
@@ -8319,75 +8229,7 @@ buildvariants:
   - name: sslSpecial
   - name: tool
   - name: unittests
-
-- name: ubuntu1404-rocksdb
-  display_name: Ubuntu 14.04 (RocksDB)
-  modules:
-  - rocksdb
-  run_on:
-  - ubuntu1404-test
-  batchtime: 1440 # 1 day
-  expansions:
-    build_rocksdb: true
-    compile_flags: -j$(grep -c ^processor /proc/cpuinfo) --dbg=off --opt=on CPPPATH=$(readlink -f ../rocksdb/include/) LIBPATH=$(readlink -f ../rocksdb/) LIBS=rocksdb CC=/opt/mongodbtoolchain/v2/bin/gcc CXX=/opt/mongodbtoolchain/v2/bin/g++ OBJCOPY=/opt/mongodbtoolchain/v2/bin/objcopy VARIANT_DIR=release --runtime-hardening=off
-    num_jobs_available: $(grep -c ^processor /proc/cpuinfo)
-    test_flags: --storageEngine=rocksdb
-    use_scons_cache: true
-    build_mongoreplay: true
-  tasks:
-  - name: compile
-    distros:
-    - ubuntu1404-build
-  - name: aggregation
-  - name: aggregation_WT
-  - name: aggregation_auth
-  - name: auth
-  - name: dbtest
-  # - name: disk
-  # - name: durability
-  # - name: dur_jscore_passthrough
-  - name: failpoints
-  - name: failpoints_auth
-  - name: gle_auth
-  - name: gle_auth_write_cmd
-  - name: gle_auth_basics_passthrough
-  - name: gle_auth_basics_passthrough_write_cmd
-  - name: sharding_gle_auth_basics_passthrough
-  - name: sharding_gle_auth_basics_passthrough_write_cmd
-  - name: jsCore
-  - name: jsCore_auth
-  - name: jsCore_compatibility
-  - name: jstestfuzz
-  - name: jstestfuzz_concurrent_WT
-  - name: jstestfuzz_concurrent_replication_WT
-  - name: jstestfuzz_concurrent_sharded_WT
-  - name: jstestfuzz_replication
-  - name: jstestfuzz_sharded
-  - name: noPassthrough
-  - name: noPassthroughWithMongod
-  - name: bulk_gle_passthrough
-  - name: parallel
-  - name: parallel_compatibility
-  - name: concurrency
-  - name: concurrency_replication
-  - name: concurrency_sharded
-    # We use machines with more RAM to avoid OOM issues when validating collections.
-    distros:
-    - ubuntu1404-build
-  - name: concurrency_simultaneous
-  - name: replica_sets
-  - name: replica_sets_auth
-  - name: replica_sets_jscore_passthrough
-  - name: master_slave
-  - name: master_slave_WT
-  - name: master_slave_auth
-  - name: master_slave_jscore_passthrough
-  - name: sharding
-  - name: sharding_auth
-  - name: slow1
-  - name: serial_run
-  - name: sharding_jscore_passthrough
-  - name: unittests
+  - name: write_concern_majority_passthrough_WT
 
 ###########################################
 #     Experimental buildvariants          #
@@ -8405,6 +8247,7 @@ buildvariants:
     tooltags: "-tags 'ssl'"
     # We need llvm-symbolizer in the PATH for ASAN for clang-3.7 or later.
     variant_path_suffix: /usr/lib/llvm-3.8/bin
+    lang_environment: LANG=C
     san_options: LSAN_OPTIONS="suppressions=etc/lsan.suppressions" ASAN_OPTIONS=detect_leaks=1
     compile_flags: CC=/usr/bin/clang-3.8 CXX=/usr/bin/clang++-3.8 CPPDEFINES="_GLIBCXX_USE_CXX11_ABI=0" --dbg=on --opt=on --allocator=system --sanitize=address --ssl -j$(grep -c ^processor /proc/cpuinfo) --nostrip VARIANT_DIR=build
     multiversion_platform_arch: "ubuntu1604"
@@ -8538,6 +8381,7 @@ buildvariants:
   - name: tool
   - name: tool_WT
   - name: unittests
+  - name: write_concern_majority_passthrough_WT
 
 - name: ubuntu1604-asan
   display_name: ~ ASAN SSL Ubuntu 16.04
@@ -8549,6 +8393,7 @@ buildvariants:
     tooltags: "-tags 'ssl'"
     # We need llvm-symbolizer in the PATH for ASAN for clang-3.7 or later.
     variant_path_suffix: /usr/lib/llvm-3.8/bin
+    lang_environment: LANG=C
     san_options: LSAN_OPTIONS="suppressions=etc/lsan.suppressions" ASAN_OPTIONS=detect_leaks=1
     compile_flags: CC=/usr/bin/clang-3.8 CXX=/usr/bin/clang++-3.8 CPPDEFINES="_GLIBCXX_USE_CXX11_ABI=0" --opt=on --allocator=system --sanitize=address --ssl -j$(grep -c ^processor /proc/cpuinfo) --nostrip VARIANT_DIR=build
     num_jobs_available: $(($(grep -c ^processor /proc/cpuinfo) / 3)) # Avoid starting too many mongod's under ASAN build.
@@ -8580,6 +8425,7 @@ buildvariants:
     variant_path_suffix: /usr/lib/llvm-3.8/bin
     gorootvars: GOROOT=/opt/go PATH="/opt/go/bin:$PATH"
     tooltags: "-tags 'ssl sasl'"
+    lang_environment: LANG=C
     san_options: UBSAN_OPTIONS="print_stacktrace=1"
     compile_flags: CC=/usr/bin/clang-3.8 CXX=/usr/bin/clang++-3.8 --dbg=on --opt=on --allocator=system --sanitize=undefined --ssl -j$(grep -c ^processor /proc/cpuinfo) --nostrip CXXFLAGS="-nostdlib -nostdinc++" LIBPATH=/opt/mongodbtoolchain/v2/lib/gcc/x86_64-mongodb-linux/5.4.0 CPPPATH="/opt/mongodbtoolchain/v2/include/c++/5.4.0/ /opt/mongodbtoolchain/v2/include/c++/5.4.0/x86_64-mongodb-linux" CPPDEFINES="_GLIBCXX_USE_CXX11_ABI=0" VARIANT_DIR=build
     multiversion_platform_arch: "ubuntu1604"
@@ -8712,6 +8558,7 @@ buildvariants:
   - name: tool
   - name: tool_WT
   - name: unittests
+  - name: write_concern_majority_passthrough_WT
 
 - name: enterprise-ubuntu-dynamic-1604-64-bit
   display_name: "* Shared Library Enterprise Ubuntu 16.04"
@@ -8720,6 +8567,7 @@ buildvariants:
   expansions:
     gorootvars: GOROOT=/opt/go PATH="/opt/go/bin:$PATH"
     tooltags: "-tags 'ssl sasl'"
+    lang_environment: LANG=C
     compile_flags: --ssl MONGO_DISTMOD=ubuntu1604 -j$(grep -c ^processor /proc/cpuinfo) CC=/opt/mongodbtoolchain/v2/bin/gcc CXX=/opt/mongodbtoolchain/v2/bin/g++ OBJCOPY=/opt/mongodbtoolchain/v2/bin/objcopy --link-model=dynamic
     num_jobs_available: $(grep -c ^processor /proc/cpuinfo)
     use_scons_cache: true
@@ -8861,6 +8709,7 @@ buildvariants:
   - name: tool
   - name: tool_WT
   - name: unittests
+  - name: write_concern_majority_passthrough_WT
 
 - name: initsync-3dot2-rhel-62
   display_name: "~ Initial Sync 3.2 Enterprise RHEL 6.2"
@@ -8962,3 +8811,6 @@ buildvariants:
     - rhel62-large
   - name: slow1
   - name: slow1_WT
+  - name: write_concern_majority_passthrough_WT
+    distros:
+    - rhel62-large
diff --git a/etc/longevity.yml b/etc/longevity.yml
index f6da942e0c1..9cdf9502fa7 100644
--- a/etc/longevity.yml
+++ b/etc/longevity.yml
@@ -9,45 +9,22 @@ post:
       params:
         working_dir: work
         script: |
-          set -v
           source ./dsienv.sh
-          $DSI_PATH/bin/make_artifact.sh
+          make_artifact.sh
     - command: s3.put
       params:
         aws_key: ${aws_key}
         aws_secret: ${aws_secret}
-        local_file: work/reports.tgz
-        remote_file: ${project}/${build_variant}/${revision}/${task_id}/${version_id}/logs/${task_name}-${build_id}.${ext|tgz}
+        local_file: work/dsi-artifacts.tgz
+        remote_file: ${project}/${build_variant}/${revision}/${task_id}/${version_id}/logs/dsi-artifacts-${task_name}-${build_id}-${execution}.${ext|tgz}
         bucket: mciuploads
         permissions: public-read
         content_type: ${content_type|application/x-gzip}
-        display_name: test-log
-    - command: s3.put
-      params:
-        aws_key: ${aws_key}
-        aws_secret: ${aws_secret}
-        local_file: work/reports/graphs/timeseries-p1.html
-        remote_file: ${project}/${build_variant}/${revision}/${task_id}/${version_id}/logs/timeseries-p1-${task_name}-${build_id}.html
-        bucket: mciuploads
-        permissions: public-read
-        content_type: text/html
-        display_name: timeseries-p1.html
+        display_name: Dsi Artifacts - Execution ${execution}
     - command: attach.results
       params:
         file_location: work/report.json
-    - command: shell.exec
-      # destroy the cluster
-      params:
-        working_dir: work
-        script: |
-          set -e
-          set -o verbose
-          # call terraform destroy twice to avoid AWS timeout
-          yes yes | ./terraform destroy --var-file=cluster.json
-          yes yes | ./terraform destroy --var-file=cluster.json
-          # clean all file to be safe
-          rm -rf *
-          echo "Cluster DESTROYED."
+    - func: "destroy cluster"
     - command: shell.exec
       params:
         working_dir: src
@@ -79,12 +56,17 @@ functions:
         working_dir: work
         script: |
           cat > bootstrap.yml <<EOF
-          cluster_type: ${cluster}
+          infrastructure_provisioning: ${cluster}
           platform: ${platform}
-          setup: ${setup}
+          mongodb_setup: ${setup}
           storageEngine: ${storageEngine}
-          test: ${test}.longevity
+          test_control: ${test}.longevity
           production: true
+          workloads_dir: ../src/workloads/workloads
+          ycsb_dir: ../src/YCSB/YCSB
+
+          # compositions of expansions
+          mongodb_binary_archive: "https://s3.amazonaws.com/mciuploads/${project}/${version_id}/${revision}/${platform}/mongod-${version_id}.tar.gz"
           EOF
 
           cat > runtime.yml <<EOF
@@ -106,11 +88,6 @@ functions:
           ext: ${ext}
           script_flags : ${script_flags}
           dsi_rev: ${dsi_rev}
-          compare_task: ${compare_task}
-          workloads_rev: ${workloads_rev}
-
-          # compositions of expansions
-          mongodb_binary_archive: "https://s3.amazonaws.com/mciuploads/${project}/${version_id}/${revision}/${platform}/mongod-${version_id}.tar.gz"
           EOF
     - command: shell.exec
       params:
@@ -133,8 +110,6 @@ functions:
     - command: shell.exec
       params:
         working_dir: work
-        # setup execution environment
-        # configure environment
         script: |
           set -e
           virtualenv ./venv
@@ -147,35 +122,25 @@ functions:
           set -v
           set -e
           source work/dsienv.sh
-          $DSI_PATH/bin/setup-dsi-env.sh
+          setup-dsi-env.sh
           ls -a work
 
-  "infrastructure provisioning":
+  "deploy cluster":
     - command: shell.exec
-      # call infrastructure-provisioning.sh. This will either create a cluster, or update tags on existing instances.
       params:
         working_dir: work
         script: |
           set -e
           set -v
           source ./dsienv.sh
-          export PRODUCTION=true
-          $DSI_PATH/bin/infrastructure_provisioning.sh ${cluster}
-
-  "configure mongodb cluster":
-    - command: shell.exec
-      # bring up the mongod
-      params:
-        working_dir: work
-        script: |
-          set -e
-          set -o verbose
-          source ./dsienv.sh
           source ./venv/bin/activate
-          $DSI_PATH/bin/mongodb_setup.py && echo "${setup} MongoDB Cluster STARTED."
+          infrastructure_provisioning.py
+          workload_setup.py
+          mongodb_setup.py
 
   "run test":
     - command: shell.exec
+      type: test
       params:
         working_dir: work
         script: |
@@ -183,9 +148,7 @@ functions:
           set -v
           source ./dsienv.sh
           source ./venv/bin/activate
-          echo "Run test for ${test}-${storageEngine} with setup ${setup}"
-          $DSI_PATH/bin/run_test.py ${storageEngine} ${test} ${cluster}
-          echo "Complete test for ${test} with setup ${setup}!"
+          test_control.py
     - command: "json.send"
       params:
           name: "perf"
@@ -199,27 +162,18 @@ functions:
         script: |
           set -e
           set -o verbose
-          source ./dsienv.sh
           # Longevity runs so rarely, we simply teardown the cluster when done.
-          # Note that nowadays infrastructure_teardown.sh is actually copying the terraform.tfstate into /data/infrastructure_provisioning
+          # Note that nowadays infrastructure_teardown.py is actually copying the terraform.tfstate into /data/infrastructure_provisioning
           # but as of this writing the rhel70-perf-longevity distro didn't actually use the teardown hook.
           source ./dsienv.sh
-          $DSI_PATH/bin/infrastructure_teardown.sh
+          source ./venv/bin/activate
+          infrastructure_teardown.py
           echo "Cluster DESTROYED."
           echo
           echo "All perf results"
           cd ..
           cat perf.json | egrep "name|ops_per_sec"
 
-  "make test log artifact":
-    - command: shell.exec
-      params:
-        working_dir: work
-        script: |
-          set -v
-          source ./dsienv.sh
-          $DSI_PATH/bin/make_artifact.sh
-
   "analyze":
     - command: json.get_history
       params:
@@ -242,8 +196,8 @@ functions:
           set -o verbose
           TAG="3.2.1-Baseline"
           PROJECT="mongo-longevity"
-          OVERRIDEFILE="../src/dsi/dsi/analysis/v3.4/longevity_override.json"
-          python -u ../src/dsi/dsi/analysis/post_run_check.py ${script_flags} --reports-analysis reports --perf-file reports/perf.json --rev ${revision} -f history.json -t tags.json --refTag $TAG --overrideFile $OVERRIDEFILE --project_id $PROJECT --task_name ${task_name} --variant ${build_variant}
+          OVERRIDEFILE="../src/dsi/dsi/analysis/${branch_name}/longevity_override.json"
+          python -u ../src/dsi/dsi/analysis/post_run_check.py ${script_flags} --reports-analysis reports --perf-file perf.json --rev ${revision} -f history.json -t tags.json --refTag $TAG --overrideFile $OVERRIDEFILE --project_id $PROJECT --task_name ${task_name} --variant ${build_variant}
 
 tasks:
 - name: compile
@@ -313,13 +267,8 @@ tasks:
       vars:
         storageEngine: "wiredTiger"
         test: "ycsb"
-    - func: "infrastructure provisioning"
-    - func: "configure mongodb cluster"
+    - func: "deploy cluster"
     - func: "run test"
-      vars:
-        storageEngine: "wiredTiger"
-        test: "ycsb"
-    - func: "make test log artifact"
     - func: "analyze"
       vars:
         script_flags: --ycsb-throughput-analysis reports
@@ -334,13 +283,8 @@ tasks:
       vars:
         storageEngine: "mmapv1"
         test: "ycsb"
-    - func: "infrastructure provisioning"
-    - func: "configure mongodb cluster"
+    - func: "deploy cluster"
     - func: "run test"
-      vars:
-        storageEngine: "mmapv1"
-        test: "ycsb"
-    - func: "make test log artifact"
     - func: "analyze"
       vars:
         script_flags: --ycsb-throughput-analysis reports
@@ -359,6 +303,11 @@ modules:
   branch: master
 
 
+
+#######################################
+#         Linux Buildvariants         #
+#######################################
+
 buildvariants:
 - name: linux-wt-shard
   display_name: Linux WT Shard
diff --git a/etc/perf.yml b/etc/perf.yml
index 3dbc7baa61d..db116a42887 100644
--- a/etc/perf.yml
+++ b/etc/perf.yml
@@ -86,7 +86,7 @@ functions:
           set -v
           chmod +x mongod
           chmod +x mongo
-          git clone https://github.com/mongodb/mongo-perf perf
+          git clone git@github.com:mongodb/mongo-perf.git perf
           cd perf
           git describe --tags
     - command: shell.exec
@@ -158,11 +158,9 @@ functions:
         script: |
           set -o errexit
           source ./venv/bin/activate
-          REFTAGS="3.4.2-Baseline 3.2.12-Baseline 3.0.14-Baseline" # These are project specific
+          REFTAGS="3.4.9-Baseline 3.2.17-Baseline 3.0.15-Baseline" # These are project specific
           # Project Opts are  project specific, but don't match expansion from elsewhere.
-          # We could change dsi/analysis/v3.4 to dsi/analysis/perf-3.4 in DSI to get dsi/analysis/${project}/perf_override.json
-          # Note project_id performance doesn't match string perf used elsewhere for ${project}
-          PROJECT_OPTS="--overrideFile ../dsi/analysis/v3.4/perf_override.json --project_id performance"
+          PROJECT_OPTS="--overrideFile ../dsi/analysis/${branch_name}/perf_override.json --project_id performance"
           python -u ../dsi/analysis/dashboard_gen.py --rev ${revision} -f history.json -t tags.json --refTag $REFTAGS $PROJECT_OPTS --variant ${build_variant} --task ${task_name} --jira-user ${perf_jira_user} --jira-password ${perf_jira_pw} || true
     - command: "json.send"
       params:
@@ -182,8 +180,8 @@ functions:
           # appropriate flags if it's `true`.
           reports_analysis_flags="--reports-analysis . --perf-file perf/perf.json"
           cmd_flags=$([ "${reports_analysis}" = "true" ] && echo "$reports_analysis_flags" || echo "")
-          REFTAG="3.2.12-Baseline"
-          OVERRIDE="../dsi/analysis/v3.4/perf_override.json" # Note use of v3.4 here cannot use ${project}
+          REFTAG="3.2.17-Baseline"
+          OVERRIDE="../dsi/analysis/${branch_name}/perf_override.json"
           python ../dsi/analysis/perf_regression_check.py $cmd_flags -f history.json --rev ${revision} -t tags.json --refTag $REFTAG --overrideFile $OVERRIDE --variant ${build_variant} --task ${task_name} --threshold 0.10 --threadThreshold 0.15
   "run perf tests":
     - command: shell.exec
@@ -196,6 +194,7 @@ functions:
           source ./venv/bin/activate
           pip install argparse
     - command: shell.exec
+      type : test
       params:
         working_dir: src
         script: |
@@ -495,7 +494,7 @@ buildvariants:
 
 - name: linux-wt-repl
   display_name: 1-Node ReplSet Linux inMemory
-  batchtime: 360 # 6 hours
+  batchtime: 90 # 1.5 hours
   expansions:
     mongod_exec_wrapper: *exec_wrapper
     perf_exec_wrapper: *perf_wrapper
@@ -511,7 +510,7 @@ buildvariants:
 
 - name: linux-mmap-repl
   display_name: 1-Node ReplSet Linux MMAPv1
-  batchtime: 360 # 6 hours
+  batchtime: 90 # 1.5 hours
   expansions:
     mongod_exec_wrapper: *exec_wrapper
     perf_exec_wrapper: *perf_wrapper
diff --git a/etc/system_perf.yml b/etc/system_perf.yml
index 79f2a3d1a89..137bc2b4a13 100644
--- a/etc/system_perf.yml
+++ b/etc/system_perf.yml
@@ -10,37 +10,41 @@ post:
         working_dir: work
         script: |
           source ./dsienv.sh
-          $DSI_PATH/bin/make_artifact.sh
+          make_artifact.sh
     - command: s3.put
       params:
         aws_key: ${aws_key}
         aws_secret: ${aws_secret}
-        local_file: work/reports.tgz
-        remote_file: ${project}/${build_variant}/${revision}/${task_id}/${version_id}/logs/${task_name}-${build_id}.${ext|tgz}
+        local_file: work/dsi-artifacts.tgz
+        remote_file: ${project}/${build_variant}/${revision}/${task_id}/${version_id}/logs/dsi-artifacts-${task_name}-${build_id}-${execution}.${ext|tgz}
         bucket: mciuploads
         permissions: public-read
         content_type: ${content_type|application/x-gzip}
-        display_name: test-log
+        display_name: Dsi Artifacts - Execution ${execution}
     - command: s3.put
       params:
         aws_key: ${aws_key}
         aws_secret: ${aws_secret}
-        local_file: work/reports/graphs/timeseries-p1.html
-        remote_file: ${project}/${build_variant}/${revision}/${task_id}/${version_id}/logs/timeseries-p1-${task_name}-${build_id}.html
+        local_file: src/workloads/workloads/jsdoc/jsdocs-redirect.html
+        remote_file: ${project}/${build_variant}/${revision}/${task_id}/${version_id}/logs/workloads-${task_name}-${build_id}.html
         bucket: mciuploads
         permissions: public-read
         content_type: text/html
-        display_name: timeseries-p1.html
+        display_name: workloads documentation
     - command: attach.results
       params:
         file_location: work/report.json
+    - command: "json.send"
+      params:
+         name: "perf"
+         file: "work/perf.json"
     - command: shell.exec
       params:
         working_dir: work
         script: |
           source ./dsienv.sh
           if [ -e /data/infrastructure_provisioning/terraform/provisioned.${cluster} ]; then
-            $DSI_PATH/bin/mark_idle.sh
+            mark_idle.sh
           fi
 
     - command: shell.exec
@@ -75,12 +79,15 @@ functions:
         working_dir: work
         script: |
           cat > bootstrap.yml <<EOF
-          cluster_type: ${cluster}
+          infrastructure_provisioning: ${cluster}
           platform: ${platform}
-          setup: ${setup}
+          mongodb_setup: ${setup}
           storageEngine: ${storageEngine}
-          test: ${test}
+          test_control: ${test}
           production: true
+          mongodb_binary_archive: "https://s3.amazonaws.com/mciuploads/${project}/${version_id}/${revision}/${platform}/mongod-${version_id}.tar.gz"
+          workloads_dir: ../src/workloads/workloads
+          ycsb_dir: ../src/YCSB/YCSB
           EOF
 
           cat > runtime.yml <<EOF
@@ -102,11 +109,7 @@ functions:
           ext: ${ext}
           script_flags : ${script_flags}
           dsi_rev: ${dsi_rev}
-          compare_task: ${compare_task}
           workloads_rev: ${workloads_rev}
-
-          # compositions of expansions
-          mongodb_binary_archive: "https://s3.amazonaws.com/mciuploads/${project}/${version_id}/${revision}/${platform}/mongod-${version_id}.tar.gz"
           EOF
     - command: shell.exec
       params:
@@ -136,44 +139,32 @@ functions:
           virtualenv ./venv
           source ./venv/bin/activate
           pip install -r ../src/dsi/dsi/requirements.txt
-          python ../src/dsi/dsi/bin/bootstrap.py --production
-          ls
-          pwd
+          python ../src/dsi/dsi/bin/bootstrap.py
     - command: shell.exec
       params:
         script: |
           set -v
           set -e
           source work/dsienv.sh
-          $DSI_PATH/bin/setup-dsi-env.sh
+          setup-dsi-env.sh
           ls -a work
 
-  "infrastructure provisioning":
+  "deploy cluster":
     - command: shell.exec
-      # call infrastructure-provisioning.sh. This will either create a cluster, or update tags on existing instances.
       params:
         working_dir: work
         script: |
           set -e
           set -v
           source ./dsienv.sh
-          export PRODUCTION=true
-          $DSI_PATH/bin/infrastructure_provisioning.sh ${cluster}
-
-  "configure mongodb cluster":
-    - command: shell.exec
-      # bring up the mongod
-      params:
-        working_dir: work
-        script: |
-          set -e
-          set -o verbose
-          source ./dsienv.sh
           source ./venv/bin/activate
-          $DSI_PATH/bin/mongodb_setup.py && echo "${setup} MongoDB Cluster STARTED."
+          infrastructure_provisioning.py
+          workload_setup.py
+          mongodb_setup.py
 
   "run test":
     - command: shell.exec
+      type: test
       params:
         working_dir: work
         script: |
@@ -181,21 +172,12 @@ functions:
           set -v
           source ./dsienv.sh
           source ./venv/bin/activate
-          echo "Run test for ${test}-${storageEngine} with setup ${setup}"
-          $DSI_PATH/bin/run_test.py ${storageEngine} ${test} ${cluster}
+          test_control.py
     - command: "json.send"
       params:
          name: "perf"
          file: "work/perf.json"
 
-  "make test log artifact":
-    - command: shell.exec
-      params:
-        working_dir: work
-        script: |
-          source ./dsienv.sh
-          $DSI_PATH/bin/make_artifact.sh
-
   "analyze":
     - command: json.get_history
       params:
@@ -216,8 +198,8 @@ functions:
         silent: true
         script: |
           set -o errexit
-          TAGS="3.2.15-Baseline"
-          OVERRIDEFILE="../src/dsi/dsi/analysis/v3.4/system_perf_override.json"
+          TAGS="3.2.17-Baseline 3.4.9-Baseline"
+          OVERRIDEFILE="../src/dsi/dsi/analysis/${branch_name}/system_perf_override.json"
           python -u ../src/dsi/dsi/analysis/dashboard_gen.py --rev ${revision} -f history.json -t tags.json --refTag $TAGS --overrideFile $OVERRIDEFILE --project_id sys-perf --variant ${build_variant} --task ${task_name} --jira-user ${perf_jira_user} --jira-password ${perf_jira_pw} || true
     - command: "json.send"
       params:
@@ -231,48 +213,9 @@ functions:
         script: |
           set -o errexit
           set -o verbose
-          TAG="3.2.15-Baseline"
-          OVERRIDEFILE="../src/dsi/dsi/analysis/v3.4/system_perf_override.json"
-          python -u ../src/dsi/dsi/analysis/post_run_check.py ${script_flags} --reports-analysis reports --perf-file reports/perf.json --rev ${revision} -f history.json -t tags.json --refTag $TAG --overrideFile $OVERRIDEFILE --project_id sys-perf --variant ${build_variant} --task ${task_name}
-
-  "compare":
-    - command: shell.exec
-      params:
-        script: |
-          set -o verbose
-          rm -rf ./src ./work
-          mkdir src
-          mkdir work
-    - command: manifest.load
-    - command: git.get_project
-      params:
-        directory: src
-        revisions: # for each module include revision as <module_name> : ${<module_name>_rev}
-          dsi: ${dsi_rev}
-    - command: json.get
-      params:
-        task: ${compare_task}
-        variant : ${variant1}
-        file: "work/standalone.json"
-        name: "perf"
-    - command: json.get
-      params:
-        task: ${compare_task}
-        variant : ${variant2}
-        file: "work/oplog.json"
-        name: "perf"
-    - command: shell.exec
-      type : test
-      params:
-        working_dir: work
-        script: |
-          set -o errexit
-          set -o verbose
-          python -u ../src/dsi/dsi/analysis/compare.py -b standalone.json -c oplog.json
-    - command: "json.send"
-      params:
-        name: "perf"
-        file: "work/perf.json"
+          TAG="3.2.17-Baseline"
+          OVERRIDEFILE="../src/dsi/dsi/analysis/${branch_name}/system_perf_override.json"
+          python -u ../src/dsi/dsi/analysis/post_run_check.py ${script_flags} --reports-analysis reports --perf-file perf.json --rev ${revision} -f history.json -t tags.json --refTag $TAG --overrideFile $OVERRIDEFILE --project_id sys-perf --variant ${build_variant} --task ${task_name}
 
 #######################################
 #               Tasks                 #
@@ -323,8 +266,17 @@ tasks:
           then
             echo "Fetching JS test DB correctness checks from directory jstests/hooks"
             cp -a jstests/hooks/* mongodb/jstests/hooks
+
+            echo "Now adding our own special run_validate_collections.js wrapper"
+            mv mongodb/jstests/hooks/run_validate_collections.js mongodb/jstests/hooks/run_validate_collections.actual.js
+
+            cat << EOF > mongodb/jstests/hooks/run_validate_collections.js
+            print("NOTE: run_validate_collections.js will skip the oplog!");
+            TestData = { skipValidationNamespaces: ['local.oplog.rs'] };
+            load('jstests/hooks/run_validate_collections.actual.js');
+          EOF
           fi
-          tar cvf mongodb.tar mongodb
+          tar cf mongodb.tar mongodb
           gzip mongodb.tar
     - command: s3.put
       params:
@@ -341,25 +293,20 @@ tasks:
   depends_on:
     - name: compile
       variant: linux-standalone
+  priority: 5
   commands:
     - func: "prepare environment"
       vars:
         storageEngine: "wiredTiger"
         test: "ycsb"
-    - func: "infrastructure provisioning"
-    - func: "configure mongodb cluster"
-      vars:
-        storageEngine: "wiredTiger"
+    - func: "deploy cluster"
     - func: "run test"
-      vars:
-        storageEngine: "wiredTiger"
-        test: "ycsb"
-    - func: "make test log artifact"
     - func: "analyze"
       vars:
         script_flags: --ycsb-throughput-analysis reports
 
 - name: industry_benchmarks_MMAPv1
+  priority: 5
   depends_on:
     - name: compile
       variant: linux-standalone
@@ -368,20 +315,46 @@ tasks:
       vars:
         storageEngine: "mmapv1"
         test: "ycsb"
-    - func: "infrastructure provisioning"
-    - func: "configure mongodb cluster"
+    - func: "deploy cluster"
+    - func: "run test"
+    - func: "analyze"
       vars:
-        storageEngine: "mmapv1"
+        script_flags: --ycsb-throughput-analysis reports
+
+- name: industry_benchmarks_wmajority_WT
+  priority: 5
+  depends_on:
+    - name: compile
+      variant: linux-standalone
+  commands:
+    - func: "prepare environment"
+      vars:
+        storageEngine: "wiredTiger"
+        test: "ycsb-wmajority"
+    - func: "deploy cluster"
     - func: "run test"
+    - func: "analyze"
+      vars:
+        script_flags: --ycsb-throughput-analysis reports
+
+- name: industry_benchmarks_wmajority_MMAPv1
+  priority: 5
+  depends_on:
+    - name: compile
+      variant: linux-standalone
+  commands:
+    - func: "prepare environment"
       vars:
         storageEngine: "mmapv1"
-        test: "ycsb"
-    - func: "make test log artifact"
+        test: "ycsb-wmajority"
+    - func: "deploy cluster"
+    - func: "run test"
     - func: "analyze"
       vars:
         script_flags: --ycsb-throughput-analysis reports
 
 - name: core_workloads_WT
+  priority: 5
   depends_on:
     - name: compile
       variant: linux-standalone
@@ -390,18 +363,12 @@ tasks:
       vars:
         storageEngine: "wiredTiger"
         test: "core"
-    - func: "infrastructure provisioning"
-    - func: "configure mongodb cluster"
-      vars:
-        storageEngine: "wiredTiger"
+    - func: "deploy cluster"
     - func: "run test"
-      vars:
-        storageEngine: "wiredTiger"
-        test: "core"
-    - func: "make test log artifact"
     - func: "analyze"
 
 - name: core_workloads_MMAPv1
+  priority: 5
   depends_on:
     - name: compile
       variant: linux-standalone
@@ -410,18 +377,12 @@ tasks:
       vars:
         storageEngine: "mmapv1"
         test: "core"
-    - func: "infrastructure provisioning"
-    - func: "configure mongodb cluster"
-      vars:
-        storageEngine: "mmapv1"
+    - func: "deploy cluster"
     - func: "run test"
-      vars:
-        storageEngine: "mmapv1"
-        test: "core"
-    - func: "make test log artifact"
     - func: "analyze"
 
 - name: non_sharded_workloads_WT
+  priority: 5
   depends_on:
     - name: compile
       variant: linux-standalone
@@ -430,18 +391,12 @@ tasks:
       vars:
         storageEngine: "wiredTiger"
         test: "non_sharded"
-    - func: "infrastructure provisioning"
-    - func: "configure mongodb cluster"
-      vars:
-        storageEngine: "wiredTiger"
+    - func: "deploy cluster"
     - func: "run test"
-      vars:
-        storageEngine: "wiredTiger"
-        test: "non_sharded"
-    - func: "make test log artifact"
     - func: "analyze"
 
 - name: non_sharded_workloads_MMAPv1
+  priority: 5
   depends_on:
     - name: compile
       variant: linux-standalone
@@ -450,18 +405,12 @@ tasks:
       vars:
         storageEngine: "mmapv1"
         test: "non_sharded"
-    - func: "infrastructure provisioning"
-    - func: "configure mongodb cluster"
-      vars:
-        storageEngine: "mmapv1"
+    - func: "deploy cluster"
     - func: "run test"
-      vars:
-        storageEngine: "mmapv1"
-        test: "non_sharded"
-    - func: "make test log artifact"
     - func: "analyze"
 
 - name: mongos_workloads_WT
+  priority: 5
   depends_on:
     - name: compile
       variant: linux-standalone
@@ -470,18 +419,12 @@ tasks:
       vars:
         storageEngine: "wiredTiger"
         test: "mongos"
-    - func: "infrastructure provisioning"
-    - func: "configure mongodb cluster"
-      vars:
-        storageEngine: "wiredTiger"
+    - func: "deploy cluster"
     - func: "run test"
-      vars:
-        storageEngine: "wiredTiger"
-        test: "mongos"
-    - func: "make test log artifact"
     - func: "analyze"
 
 - name: mongos_workloads_MMAPv1
+  priority: 5
   depends_on:
     - name: compile
       variant: linux-standalone
@@ -491,18 +434,12 @@ tasks:
       vars:
         storageEngine: "mmapv1"
         test: "mongos"
-    - func: "infrastructure provisioning"
-    - func: "configure mongodb cluster"
-      vars:
-        storageEngine: "mmapv1"
+    - func: "deploy cluster"
     - func: "run test"
-      vars:
-        storageEngine: "mmapv1"
-        test: "mongos"
-    - func: "make test log artifact"
     - func: "analyze"
 
 - name: move_chunk_workloads_WT
+  priority: 5
   depends_on:
     - name: compile
       variant: linux-standalone
@@ -511,18 +448,12 @@ tasks:
       vars:
         storageEngine: "wiredTiger"
         test: "move_chunk"
-    - func: "infrastructure provisioning"
-    - func: "configure mongodb cluster"
-      vars:
-        storageEngine: "wiredTiger"
+    - func: "deploy cluster"
     - func: "run test"
-      vars:
-        storageEngine: "wiredTiger"
-        test: "move_chunk"
-    - func: "make test log artifact"
     - func: "analyze"
 
 - name: move_chunk_workloads_MMAPv1
+  priority: 5
   depends_on:
     - name: compile
       variant: linux-standalone
@@ -531,18 +462,12 @@ tasks:
       vars:
         storageEngine: "mmapv1"
         test: "move_chunk"
-    - func: "infrastructure provisioning"
-    - func: "configure mongodb cluster"
-      vars:
-        storageEngine: "mmapv1"
+    - func: "deploy cluster"
     - func: "run test"
-      vars:
-        storageEngine: "mmapv1"
-        test: "move_chunk"
-    - func: "make test log artifact"
     - func: "analyze"
 
 - name: secondary_performance_WT
+  priority: 5
   depends_on:
     - name: compile
       variant: linux-standalone
@@ -550,21 +475,15 @@ tasks:
     - func: "prepare environment"
       vars:
         storageEngine: "wiredTiger"
+        # Unfortunately the dash/underscore style is different for mongodb_setup and test_control
         test: "secondary_performance"
-        setup: "replica-2node"
-    - func: "infrastructure provisioning"
-    - func: "configure mongodb cluster"
-      vars:
-        storageEngine: "wiredTiger"
-        setup: "replica-2node"
+        setup: "secondary-performance"
+    - func: "deploy cluster"
     - func: "run test"
-      vars:
-        storageEngine: "wiredTiger"
-        test: "secondary_performance"
-    - func: "make test log artifact"
     - func: "analyze"
 
 - name: secondary_performance_MMAPv1
+  priority: 5
   depends_on:
     - name: compile
       variant: linux-standalone
@@ -572,85 +491,15 @@ tasks:
     - func: "prepare environment"
       vars:
         storageEngine: "mmapv1"
+        # Unfortunately the dash/underscore style is different for mongodb_setup and test_control
         test: "secondary_performance"
-        setup: "replica-2node"
-    - func: "infrastructure provisioning"
-    - func: "configure mongodb cluster"
-      vars:
-        storageEngine: "mmapv1"
-        setup: "replica-2node"
+        setup: "secondary-performance"
+    - func: "deploy cluster"
     - func: "run test"
-      vars:
-        storageEngine: "mmapv1"
-        test: "secondary_performance"
-    - func: "make test log artifact"
-    - func: "analyze"
-
-- name: industry_benchmarks_WT_oplog_comp
-  depends_on:
-    - name: industry_benchmarks_WT
-      variant: linux-standalone
-      status : "*"
-    - name: industry_benchmarks_WT
-      variant: linux-1-node-replSet
-      status: "*"
-  commands:
-    - func: "compare"
-      vars:
-        compare_task: "industry_benchmarks_WT"
-        variant1: "linux-standalone"
-        variant2: "linux-1-node-replSet"
-    - func: "analyze"
-
-- name: industry_benchmarks_MMAPv1_oplog_comp
-  depends_on:
-    - name: industry_benchmarks_MMAPv1
-      variant: linux-standalone
-      status: "*"
-    - name: industry_benchmarks_MMAPv1
-      variant: linux-1-node-replSet
-      status: "*"
-  commands:
-    - func: "compare"
-      vars:
-        compare_task: "industry_benchmarks_MMAPv1"
-        variant1: "linux-standalone"
-        variant2: "linux-1-node-replSet"
-    - func: "analyze"
-
-- name: core_workloads_WT_oplog_comp
-  depends_on:
-    - name: core_workloads_WT
-      variant: linux-standalone
-      status: "*"
-    - name: core_workloads_WT
-      variant: linux-1-node-replSet
-      status: "*"
-  commands:
-    - func: "compare"
-      vars:
-        compare_task: "core_workloads_WT"
-        variant1: "linux-standalone"
-        variant2: "linux-1-node-replSet"
-    - func: "analyze"
-
-- name: core_workloads_MMAPv1_oplog_comp
-  depends_on:
-    - name: core_workloads_MMAPv1
-      variant: linux-standalone
-      status: "*"
-    - name: core_workloads_MMAPv1
-      variant: linux-1-node-replSet
-      status: "*"
-  commands:
-    - func: "compare"
-      vars:
-        compare_task: "core_workloads_MMAPv1"
-        variant1: "linux-standalone"
-        variant2: "linux-1-node-replSet"
     - func: "analyze"
 
 - name: initialsync_WT
+  priority: 5
   depends_on:
     - name: compile
       variant: linux-standalone
@@ -658,19 +507,13 @@ tasks:
     - func: "prepare environment"
       vars:
         storageEngine: "wiredTiger"
-        test: "initialSync"
-    - func: "infrastructure provisioning"
-    - func: "configure mongodb cluster"
-      vars:
-        storageEngine: "wiredTiger"
+        test: "initialsync"
+    - func: "deploy cluster"
     - func: "run test"
-      vars:
-        storageEngine: "wiredTiger"
-        test: "initialSync"
-    - func: "make test log artifact"
     - func: "analyze"
 
 - name: initialsync_MMAPv1
+  priority: 5
   depends_on:
     - name: compile
       variant : linux-standalone
@@ -678,19 +521,13 @@ tasks:
     - func: "prepare environment"
       vars:
         storageEngine: "mmapv1"
-        test: "initialSync"
-    - func: "infrastructure provisioning"
-    - func: "configure mongodb cluster"
-      vars:
-        storageEngine: "mmapv1"
+        test: "initialsync"
+    - func: "deploy cluster"
     - func: "run test"
-      vars:
-        storageEngine: "mmapv1"
-        test: "initialSync"
-    - func: "make test log artifact"
     - func: "analyze"
 
 - name: initialsync-logkeeper_WT
+  priority: 5
   depends_on:
     - name: compile
       variant: linux-standalone
@@ -700,18 +537,10 @@ tasks:
       vars:
         storageEngine: "wiredTiger"
         test: "initialsync-logkeeper"
-    - func: "infrastructure provisioning"
-    - func: "configure mongodb cluster"
-      vars:
-        storageEngine: "wiredTiger"
+    - func: "deploy cluster"
     - func: "run test"
-      vars:
-        storageEngine: "wiredTiger"
-        test: "initialsync-logkeeper"
-    - func: "make test log artifact"
     - func: "analyze"
 
-
 #######################################
 #               Modules               #
 #######################################
@@ -798,6 +627,8 @@ buildvariants:
     - name: industry_benchmarks_WT
     - name: core_workloads_WT
     - name: industry_benchmarks_MMAPv1
+    - name: industry_benchmarks_wmajority_WT
+    - name: industry_benchmarks_wmajority_MMAPv1
     - name: core_workloads_MMAPv1
     - name: mongos_workloads_WT
     - name: mongos_workloads_MMAPv1
@@ -820,6 +651,8 @@ buildvariants:
     - name: industry_benchmarks_WT
     - name: core_workloads_WT
     - name: industry_benchmarks_MMAPv1
+    - name: industry_benchmarks_wmajority_WT
+    - name: industry_benchmarks_wmajority_MMAPv1
     - name: core_workloads_MMAPv1
     - name: secondary_performance_WT
     - name: secondary_performance_MMAPv1
@@ -858,19 +691,3 @@ buildvariants:
       - "rhel70-perf-initialsync-logkeeper"
   tasks:
     - name: initialsync-logkeeper_WT
-
-- name: linux-oplog-compare
-  display_name: Linux Oplog Compare
-  batchtime: 10080 # 7 days
-  modules: *modules
-  expansions:
-    compile_flags: -j$(grep -c ^processor /proc/cpuinfo) CC=/opt/mongodbtoolchain/v2/bin/gcc CXX=/opt/mongodbtoolchain/v2/bin/g++ OBJCOPY=/opt/mongodbtoolchain/v2/bin/objcopy
-    use_scons_cache: true
-    project: *project
-  run_on:
-      - "rhel70-perf-single"
-  tasks:
-    - name: industry_benchmarks_WT_oplog_comp
-    - name: core_workloads_WT_oplog_comp
-    - name: industry_benchmarks_MMAPv1_oplog_comp
-    - name: core_workloads_MMAPv1_oplog_comp
diff --git a/jstests/aggregation/extras/utils.js b/jstests/aggregation/extras/utils.js
index 2dd1038388c..68b2597c533 100644
--- a/jstests/aggregation/extras/utils.js
+++ b/jstests/aggregation/extras/utils.js
@@ -276,7 +276,7 @@ function assertErrorCode(coll, pipe, code, errmsg) {
     var cursorRes = coll.runCommand("aggregate", cmd);
     if (cursorRes.ok) {
         var followupBatchSize = 0;  // default
-        var cursor = new DBCommandCursor(coll.getMongo(), cursorRes, followupBatchSize);
+        var cursor = new DBCommandCursor(cursorRes._mongo, cursorRes, followupBatchSize);
 
         var error = assert.throws(function() {
             cursor.itcount();
diff --git a/jstests/auth/mongos_cache_invalidation.js b/jstests/auth/mongos_cache_invalidation.js
index e36445cac5e..b53656de3a1 100644
--- a/jstests/auth/mongos_cache_invalidation.js
+++ b/jstests/auth/mongos_cache_invalidation.js
@@ -61,7 +61,7 @@ db3.auth('spencer', 'pwd');
  * At this point we have 3 handles to the "test" database, each of which are on connections to
  * different mongoses.  "db1", "db2", and "db3" are all auth'd as spencer@test and will be used
  * to verify that user and role data changes get propaged to their mongoses.
- * "db2" is connected to a mongos with a 10 second user cache invalidation interval,
+ * "db2" is connected to a mongos with a 5 second user cache invalidation interval,
  * while "db3" is connected to a mongos with a 10 minute cache invalidation interval.
  */
 
@@ -202,12 +202,9 @@ db3.auth('spencer', 'pwd');
     assert.commandFailedWithCode(db1.foo.runCommand("collStats"), authzErrorCode);
 
     // s1/db2 should update its cache in 10 seconds.
-    assert.soon(
-        function() {
-            return db2.foo.runCommand("collStats").code == authzErrorCode;
-        },
-        "Mongos did not update its user cache after 10 seconds",
-        6 * 1000);  // Give an extra 1 second to avoid races
+    assert.soon(function() {
+        return db2.foo.runCommand("collStats").code == authzErrorCode;
+    }, "Mongos did not update its user cache after 10 seconds", 10 * 1000);
 
     // We manually invalidate the cache on s2/db3.
     db3.adminCommand("invalidateUserCache");
diff --git a/jstests/auth/scram-credentials-invalid.js b/jstests/auth/scram-credentials-invalid.js
new file mode 100644
index 00000000000..16c0c204d12
--- /dev/null
+++ b/jstests/auth/scram-credentials-invalid.js
@@ -0,0 +1,45 @@
+// Ensure that attempting to use SCRAM-SHA-1 auth on a
+// user with invalid SCRAM-SHA-1 credentials fails gracefully.
+
+(function() {
+    'use strict';
+
+    function runTest(mongod) {
+        assert(mongod);
+        const admin = mongod.getDB('admin');
+        const test = mongod.getDB('test');
+
+        admin.createUser({user: 'admin', pwd: 'pass', roles: jsTest.adminUserRoles});
+        assert(admin.auth('admin', 'pass'));
+
+        test.createUser({user: 'user', pwd: 'pass', roles: jsTest.basicUserRoles});
+
+        // Give the test user an invalid set of SCRAM-SHA-1 credentials.
+        assert.eq(admin.system.users
+                      .update({_id: "test.user"}, {
+                          $set: {
+                              "credentials.SCRAM-SHA-1": {
+                                  salt: "AAAA",
+                                  storedKey: "AAAA",
+                                  serverKey: "AAAA",
+                                  iterationCount: 10000
+                              }
+                          }
+                      })
+                      .nModified,
+                  1,
+                  "Should have updated one document for user@test");
+        admin.logout();
+
+        assert(!test.auth({user: 'user', pwd: 'pass'}));
+
+        assert.soon(function() {
+            const log = cat(mongod.fullOptions.logFile);
+            return /Unable to perform SCRAM-SHA-1 auth.* invalid SCRAM credentials/.test(log);
+        }, "No warning issued for invalid SCRAM-SHA-1 credendials doc", 30 * 1000, 5 * 1000);
+    }
+
+    const mongod = MongoRunner.runMongod({auth: "", useLogFiles: true});
+    runTest(mongod);
+    MongoRunner.stopMongod(mongod);
+})();
diff --git a/jstests/auth/system_authorization_indexes.js b/jstests/auth/system_authorization_indexes.js
new file mode 100644
index 00000000000..496f78578a4
--- /dev/null
+++ b/jstests/auth/system_authorization_indexes.js
@@ -0,0 +1,66 @@
+/** Ensure that authorization system collections' indexes are correctly generated.
+ *
+ * This test requires users to persist across a restart.
+ * @tags: [requires_persistence]
+ */
+
+(function() {
+    let conn = MongoRunner.runMongod({smallfiles: ""});
+    let db = conn.getDB("admin");
+
+    // TEST: User and role collections start off with no indexes
+    assert.eq(0, db.system.users.getIndexes().length);
+    assert.eq(0, db.system.roles.getIndexes().length);
+
+    // TEST: User and role creation generates indexes
+    db.createUser({user: "user", pwd: "pwd", roles: []});
+    assert.eq(2, db.system.users.getIndexes().length);
+
+    db.createRole({role: "role", privileges: [], roles: []});
+    assert.eq(2, db.system.roles.getIndexes().length);
+
+    // TEST: Destroying admin.system.users index and restarting will recreate it
+    assert.commandWorked(db.system.users.dropIndexes());
+    MongoRunner.stopMongod(conn);
+    conn = MongoRunner.runMongod({restart: conn, cleanData: false});
+    db = conn.getDB("admin");
+    assert.eq(2, db.system.users.getIndexes().length);
+    assert.eq(2, db.system.roles.getIndexes().length);
+
+    // TEST: Destroying admin.system.roles index and restarting will recreate it
+    assert.commandWorked(db.system.roles.dropIndexes());
+    MongoRunner.stopMongod(conn);
+    conn = MongoRunner.runMongod({restart: conn, cleanData: false});
+    db = conn.getDB("admin");
+    assert.eq(2, db.system.users.getIndexes().length);
+    assert.eq(2, db.system.roles.getIndexes().length);
+
+    // TEST: Destroying both authorization indexes and restarting will recreate them
+    assert.commandWorked(db.system.users.dropIndexes());
+    assert.commandWorked(db.system.roles.dropIndexes());
+    MongoRunner.stopMongod(conn);
+    conn = MongoRunner.runMongod({restart: conn, cleanData: false});
+    db = conn.getDB("admin");
+    assert.eq(2, db.system.users.getIndexes().length);
+    assert.eq(2, db.system.roles.getIndexes().length);
+
+    // TEST: Destroying the admin.system.users index and restarting will recreate it, even if
+    // admin.system.roles does not exist
+    db.dropDatabase();
+    db.createUser({user: "user", pwd: "pwd", roles: []});
+    assert.commandWorked(db.system.users.dropIndexes());
+    MongoRunner.stopMongod(conn);
+    conn = MongoRunner.runMongod({restart: conn, cleanData: false});
+    db = conn.getDB("admin");
+    assert.eq(2, db.system.users.getIndexes().length);
+
+    // TEST: Destroying the admin.system.roles index and restarting will recreate it, even if
+    // admin.system.users does not exist
+    db.dropDatabase();
+    db.createRole({role: "role", privileges: [], roles: []});
+    assert.commandWorked(db.system.roles.dropIndexes());
+    MongoRunner.stopMongod(conn);
+    conn = MongoRunner.runMongod({restart: conn, cleanData: false});
+    db = conn.getDB("admin");
+    assert.eq(2, db.system.roles.getIndexes().length);
+})();
diff --git a/jstests/concurrency/fsm_libs/cluster.js b/jstests/concurrency/fsm_libs/cluster.js
index a5705a409af..f3e2429f898 100644
--- a/jstests/concurrency/fsm_libs/cluster.js
+++ b/jstests/concurrency/fsm_libs/cluster.js
@@ -573,6 +573,25 @@ var Cluster = function(options) {
 
         return data;
     };
+
+    this.isRunningWiredTigerLSM = function isRunningWiredTigerLSM() {
+        var adminDB = this.getDB('admin');
+
+        if (this.isSharded()) {
+            // Get the storage engine the sharded cluster is configured to use from one of the
+            // shards since mongos won't report it.
+            adminDB = st.shard0.getDB('admin');
+        }
+
+        var res = adminDB.runCommand({getCmdLineOpts: 1});
+        assert.commandWorked(res, 'failed to get command line options');
+
+        var wiredTigerOptions = res.parsed.storage.wiredTiger || {};
+        var wiredTigerCollectionConfig = wiredTigerOptions.collectionConfig || {};
+        var wiredTigerConfigString = wiredTigerCollectionConfig.configString || '';
+
+        return wiredTigerConfigString === 'type=lsm';
+    };
 };
 
 /**
diff --git a/jstests/concurrency/fsm_workloads/compact.js b/jstests/concurrency/fsm_workloads/compact.js
index 8f91f52bf5e..e86e6ef555f 100644
--- a/jstests/concurrency/fsm_workloads/compact.js
+++ b/jstests/concurrency/fsm_workloads/compact.js
@@ -87,12 +87,27 @@ var $config = (function() {
         dropCollections(db, pattern);
     };
 
+    var skip = function skip(cluster) {
+        if (cluster.isRunningWiredTigerLSM()) {
+            // There is a known hang during concurrent FSM workloads with the compact command used
+            // with wiredTiger LSM variants. Bypass this command for the wiredTiger LSM variant
+            // until a fix is available for WT-2523.
+            return {
+                skip: true,
+                msg: 'WT-2523: compact command can cause hang using WT LSM index during ' +
+                    'concurrent workloads'
+            };
+        }
+        return {skip: false};
+    };
+
     return {
         threadCount: 15,
         iterations: 10,
         states: states,
         transitions: transitions,
         teardown: teardown,
-        data: data
+        data: data,
+        skip: skip
     };
 })();
diff --git a/jstests/concurrency/fsm_workloads/remove_multiple_documents.js b/jstests/concurrency/fsm_workloads/remove_multiple_documents.js
index bfd64cd6790..1461ca0cc03 100644
--- a/jstests/concurrency/fsm_workloads/remove_multiple_documents.js
+++ b/jstests/concurrency/fsm_workloads/remove_multiple_documents.js
@@ -34,8 +34,17 @@ var $config = (function() {
         }
     };
 
+    var skip = function skip(cluster) {
+        // When the balancer is enabled, the nRemoved result may be inaccurate as
+        // a chunk migration may be active, causing the count function to assert.
+        if (cluster.isBalancerEnabled()) {
+            return {skip: true, msg: 'does not run when balancer is enabled.'};
+        }
+        return {skip: false};
+    };
+
     var transitions = {init: {count: 1}, count: {remove: 1}, remove: {remove: 0.825, count: 0.125}};
 
-    return {threadCount: 10, iterations: 20, states: states, transitions: transitions};
+    return {threadCount: 10, iterations: 20, states: states, transitions: transitions, skip: skip};
 
 })();
diff --git a/jstests/concurrency/fsm_workloads/remove_where.js b/jstests/concurrency/fsm_workloads/remove_where.js
index f9c0e6a2c03..3ef214c769b 100644
--- a/jstests/concurrency/fsm_workloads/remove_where.js
+++ b/jstests/concurrency/fsm_workloads/remove_where.js
@@ -38,5 +38,14 @@ var $config = extendWorkload($config, function($config, $super) {
         /* no-op to prevent index from being created */
     };
 
+    $config.skip = function skip(cluster) {
+        // When the balancer is enabled, the nRemoved result may be inaccurate as
+        // a chunk migration may be active, causing the count function to assert.
+        if (cluster.isBalancerEnabled()) {
+            return {skip: true, msg: 'does not run when balancer is enabled.'};
+        }
+        return {skip: false};
+    };
+
     return $config;
 });
diff --git a/jstests/core/apply_ops1.js b/jstests/core/apply_ops1.js
index 8ce15b1df22..7088145c718 100644
--- a/jstests/core/apply_ops1.js
+++ b/jstests/core/apply_ops1.js
@@ -286,9 +286,54 @@
         'applyOps should fail on unknown operation type "x" with valid "ns" value');
 
     assert.eq(0, t.find().count(), "Non-zero amount of documents in collection to start");
+
+    /**
+     * Test function for running CRUD operations on non-existent namespaces using various
+     * combinations of invalid namespaces (collection/database), allowAtomic and alwaysUpsert.
+     *
+     * Leave 'expectedErrorCode' undefined if this command is expected to run successfully.
+     */
+    function testCrudOperationOnNonExistentNamespace(optype, o, o2, expectedErrorCode) {
+        expectedErrorCode = expectedErrorCode || ErrorCodes.OK;
+        const t2 = db.getSiblingDB('apply_ops1_no_such_db').getCollection('t');
+        [t, t2].forEach(coll => {
+            const op = {op: optype, ns: coll.getFullName(), o: o, o2: o2};
+            [false, true].forEach(allowAtomic => {
+                [false, true].forEach(alwaysUpsert => {
+                    const cmd = {
+                        applyOps: [op],
+                        allowAtomic: allowAtomic,
+                        alwaysUpsert: alwaysUpsert
+                    };
+                    jsTestLog('Testing applyOps on non-existent namespace: ' + tojson(cmd));
+                    if (expectedErrorCode === ErrorCodes.OK) {
+                        assert.commandWorked(db.adminCommand(cmd));
+                    } else {
+                        assert.commandFailedWithCode(db.adminCommand(cmd), expectedErrorCode);
+                    }
+                });
+            });
+        });
+    }
+
+    // Insert and update operations on non-existent collections/databases should return
+    // NamespaceNotFound.
+    testCrudOperationOnNonExistentNamespace('i', {_id: 0}, {}, ErrorCodes.NamespaceNotFound);
+    testCrudOperationOnNonExistentNamespace('u', {x: 0}, {_id: 0}, ErrorCodes.NamespaceNotFound);
+
+    // Delete operations on non-existent collections/databases should return OK for idempotency
+    // reasons.
+    testCrudOperationOnNonExistentNamespace('d', {_id: 0}, {});
+
     assert.commandFailed(
-        db.adminCommand({applyOps: [{"op": "i", "ns": t.getFullName(), "o": {_id: 5, x: 17}}]}),
-        "Applying an insert operation on a non-existent collection should fail");
+        db.adminCommand({
+            applyOps: [{
+                "op": "c",
+                "ns": "admin.$cmd",
+                "o": {applyOps: [{"op": "i", "ns": t.getFullName(), "o": {_id: 5, x: 17}}]}
+            }]
+        }),
+        "Applying a nested insert operation on a non-existent collection should fail");
 
     assert.commandWorked(db.createCollection(t.getName()));
     var a = assert.commandWorked(
diff --git a/jstests/core/apply_ops_atomicity.js b/jstests/core/apply_ops_atomicity.js
index 911ce32a311..e704815a889 100644
--- a/jstests/core/apply_ops_atomicity.js
+++ b/jstests/core/apply_ops_atomicity.js
@@ -29,10 +29,12 @@
     var newDBName = "apply_ops_atomicity";
     var newDB = db.getSiblingDB(newDBName);
     assert.commandWorked(newDB.dropDatabase());
-    // Do an update on a non-existent database, since only 'u' ops can implicitly create
-    // collections.
-    assert.commandWorked(newDB.runCommand(
-        {applyOps: [{op: "u", ns: newDBName + ".foo", o: {_id: 5, x: 17}, o2: {_id: 5, x: 16}}]}));
+    // Updates on a non-existent database no longer implicitly create collections and will fail with
+    // a NamespaceNotFound error.
+    assert.commandFailedWithCode(newDB.runCommand({
+        applyOps: [{op: "u", ns: newDBName + ".foo", o: {_id: 5, x: 17}, o2: {_id: 5, x: 16}}]
+    }),
+                                 ErrorCodes.NamespaceNotFound);
 
     var sawTooManyLocksError = false;
 
diff --git a/jstests/core/autocomplete.js b/jstests/core/autocomplete.js
new file mode 100644
index 00000000000..6eb6e21a7a3
--- /dev/null
+++ b/jstests/core/autocomplete.js
@@ -0,0 +1,42 @@
+/**
+ * Validate auto complete works for various javascript types implemented by C++.
+ */
+(function() {
+    'use strict';
+
+    function testAutoComplete(prefix) {
+        // This method updates a global object with an array of strings on success.
+        shellAutocomplete(prefix);
+        return __autocomplete__;
+    }
+
+    // Create a collection
+    db.auto_complete_coll.insert({});
+
+    // Validate DB auto completion
+    const db_stuff = testAutoComplete('db.');
+
+    // Verify we enumerate built-in methods
+    assert.contains('db.prototype', db_stuff);
+    assert.contains('db.hasOwnProperty', db_stuff);
+    assert.contains('db.toString(', db_stuff);
+
+    // Verify we have some methods we added
+    assert.contains('db.adminCommand(', db_stuff);
+    assert.contains('db.runCommand(', db_stuff);
+
+    // Verify we enumerate collections
+    assert.contains('db.auto_complete_coll', db_stuff);
+
+    // Validate Collection autocompletion
+    const coll_stuff = testAutoComplete('db.auto_complete_coll.');
+
+    // Verify we enumerate built-in methods
+    assert.contains('db.auto_complete_coll.prototype', coll_stuff);
+    assert.contains('db.auto_complete_coll.hasOwnProperty', coll_stuff);
+    assert.contains('db.auto_complete_coll.toString(', coll_stuff);
+
+    // Verify we have some methods we added
+    assert.contains('db.auto_complete_coll.aggregate(', coll_stuff);
+    assert.contains('db.auto_complete_coll.runCommand(', coll_stuff);
+})();
\ No newline at end of file
diff --git a/jstests/core/batch_write_command_delete.js b/jstests/core/batch_write_command_delete.js
index 2aefcea6a7f..99b5f8e3a61 100644
--- a/jstests/core/batch_write_command_delete.js
+++ b/jstests/core/batch_write_command_delete.js
@@ -1,3 +1,5 @@
+// @tags: [assumes_write_concern_unchanged]
+
 //
 // Ensures that mongod respects the batch write protocols for delete
 //
diff --git a/jstests/core/batch_write_command_insert.js b/jstests/core/batch_write_command_insert.js
index 274f35513e7..3fa6cf98756 100644
--- a/jstests/core/batch_write_command_insert.js
+++ b/jstests/core/batch_write_command_insert.js
@@ -1,3 +1,5 @@
+// @tags: [assumes_write_concern_unchanged]
+
 //
 // Ensures that mongod respects the batch write protocol for inserts
 //
diff --git a/jstests/core/batch_write_command_update.js b/jstests/core/batch_write_command_update.js
index 2d9d2d699b2..987525e2515 100644
--- a/jstests/core/batch_write_command_update.js
+++ b/jstests/core/batch_write_command_update.js
@@ -1,3 +1,5 @@
+// @tags: [assumes_write_concern_unchanged]
+
 //
 // Ensures that mongod respects the batch write protocols for updates
 //
diff --git a/jstests/core/bypass_doc_validation.js b/jstests/core/bypass_doc_validation.js
index d9bca81ab6d..c2cdfff0bc9 100644
--- a/jstests/core/bypass_doc_validation.js
+++ b/jstests/core/bypass_doc_validation.js
@@ -1,5 +1,7 @@
 // Test the bypassDocumentValidation flag with some database commands. The test uses relevant shell
 // helpers when they're available for the respective server commands.
+//
+// @tags: [requires_collmod_command]
 
 (function() {
     'use strict';
diff --git a/jstests/core/capped6.js b/jstests/core/capped6.js
index e94e7ea44e8..ca216fbe96a 100644
--- a/jstests/core/capped6.js
+++ b/jstests/core/capped6.js
@@ -1,4 +1,11 @@
 // Test NamespaceDetails::cappedTruncateAfter via "captrunc" command
+//
+// @tags: [
+//   # This test attempts to perform read operations on a capped collection after truncating
+//   # documents using the captrunc command. The writes from the captrunc command aren't guaranteed
+//   # to become visible until a later w="majority" write occurs.
+//   assumes_write_concern_unchanged,
+// ]
 (function() {
     var coll = db.capped6;
 
diff --git a/jstests/core/collation_plan_cache.js b/jstests/core/collation_plan_cache.js
index 0eec77388e4..790bbbadaa6 100644
--- a/jstests/core/collation_plan_cache.js
+++ b/jstests/core/collation_plan_cache.js
@@ -1,4 +1,11 @@
 // Integration testing for the plan cache and index filter commands with collation.
+//
+// @tags: [
+//   # This test attempts to perform queries and introspect the server's plan cache entries. The
+//   # former operation may be routed to a secondary in the replica set, whereas the latter must be
+//   # routed to the primary.
+//   assumes_read_preference_unchanged,
+// ]
 (function() {
     'use strict';
 
@@ -237,4 +244,4 @@
     assert.eq(0,
               coll.runCommand('planCacheListFilters').filters.length,
               'unexpected number of plan cache filters');
-})();
\ No newline at end of file
+})();
diff --git a/jstests/core/collection_info_cache_race.js b/jstests/core/collection_info_cache_race.js
index d57fc3340db..8fcde050e99 100644
--- a/jstests/core/collection_info_cache_race.js
+++ b/jstests/core/collection_info_cache_race.js
@@ -5,9 +5,9 @@ var coll = db.collection_info_cache_race;
 coll.drop();
 assert.commandWorked(db.createCollection(coll.getName(), {autoIndexId: false}));
 // Fails when SERVER-16502 was not fixed, due to invariant
-assert.writeOK(coll.save({_id: false}, {writeConcern: {w: 1}}));
+assert.writeOK(coll.save({_id: false}));
 
 coll.drop();
 assert.commandWorked(db.createCollection(coll.getName(), {autoIndexId: false}));
 assert.eq(null, coll.findOne());
-assert.writeOK(coll.save({_id: false}, {writeConcern: {w: 1}}));
+assert.writeOK(coll.save({_id: false}));
diff --git a/jstests/core/collmod.js b/jstests/core/collmod.js
index 16f9694560c..e366041bd99 100644
--- a/jstests/core/collmod.js
+++ b/jstests/core/collmod.js
@@ -1,5 +1,7 @@
 // Basic js tests for the collMod command.
 // Test setting the usePowerOf2Sizes flag, and modifying TTL indexes.
+//
+// @tags: [requires_collmod_command]
 
 function debug(x) {
     // printjson( x );
diff --git a/jstests/core/collmod_bad_spec.js b/jstests/core/collmod_bad_spec.js
index ccce81fd4b1..c3d5e7a148e 100644
--- a/jstests/core/collmod_bad_spec.js
+++ b/jstests/core/collmod_bad_spec.js
@@ -2,6 +2,8 @@
 //
 // Tests that a collMod with a bad specification does not cause any changes, and does not crash the
 // server.
+//
+// @tags: [requires_collmod_command]
 (function() {
     "use strict";
 
diff --git a/jstests/core/commands_that_do_not_write_do_not_accept_wc.js b/jstests/core/commands_that_do_not_write_do_not_accept_wc.js
index ef5f8762a42..a9c72e108a1 100644
--- a/jstests/core/commands_that_do_not_write_do_not_accept_wc.js
+++ b/jstests/core/commands_that_do_not_write_do_not_accept_wc.js
@@ -2,6 +2,8 @@
  * This file tests commands that do not support write concern. It passes both valid and invalid
  * writeConcern fields to commands and expects the commands to fail with a writeConcernNotSupported
  * error.
+ *
+ * @tags: [assumes_write_concern_unchanged]
  */
 
 (function() {
diff --git a/jstests/core/constructors.js b/jstests/core/constructors.js
index 9e2cd26bbe8..93842780b42 100644
--- a/jstests/core/constructors.js
+++ b/jstests/core/constructors.js
@@ -1,4 +1,6 @@
 // Tests to see what validity checks are done for 10gen specific object construction
+//
+// @tags: [requires_eval_command]
 
 // Takes a list of constructors and returns a new list with an extra entry for each constructor with
 // "new" prepended
diff --git a/jstests/core/count10.js b/jstests/core/count10.js
index 2a1853c399a..453775c97f5 100644
--- a/jstests/core/count10.js
+++ b/jstests/core/count10.js
@@ -1,4 +1,11 @@
 // Test that interrupting a count returns an error code.
+//
+// @tags: [
+//   # This test attempts to perform a count command and find it using the currentOp command. The
+//   # former operation may be routed to a secondary in the replica set, whereas the latter must be
+//   # routed to the primary.
+//   assumes_read_preference_unchanged,
+// ]
 
 t = db.count10;
 t.drop();
diff --git a/jstests/core/count_plan_summary.js b/jstests/core/count_plan_summary.js
index 48891d21e8e..365f289c457 100644
--- a/jstests/core/count_plan_summary.js
+++ b/jstests/core/count_plan_summary.js
@@ -1,5 +1,11 @@
-// Test that the plan summary string appears in db.currentOp() for
-// count operations. SERVER-14064.
+// Test that the plan summary string appears in db.currentOp() for count operations. SERVER-14064.
+//
+// @tags: [
+//   # This test attempts to perform a find command and find it using the currentOp command. The
+//   # former operation may be routed to a secondary in the replica set, whereas the latter must be
+//   # routed to the primary.
+//   assumes_read_preference_unchanged,
+// ]
 
 var t = db.jstests_count_plan_summary;
 t.drop();
diff --git a/jstests/core/crud_api.js b/jstests/core/crud_api.js
index c9dbfb40c85..d572d4b90a5 100644
--- a/jstests/core/crud_api.js
+++ b/jstests/core/crud_api.js
@@ -1,3 +1,5 @@
+// @tags: [assumes_write_concern_unchanged]
+
 (function() {
     "use strict";
 
diff --git a/jstests/core/diagdata.js b/jstests/core/diagdata.js
index 490e4a3eb2b..6938f8c5102 100644
--- a/jstests/core/diagdata.js
+++ b/jstests/core/diagdata.js
@@ -1,4 +1,5 @@
 // Test that verifies getDiagnosticData returns FTDC data
+load('jstests/libs/ftdc.js');
 
 (function() {
     "use strict";
@@ -6,28 +7,5 @@
     // Verify we require admin database
     assert.commandFailed(db.diagdata.runCommand("getDiagnosticData"));
 
-    // We need to retry a few times if run this test immediately after mongod is started as FTDC may
-    // not have run yet.
-    var foundGoodDocument = false;
-
-    for (var i = 0; i < 60; ++i) {
-        var result = db.adminCommand("getDiagnosticData");
-        assert.commandWorked(result);
-
-        var data = result.data;
-
-        if (!data.hasOwnProperty("start")) {
-            // Wait a little longer for FTDC to start
-            sleep(500);
-        } else {
-            // Check for a few common properties to ensure we got data
-            assert(data.hasOwnProperty("serverStatus"),
-                   "does not have 'serverStatus' in '" + tojson(data) + "'");
-            assert(data.hasOwnProperty("end"), "does not have 'end' in '" + tojson(data) + "'");
-            foundGoodDocument = true;
-        }
-    }
-    assert(foundGoodDocument,
-           "getDiagnosticData failed to return a non-empty command, is FTDC running?");
-
+    verifyGetDiagnosticData(db.getSiblingDB('admin'));
 })();
diff --git a/jstests/core/doc_validation.js b/jstests/core/doc_validation.js
index a30763869e7..95dbae1b219 100644
--- a/jstests/core/doc_validation.js
+++ b/jstests/core/doc_validation.js
@@ -1,4 +1,6 @@
 // Test basic inserts and updates with document validation.
+//
+// @tags: [requires_collmod_command]
 (function() {
     "use strict";
 
diff --git a/jstests/core/doc_validation_invalid_validators.js b/jstests/core/doc_validation_invalid_validators.js
index b78b31c0977..70262f5de43 100644
--- a/jstests/core/doc_validation_invalid_validators.js
+++ b/jstests/core/doc_validation_invalid_validators.js
@@ -1,5 +1,7 @@
 // Verify invalid validator statements won't work and that we
 // can't create validated collections on restricted databases.
+//
+// @tags: [requires_collmod_command]
 (function() {
     "use strict";
 
diff --git a/jstests/core/doc_validation_options.js b/jstests/core/doc_validation_options.js
index 8a96685e48f..d1434af6d45 100644
--- a/jstests/core/doc_validation_options.js
+++ b/jstests/core/doc_validation_options.js
@@ -1,3 +1,4 @@
+// @tags: [requires_collmod_command]
 (function() {
     "use strict";
 
diff --git a/jstests/core/dropdb_race.js b/jstests/core/dropdb_race.js
index bd5e7e5ddba..d8b49d08174 100644
--- a/jstests/core/dropdb_race.js
+++ b/jstests/core/dropdb_race.js
@@ -1,4 +1,6 @@
 // test dropping a db with simultaneous commits
+//
+// @tags: [assumes_write_concern_unchanged]
 
 m = db.getMongo();
 baseName = "jstests_dur_droprace";
diff --git a/jstests/core/elemMatchProjection.js b/jstests/core/elemMatchProjection.js
index 4e80e8a296e..5060d8a6346 100644
--- a/jstests/core/elemMatchProjection.js
+++ b/jstests/core/elemMatchProjection.js
@@ -4,35 +4,59 @@ t.drop();
 
 date1 = new Date();
 
+// Generate monotonically increasing _id values. ObjectIds generated by the shell are not guaranteed
+// to be monotically increasing, and we will depend on the _id sort order later in the test.
+var currentId = 0;
+function nextId() {
+    return ++currentId;
+}
+
 // Insert various styles of arrays
 for (i = 0; i < 100; i++) {
-    t.insert({group: 1, x: [1, 2, 3, 4, 5]});
-    t.insert({group: 2, x: [{a: 1, b: 2}, {a: 2, c: 3}, {a: 1, d: 5}]});
+    t.insert({_id: nextId(), group: 1, x: [1, 2, 3, 4, 5]});
+    t.insert({_id: nextId(), group: 2, x: [{a: 1, b: 2}, {a: 2, c: 3}, {a: 1, d: 5}]});
     t.insert({
+        _id: nextId(),
         group: 3,
         x: [{a: 1, b: 2}, {a: 2, c: 3}, {a: 1, d: 5}],
         y: [{aa: 1, bb: 2}, {aa: 2, cc: 3}, {aa: 1, dd: 5}]
     });
-    t.insert({group: 3, x: [{a: 1, b: 3}, {a: -6, c: 3}]});
-    t.insert({group: 4, x: [{a: 1, b: 4}, {a: -6, c: 3}]});
-    t.insert({group: 5, x: [new Date(), 5, 10, 'string', new ObjectId(), 123.456]});
+    t.insert({_id: nextId(), group: 3, x: [{a: 1, b: 3}, {a: -6, c: 3}]});
+    t.insert({_id: nextId(), group: 4, x: [{a: 1, b: 4}, {a: -6, c: 3}]});
+    t.insert({_id: nextId(), group: 5, x: [new Date(), 5, 10, 'string', new ObjectId(), 123.456]});
     t.insert({
+        _id: nextId(),
         group: 6,
         x: [{a: 'string', b: date1}, {a: new ObjectId(), b: 1.2345}, {a: 'string2', b: date1}]
     });
-    t.insert({group: 7, x: [{y: [1, 2, 3, 4]}]});
-    t.insert({group: 8, x: [{y: [{a: 1, b: 2}, {a: 3, b: 4}]}]});
-    t.insert({group: 9, x: [{y: [{a: 1, b: 2}, {a: 3, b: 4}]}, {z: [{a: 1, b: 2}, {a: 3, b: 4}]}]});
-    t.insert({group: 10, x: [{a: 1, b: 2}, {a: 3, b: 4}], y: [{c: 1, d: 2}, {c: 3, d: 4}]});
-    t.insert({group: 10, x: [{a: 1, b: 2}, {a: 3, b: 4}], y: [{c: 1, d: 2}, {c: 3, d: 4}]});
+    t.insert({_id: nextId(), group: 7, x: [{y: [1, 2, 3, 4]}]});
+    t.insert({_id: nextId(), group: 8, x: [{y: [{a: 1, b: 2}, {a: 3, b: 4}]}]});
+    t.insert({
+        _id: nextId(),
+        group: 9,
+        x: [{y: [{a: 1, b: 2}, {a: 3, b: 4}]}, {z: [{a: 1, b: 2}, {a: 3, b: 4}]}]
+    });
+    t.insert({
+        _id: nextId(),
+        group: 10,
+        x: [{a: 1, b: 2}, {a: 3, b: 4}],
+        y: [{c: 1, d: 2}, {c: 3, d: 4}]
+    });
+    t.insert({
+        _id: nextId(),
+        group: 10,
+        x: [{a: 1, b: 2}, {a: 3, b: 4}],
+        y: [{c: 1, d: 2}, {c: 3, d: 4}]
+    });
     t.insert({
+        _id: nextId(),
         group: 11,
         x: [{a: 1, b: 2}, {a: 2, c: 3}, {a: 1, d: 5}],
         covered: [{aa: 1, bb: 2}, {aa: 2, cc: 3}, {aa: 1, dd: 5}]
     });
-    t.insert({group: 12, x: {y: [{a: 1, b: 1}, {a: 1, b: 2}]}});
-    t.insert({group: 13, x: [{a: 1, b: 1}, {a: 1, b: 2}]});
-    t.insert({group: 13, x: [{a: 1, b: 2}, {a: 1, b: 1}]});
+    t.insert({_id: nextId(), group: 12, x: {y: [{a: 1, b: 1}, {a: 1, b: 2}]}});
+    t.insert({_id: nextId(), group: 13, x: [{a: 1, b: 1}, {a: 1, b: 2}]});
+    t.insert({_id: nextId(), group: 13, x: [{a: 1, b: 2}, {a: 1, b: 1}]});
 }
 t.ensureIndex({
     group: 1,
diff --git a/jstests/core/error2.js b/jstests/core/error2.js
index 6f0b95bc17e..fb6a8e6e3b2 100644
--- a/jstests/core/error2.js
+++ b/jstests/core/error2.js
@@ -1,4 +1,5 @@
 // Test that client gets stack trace on failed invoke
+// @tags: [requires_eval_command]
 
 f = db.jstests_error2;
 
diff --git a/jstests/core/eval0.js b/jstests/core/eval0.js
index 5802f2597cb..c21c6be66c6 100644
--- a/jstests/core/eval0.js
+++ b/jstests/core/eval0.js
@@ -1,3 +1,5 @@
+// @tags: [requires_eval_command]
+
 assert.writeOK(db.evalprep.insert({}), "db must exist for eval to succeed");
 db.evalprep.drop();
 assert.eq(17,
diff --git a/jstests/core/eval1.js b/jstests/core/eval1.js
index 8b139cae02a..b5bffac892e 100644
--- a/jstests/core/eval1.js
+++ b/jstests/core/eval1.js
@@ -1,3 +1,4 @@
+// @tags: [requires_eval_command]
 
 t = db.eval1;
 t.drop();
diff --git a/jstests/core/eval3.js b/jstests/core/eval3.js
index c4f8be21056..b95837a6817 100644
--- a/jstests/core/eval3.js
+++ b/jstests/core/eval3.js
@@ -1,3 +1,4 @@
+// @tags: [requires_eval_command]
 
 t = db.eval3;
 t.drop();
diff --git a/jstests/core/eval4.js b/jstests/core/eval4.js
index 0d120b393de..9b0c2a49d82 100644
--- a/jstests/core/eval4.js
+++ b/jstests/core/eval4.js
@@ -1,3 +1,4 @@
+// @tags: [requires_eval_command]
 
 t = db.eval4;
 t.drop();
diff --git a/jstests/core/eval5.js b/jstests/core/eval5.js
index 46bd679dd77..815365ac8b9 100644
--- a/jstests/core/eval5.js
+++ b/jstests/core/eval5.js
@@ -1,3 +1,4 @@
+// @tags: [requires_eval_command]
 
 t = db.eval5;
 t.drop();
diff --git a/jstests/core/eval6.js b/jstests/core/eval6.js
index 31258f6917b..96b43b3516c 100644
--- a/jstests/core/eval6.js
+++ b/jstests/core/eval6.js
@@ -1,3 +1,4 @@
+// @tags: [requires_eval_command]
 
 t = db.eval6;
 t.drop();
diff --git a/jstests/core/eval7.js b/jstests/core/eval7.js
index 80197fcdde6..3bace093db1 100644
--- a/jstests/core/eval7.js
+++ b/jstests/core/eval7.js
@@ -1,3 +1,5 @@
+// @tags: [requires_eval_command]
+
 assert.writeOK(db.evalprep.insert({}), "db must exist for eval to succeed");
 db.evalprep.drop();
 
diff --git a/jstests/core/eval9.js b/jstests/core/eval9.js
index 1480ff6519c..82c230a8a95 100644
--- a/jstests/core/eval9.js
+++ b/jstests/core/eval9.js
@@ -1,3 +1,5 @@
+// @tags: [requires_eval_command]
+
 assert.writeOK(db.evalprep.insert({}), "db must exist for eval to succeed");
 db.evalprep.drop();
 
diff --git a/jstests/core/eval_mr.js b/jstests/core/eval_mr.js
index 4a3dc8dad6c..33ca96eea01 100644
--- a/jstests/core/eval_mr.js
+++ b/jstests/core/eval_mr.js
@@ -1,4 +1,6 @@
 // Test that the eval command can't be used to invoke the mapReduce command.  SERVER-17889.
+//
+// @tags: [requires_eval_command]
 (function() {
     "use strict";
     db.eval_mr.drop();
diff --git a/jstests/core/eval_nolock.js b/jstests/core/eval_nolock.js
index 9511784becb..0fde2666f5d 100644
--- a/jstests/core/eval_nolock.js
+++ b/jstests/core/eval_nolock.js
@@ -1,3 +1,4 @@
+// @tags: [requires_eval_command]
 
 t = db.eval_nolock;
 t.drop();
diff --git a/jstests/core/evala.js b/jstests/core/evala.js
index 7ccf33ac754..09241eeedee 100644
--- a/jstests/core/evala.js
+++ b/jstests/core/evala.js
@@ -1,3 +1,4 @@
+// @tags: [requires_eval_command]
 
 t = db.evala;
 t.drop();
diff --git a/jstests/core/evalb.js b/jstests/core/evalb.js
index 3391c4cc4f2..2de16f4ea83 100644
--- a/jstests/core/evalb.js
+++ b/jstests/core/evalb.js
@@ -1,5 +1,7 @@
 // Check the return value of a db.eval function running a database query, and ensure the function's
 // contents are logged in the profile log.
+//
+// @tags: [requires_eval_command]
 
 // Use a reserved database name to avoid a conflict in the parallel test suite.
 var stddb = db;
diff --git a/jstests/core/evald.js b/jstests/core/evald.js
index 8049d2ba8ae..43e74fc4600 100644
--- a/jstests/core/evald.js
+++ b/jstests/core/evald.js
@@ -1,3 +1,5 @@
+// @tags: [requires_eval_command]
+
 t = db.jstests_evald;
 t.drop();
 
diff --git a/jstests/core/evale.js b/jstests/core/evale.js
index 1ddc8519fc6..20384e0d741 100644
--- a/jstests/core/evale.js
+++ b/jstests/core/evale.js
@@ -1,3 +1,5 @@
+// @tags: [requires_eval_command]
+
 t = db.jstests_evale;
 t.drop();
 
diff --git a/jstests/core/evalg.js b/jstests/core/evalg.js
index 18503659217..56cb1dedd5f 100644
--- a/jstests/core/evalg.js
+++ b/jstests/core/evalg.js
@@ -1,4 +1,6 @@
 // SERVER-17499: Test behavior of getMore on aggregation cursor under eval command.
+//
+// @tags: [requires_eval_command]
 db.evalg.drop();
 for (var i = 0; i < 102; ++i) {
     db.evalg.insert({});
diff --git a/jstests/core/evalh.js b/jstests/core/evalh.js
index 11e672f6bf4..b9c0d486d59 100644
--- a/jstests/core/evalh.js
+++ b/jstests/core/evalh.js
@@ -1,5 +1,7 @@
 /**
  * Test that db.eval does not support auth.
+ *
+ * @tags: [requires_eval_command]
  */
 (function() {
     'use strict';
diff --git a/jstests/core/evalj.js b/jstests/core/evalj.js
index f2326fff365..d6ef46430de 100644
--- a/jstests/core/evalj.js
+++ b/jstests/core/evalj.js
@@ -1,3 +1,5 @@
+// @tags: [requires_eval_command]
+
 (function() {
     "use strict";
 
diff --git a/jstests/core/existsa.js b/jstests/core/existsa.js
index e9430b489a3..d98fd3f2d68 100644
--- a/jstests/core/existsa.js
+++ b/jstests/core/existsa.js
@@ -1,101 +1,111 @@
-// Sparse indexes are disallowed for $exists:false queries.  SERVER-3918
-
-t = db.jstests_existsa;
-t.drop();
-
-t.save({});
-t.save({a: 1});
-t.save({a: {x: 1}, b: 1});
-
-/** Configure testing of an index { <indexKeyField>:1 }. */
-function setIndex(_indexKeyField) {
-    indexKeyField = _indexKeyField;
-    indexKeySpec = {};
-    indexKeySpec[indexKeyField] = 1;
-    t.ensureIndex(indexKeySpec, {sparse: true});
-}
-setIndex('a');
-
-/** @return count when hinting the index to use. */
-function hintedCount(query) {
-    return t.find(query).hint(indexKeySpec).itcount();
-}
-
-/** The query field does not exist and the sparse index is not used without a hint. */
-function assertMissing(query, expectedMissing, expectedIndexedMissing) {
-    expectedMissing = expectedMissing || 1;
-    expectedIndexedMissing = expectedIndexedMissing || 0;
-    assert.eq(expectedMissing, t.count(query));
-    // We also shouldn't get a different count depending on whether
-    // an index is used or not.
-    assert.eq(expectedIndexedMissing, hintedCount(query));
-}
-
-/** The query field exists and the sparse index is used without a hint. */
-function assertExists(query, expectedExists) {
-    expectedExists = expectedExists || 2;
-    assert.eq(expectedExists, t.count(query));
-    // An $exists:true predicate generates no index filters.  Add another predicate on the index key
-    // to trigger use of the index.
-    andClause = {};
-    andClause[indexKeyField] = {$ne: null};
-    Object.extend(query, {$and: [andClause]});
-    assert.eq(expectedExists, t.count(query));
-    assert.eq(expectedExists, hintedCount(query));
-}
-
-/** The query field exists and the sparse index is not used without a hint. */
-function assertExistsUnindexed(query, expectedExists) {
-    expectedExists = expectedExists || 2;
-    assert.eq(expectedExists, t.count(query));
-    // Even with another predicate on the index key, the sparse index is disallowed.
-    andClause = {};
-    andClause[indexKeyField] = {$ne: null};
-    Object.extend(query, {$and: [andClause]});
-    assert.eq(expectedExists, t.count(query));
-    assert.eq(expectedExists, hintedCount(query));
-}
-
-// $exists:false queries match the proper number of documents and disallow the sparse index.
-assertMissing({a: {$exists: false}});
-assertMissing({a: {$not: {$exists: true}}});
-assertMissing({$and: [{a: {$exists: false}}]});
-assertMissing({$or: [{a: {$exists: false}}]});
-assertMissing({$nor: [{a: {$exists: true}}]});
-assertMissing({'a.x': {$exists: false}}, 2, 1);
-
-// Currently a sparse index is disallowed even if the $exists:false query is on a different field.
-assertMissing({b: {$exists: false}}, 2, 1);
-assertMissing({b: {$exists: false}, a: {$ne: 6}}, 2, 1);
-assertMissing({b: {$not: {$exists: true}}}, 2, 1);
-
-// Top level $exists:true queries match the proper number of documents
-// and use the sparse index on { a : 1 }.
-assertExists({a: {$exists: true}});
-
-// Nested $exists queries match the proper number of documents and disallow the sparse index.
-assertExistsUnindexed({$nor: [{a: {$exists: false}}]});
-assertExistsUnindexed({$nor: [{'a.x': {$exists: false}}]}, 1);
-assertExistsUnindexed({a: {$not: {$exists: false}}});
-
-// Nested $exists queries disallow the sparse index in some cases where it is not strictly
-// necessary to do so.  (Descriptive tests.)
-assertExistsUnindexed({$nor: [{b: {$exists: false}}]}, 1);  // Unindexed field.
-assertExists({$or: [{a: {$exists: true}}]});                // $exists:true not $exists:false.
-
-// Behavior is similar with $elemMatch.
-t.drop();
-t.save({a: [{}]});
-t.save({a: [{b: 1}]});
-t.save({a: [{b: 1}]});
-setIndex('a.b');
-
-assertMissing({a: {$elemMatch: {b: {$exists: false}}}});
-// A $elemMatch predicate is treated as nested, and the index should be used for $exists:true.
-assertExists({a: {$elemMatch: {b: {$exists: true}}}});
-
-// A non sparse index will not be disallowed.
-t.drop();
-t.save({});
-t.ensureIndex({a: 1});
-assert.eq(1, t.find({a: {$exists: false}}).itcount());
+/**
+ * Tests that sparse indexes are disallowed for $exists:false queries.
+ */
+(function() {
+    "use strict";
+
+    const coll = db.jstests_existsa;
+    coll.drop();
+
+    assert.writeOK(coll.insert({}));
+    assert.writeOK(coll.insert({a: 1}));
+    assert.writeOK(coll.insert({a: {x: 1}, b: 1}));
+
+    let indexKeySpec = {};
+    let indexKeyField = '';
+
+    /** Configure testing of an index { <indexKeyField>:1 }. */
+    function setIndex(_indexKeyField) {
+        indexKeyField = _indexKeyField;
+        indexKeySpec = {};
+        indexKeySpec[indexKeyField] = 1;
+        coll.ensureIndex(indexKeySpec, {sparse: true});
+    }
+    setIndex('a');
+
+    /** @return count when hinting the index to use. */
+    function hintedCount(query) {
+        return coll.find(query).hint(indexKeySpec).itcount();
+    }
+
+    /** The query field does not exist and the sparse index is not used without a hint. */
+    function assertMissing(query, expectedMissing = 1, expectedIndexedMissing = 0) {
+        assert.eq(expectedMissing, coll.count(query));
+        // We also shouldn't get a different count depending on whether
+        // an index is used or not.
+        assert.eq(expectedIndexedMissing, hintedCount(query));
+    }
+
+    /** The query field exists and the sparse index is used without a hint. */
+    function assertExists(query, expectedExists = 2) {
+        assert.eq(expectedExists, coll.count(query));
+        // An $exists:true predicate generates no index filters. Add another predicate on the index
+        // key to trigger use of the index.
+        let andClause = {};
+        andClause[indexKeyField] = {$ne: null};
+        Object.extend(query, {$and: [andClause]});
+        assert.eq(expectedExists, coll.count(query));
+        assert.eq(expectedExists, hintedCount(query));
+    }
+
+    /** The query field exists and the sparse index is not used without a hint. */
+    function assertExistsUnindexed(query, expectedExists = 2) {
+        assert.eq(expectedExists, coll.count(query));
+        // Even with another predicate on the index key, the sparse index is disallowed.
+        let andClause = {};
+        andClause[indexKeyField] = {$ne: null};
+        Object.extend(query, {$and: [andClause]});
+        assert.eq(expectedExists, coll.count(query));
+        assert.eq(expectedExists, hintedCount(query));
+    }
+
+    // $exists:false queries match the proper number of documents and disallow the sparse index.
+    assertMissing({a: {$exists: false}});
+    assertMissing({a: {$not: {$exists: true}}});
+    assertMissing({$and: [{a: {$exists: false}}]});
+    assertMissing({$or: [{a: {$exists: false}}]});
+    assertMissing({$nor: [{a: {$exists: true}}]});
+    assertMissing({'a.x': {$exists: false}}, 2, 1);
+
+    // Currently a sparse index is disallowed even if the $exists:false query is on a different
+    // field.
+    assertMissing({b: {$exists: false}}, 2, 1);
+    assertMissing({b: {$exists: false}, a: {$ne: 6}}, 2, 1);
+    assertMissing({b: {$not: {$exists: true}}}, 2, 1);
+
+    // Top level $exists:true queries match the proper number of documents
+    // and use the sparse index on { a : 1 }.
+    assertExists({a: {$exists: true}});
+
+    // Nested $exists queries match the proper number of documents and disallow the sparse index.
+    assertExistsUnindexed({$nor: [{a: {$exists: false}}]});
+    assertExistsUnindexed({$nor: [{'a.x': {$exists: false}}]}, 1);
+    assertExistsUnindexed({a: {$not: {$exists: false}}});
+
+    // Nested $exists queries disallow the sparse index in some cases where it is not strictly
+    // necessary to do so.  (Descriptive tests.)
+    assertExistsUnindexed({$nor: [{b: {$exists: false}}]}, 1);  // Unindexed field.
+    assertExists({$or: [{a: {$exists: true}}]});                // $exists:true not $exists:false.
+
+    // Behavior is similar with $elemMatch.
+    coll.drop();
+    assert.writeOK(coll.insert({a: [{}]}));
+    assert.writeOK(coll.insert({a: [{b: 1}]}));
+    assert.writeOK(coll.insert({a: [{b: [1]}]}));
+    setIndex('a.b');
+
+    assertMissing({a: {$elemMatch: {b: {$exists: false}}}});
+
+    // A $elemMatch predicate is treated as nested, and the index should be used for $exists:true.
+    assertExists({a: {$elemMatch: {b: {$exists: true}}}});
+
+    // A $not within $elemMatch should not attempt to use a sparse index for $exists:false.
+    assertExistsUnindexed({'a.b': {$elemMatch: {$not: {$exists: false}}}}, 1);
+    assertExistsUnindexed({'a.b': {$elemMatch: {$gt: 0, $not: {$exists: false}}}}, 1);
+
+    // A non sparse index will not be disallowed.
+    coll.drop();
+    assert.writeOK(coll.insert({}));
+    coll.ensureIndex({a: 1});
+    assert.eq(1, coll.find({a: {$exists: false}}).itcount());
+})();
diff --git a/jstests/core/fsync.js b/jstests/core/fsync.js
index 13ff20e4177..2ad625d88ba 100644
--- a/jstests/core/fsync.js
+++ b/jstests/core/fsync.js
@@ -6,6 +6,8 @@
  * - Confirm that writes can progress after fsyncUnlock
  * - Confirm that the command can be run repeatedly without breaking things
  * - Confirm that the pseudo commands and eval can perform fsyncLock/Unlock
+ *
+ * @tags: [requires_eval_command]
  */
 (function() {
     "use strict";
diff --git a/jstests/core/fts_dotted_prefix_fields.js b/jstests/core/fts_dotted_prefix_fields.js
new file mode 100644
index 00000000000..f811c4a7203
--- /dev/null
+++ b/jstests/core/fts_dotted_prefix_fields.js
@@ -0,0 +1,15 @@
+// Test that text search works correct when the text index has dotted paths as the non-text
+// prefixes.
+(function() {
+    "use strict";
+
+    let coll = db.fts_dotted_prefix_fields;
+    coll.drop();
+    assert.commandWorked(coll.createIndex({"a.x": 1, "a.y": 1, "b.x": 1, "b.y": 1, words: "text"}));
+    assert.writeOK(coll.insert({a: {x: 1, y: 2}, b: {x: 3, y: 4}, words: "lorem ipsum dolor sit"}));
+    assert.writeOK(coll.insert({a: {x: 1, y: 2}, b: {x: 5, y: 4}, words: "lorem ipsum dolor sit"}));
+
+    assert.eq(1,
+              coll.find({$text: {$search: "lorem ipsum"}, "a.x": 1, "a.y": 2, "b.x": 3, "b.y": 4})
+                  .itcount());
+}());
diff --git a/jstests/core/fts_trailing_fields.js b/jstests/core/fts_trailing_fields.js
new file mode 100644
index 00000000000..2c7f79b423d
--- /dev/null
+++ b/jstests/core/fts_trailing_fields.js
@@ -0,0 +1,22 @@
+// Tests for predicates which can use the trailing field of a text index.
+(function() {
+    "use strict";
+
+    const coll = db.fts_trailing_fields;
+
+    coll.drop();
+    assert.commandWorked(coll.createIndex({a: 1, b: "text", c: 1}));
+    assert.writeOK(coll.insert({a: 2, b: "lorem ipsum"}));
+
+    assert.eq(0, coll.find({a: 2, $text: {$search: "lorem"}, c: {$exists: true}}).itcount());
+    assert.eq(1, coll.find({a: 2, $text: {$search: "lorem"}, c: null}).itcount());
+    assert.eq(1, coll.find({a: 2, $text: {$search: "lorem"}, c: {$exists: false}}).itcount());
+
+    // An equality predicate on the leading field isn't useful, but it shouldn't cause any problems.
+    // Same with an $elemMatch predicate on one of the trailing fields.
+    coll.drop();
+    assert.commandWorked(coll.createIndex({a: 1, b: "text", "c.d": 1}));
+    assert.writeOK(coll.insert({a: 2, b: "lorem ipsum", c: {d: 3}}));
+    assert.eq(0, coll.find({a: [1, 2], $text: {$search: "lorem"}}).itcount());
+    assert.eq(0, coll.find({a: 2, $text: {$search: "lorem"}, c: {$elemMatch: {d: 3}}}).itcount());
+}());
diff --git a/jstests/core/function_string_representations.js b/jstests/core/function_string_representations.js
new file mode 100644
index 00000000000..af66e9160a9
--- /dev/null
+++ b/jstests/core/function_string_representations.js
@@ -0,0 +1,38 @@
+/** Demonstrate that mapReduce can accept functions represented by strings.
+ * Some drivers do not have a type which represents a Javascript function. These languages represent
+ * the arguments to mapReduce as strings.
+ */
+
+(function() {
+    "use strict";
+
+    var col = db.function_string_representations;
+    col.drop();
+    assert.writeOK(col.insert({
+        _id: "abc123",
+        ord_date: new Date("Oct 04, 2012"),
+        status: 'A',
+        price: 25,
+        items: [{sku: "mmm", qty: 5, price: 2.5}, {sku: "nnn", qty: 5, price: 2.5}]
+    }));
+
+    var mapFunction = "function() {emit(this._id, this.price);}";
+    var reduceFunction = "function(keyCustId, valuesPrices) {return Array.sum(valuesPrices);}";
+    assert.commandWorked(col.mapReduce(mapFunction, reduceFunction, {out: "map_reduce_example"}));
+
+    // Provided strings may end with semicolons and/or whitespace
+    mapFunction += " ; ";
+    reduceFunction += " ; ";
+    assert.commandWorked(col.mapReduce(mapFunction, reduceFunction, {out: "map_reduce_example"}));
+
+    // $where exhibits the same behavior
+    var whereFunction = "function() {return this.price === 25;}";
+    assert.eq(1, col.find({$where: whereFunction}).itcount());
+
+    whereFunction += ";";
+    assert.eq(1, col.find({$where: whereFunction}).itcount());
+
+    // db.eval does not need to be tested, as it accepts code fragments, not functions.
+    // system.js does not need to be tested, as its contents types' are preserved, and
+    // strings are not promoted into functions.
+})();
diff --git a/jstests/core/geo_2d_trailing_fields.js b/jstests/core/geo_2d_trailing_fields.js
new file mode 100644
index 00000000000..8f9c881ae4c
--- /dev/null
+++ b/jstests/core/geo_2d_trailing_fields.js
@@ -0,0 +1,45 @@
+// Tests for predicates which can use the trailing field of a 2d index.
+(function() {
+    "use strict";
+
+    const coll = db.geo_2d_trailing_fields;
+
+    const isMaster = assert.commandWorked(db.adminCommand({isMaster: 1}));
+    const isMongos = (isMaster.msg === "isdbgrid");
+
+    coll.drop();
+    assert.commandWorked(coll.createIndex({a: "2d", b: 1}));
+    assert.writeOK(coll.insert({a: [0, 0]}));
+
+    // Verify that $near queries handle existence predicates over the trailing fields correctly.
+    if (!isMongos) {
+        assert.eq(0, coll.find({a: {$near: [0, 0]}, b: {$exists: true}}).itcount());
+        assert.eq(1, coll.find({a: {$near: [0, 0]}, b: null}).itcount());
+        assert.eq(1, coll.find({a: {$near: [0, 0]}, b: {$exists: false}}).itcount());
+    }
+
+    // Verify that non-near 2d queries handle existence predicates over the trailing fields
+    // correctly.
+    assert.eq(0,
+              coll.find({a: {$geoWithin: {$center: [[0, 0], 1]}}, b: {$exists: true}}).itcount());
+    assert.eq(1, coll.find({a: {$geoWithin: {$center: [[0, 0], 1]}}, b: null}).itcount());
+    assert.eq(1,
+              coll.find({a: {$geoWithin: {$center: [[0, 0], 1]}}, b: {$exists: false}}).itcount());
+
+    coll.drop();
+    assert.commandWorked(coll.createIndex({a: "2d", "b.c": 1}));
+    assert.writeOK(coll.insert({a: [0, 0], b: [{c: 2}, {c: 3}]}));
+
+    // Verify that $near queries correctly handle predicates which cannot be covered due to array
+    // semantics.
+    if (!isMongos) {
+        assert.eq(0, coll.find({a: {$near: [0, 0]}, "b.c": [2, 3]}).itcount());
+        assert.eq(0, coll.find({a: {$near: [0, 0]}, "b.c": {$type: "array"}}).itcount());
+    }
+
+    // Verify that non-near 2d queries correctly handle predicates which cannot be covered due to
+    // array semantics.
+    assert.eq(0, coll.find({a: {$geoWithin: {$center: [[0, 0], 1]}}, "b.c": [2, 3]}).itcount());
+    assert.eq(
+        0, coll.find({a: {$geoWithin: {$center: [[0, 0], 1]}}, "b.c": {$type: "array"}}).itcount());
+}());
diff --git a/jstests/core/geo_s2cursorlimitskip.js b/jstests/core/geo_s2cursorlimitskip.js
index 427fbf8fe29..dc645dc68af 100644
--- a/jstests/core/geo_s2cursorlimitskip.js
+++ b/jstests/core/geo_s2cursorlimitskip.js
@@ -1,4 +1,11 @@
 // Test various cursor behaviors
+//
+// @tags: [
+//   # This test attempts to enable profiling on a server and then get profiling data by reading
+//   # from the "system.profile" collection. The former operation must be routed to the primary in
+//   # a replica set, whereas the latter may be routed to a secondary.
+//   assumes_read_preference_unchanged,
+// ]
 
 var testDB = db.getSiblingDB("geo_s2cursorlimitskip");
 var t = testDB.geo_s2getmmm;
diff --git a/jstests/core/geo_update_btree.js b/jstests/core/geo_update_btree.js
index a85d4274415..b4fa24df57e 100644
--- a/jstests/core/geo_update_btree.js
+++ b/jstests/core/geo_update_btree.js
@@ -1,4 +1,6 @@
 // Tests whether the geospatial search is stable under btree updates
+//
+// @tags: [assumes_write_concern_unchanged]
 
 var coll = db.getCollection("jstests_geo_update_btree");
 coll.drop();
diff --git a/jstests/core/getlog2.js b/jstests/core/getlog2.js
index 597a85e20ee..e5287ea8c1b 100644
--- a/jstests/core/getlog2.js
+++ b/jstests/core/getlog2.js
@@ -1,4 +1,11 @@
 // tests getlog as well as slow querying logging
+//
+// @tags: [
+//   # This test attempts to perform a find command and see that it ran using the getLog command.
+//   # The former operation may be routed to a secondary in the replica set, whereas the latter must
+//   # be routed to the primary.
+//   assumes_read_preference_unchanged,
+// ]
 
 glcol = db.getLogTest2;
 glcol.drop();
diff --git a/jstests/core/index_elemmatch2.js b/jstests/core/index_elemmatch2.js
new file mode 100644
index 00000000000..ecd24035284
--- /dev/null
+++ b/jstests/core/index_elemmatch2.js
@@ -0,0 +1,63 @@
+/**
+ * Test that queries containing $elemMatch correctly use an index if each child expression is
+ * compatible with the index.
+ */
+(function() {
+    "use strict";
+
+    load("jstests/libs/analyze_plan.js");
+
+    const coll = db.elemMatch_index;
+    coll.drop();
+
+    assert.writeOK(coll.insert({a: 1}));
+    assert.writeOK(coll.insert({a: [{}]}));
+    assert.writeOK(coll.insert({a: [1, null]}));
+    assert.writeOK(coll.insert({a: [{type: "Point", coordinates: [0, 0]}]}));
+
+    assert.commandWorked(coll.createIndex({a: 1}, {sparse: true}));
+
+    function assertIndexResults(coll, query, useIndex, nReturned) {
+        const explainPlan = coll.find(query).explain("executionStats");
+        assert.eq(isIxscan(explainPlan.queryPlanner.winningPlan), useIndex);
+        assert.eq(explainPlan.executionStats.nReturned, nReturned);
+    }
+
+    assertIndexResults(coll, {a: {$elemMatch: {$exists: false}}}, false, 0);
+
+    // An $elemMatch predicate is treated as nested, and the index should be used for $exists:true.
+    assertIndexResults(coll, {a: {$elemMatch: {$exists: true}}}, true, 3);
+
+    // $not within $elemMatch should not attempt to use a sparse index for $exists:false.
+    assertIndexResults(coll, {a: {$elemMatch: {$not: {$exists: false}}}}, false, 3);
+    assertIndexResults(coll, {a: {$elemMatch: {$gt: 0, $not: {$exists: false}}}}, false, 1);
+
+    // $geo within $elemMatch should not attempt to use a non-geo index.
+    assertIndexResults(
+        coll,
+        {
+          a: {
+              $elemMatch: {
+                  $geoWithin: {
+                      $geometry:
+                          {type: "Polygon", coordinates: [[[0, 0], [0, 1], [1, 0], [0, 0]]]}
+                  }
+              }
+          }
+        },
+        false,
+        1);
+
+    // $in with a null value within $elemMatch should use a sparse index.
+    assertIndexResults(coll, {a: {$elemMatch: {$in: [null]}}}, true, 1);
+
+    // $eq with a null value within $elemMatch should use a sparse index.
+    assertIndexResults(coll, {a: {$elemMatch: {$eq: null}}}, true, 1);
+
+    // A negated regex within $elemMatch should not use an index, sparse or not.
+    assertIndexResults(coll, {a: {$elemMatch: {$not: {$in: [/^a/]}}}}, false, 3);
+
+    coll.dropIndexes();
+    assert.commandWorked(coll.createIndex({a: 1}));
+    assertIndexResults(coll, {a: {$elemMatch: {$not: {$in: [/^a/]}}}}, false, 3);
+})();
diff --git a/jstests/core/index_filter_commands.js b/jstests/core/index_filter_commands.js
index 8684be3b2b9..58f78d0514e 100644
--- a/jstests/core/index_filter_commands.js
+++ b/jstests/core/index_filter_commands.js
@@ -6,20 +6,24 @@
  *   Displays index filters for all query shapes in a collection.
  *
  * - planCacheClearFilters
- *   Clears index filter for a single query shape or,
- *   if the query shape is omitted, all filters for the collection.
+ *   Clears index filter for a single query shape or, if the query shape is omitted, all filters for
+ *   the collection.
  *
  * - planCacheSetFilter
  *   Sets index filter for a query shape. Overrides existing filter.
  *
- * Not a lot of data access in this test suite. Hint commands
- * manage a non-persistent mapping in the server of
- * query shape to list of index specs.
+ * Not a lot of data access in this test suite. Hint commands manage a non-persistent mapping in the
+ * server of query shape to list of index specs.
  *
- * Only time we might need to execute a query is to check the plan
- * cache state. We would do this with the planCacheListPlans command
- * on the same query shape with the index filters.
+ * Only time we might need to execute a query is to check the plan cache state. We would do this
+ * with the planCacheListPlans command on the same query shape with the index filters.
  *
+ * @tags: [
+ *   # This test attempts to perform queries with plan cache filters set up. The former operation
+ *   # may be routed to a secondary in the replica set, whereas the latter must be routed to the
+ *   # primary.
+ *   assumes_read_preference_unchanged,
+ * ]
  */
 
 load("jstests/libs/analyze_plan.js");
diff --git a/jstests/core/index_stats.js b/jstests/core/index_stats.js
index 60b37fd571e..ee4d13d4d0a 100644
--- a/jstests/core/index_stats.js
+++ b/jstests/core/index_stats.js
@@ -1,3 +1,10 @@
+// @tags: [
+//   # This test attempts to perform write operations and get index usage statistics using the
+//   # $indexStats stage. The former operation must be routed to the primary in a replica set,
+//   # whereas the latter may be routed to a secondary.
+//   assumes_read_preference_unchanged,
+// ]
+
 (function() {
     "use strict";
 
diff --git a/jstests/core/js3.js b/jstests/core/js3.js
index 4d46c25bbf7..c808e7ec75a 100644
--- a/jstests/core/js3.js
+++ b/jstests/core/js3.js
@@ -1,3 +1,4 @@
+// @tags: [requires_eval_command]
 
 t = db.jstests_js3;
 
diff --git a/jstests/core/js7.js b/jstests/core/js7.js
index 810f4692d4f..99cddd114f8 100644
--- a/jstests/core/js7.js
+++ b/jstests/core/js7.js
@@ -1,3 +1,5 @@
+// @tags: [requires_eval_command]
+
 t = db.jstests_js7;
 t.drop();
 
diff --git a/jstests/core/js9.js b/jstests/core/js9.js
index 515fa883aea..e0703ac39ea 100644
--- a/jstests/core/js9.js
+++ b/jstests/core/js9.js
@@ -1,3 +1,5 @@
+// @tags: [requires_eval_command]
+
 c = db.jstests_js9;
 c.drop();
 
diff --git a/jstests/core/js_jit.js b/jstests/core/js_jit.js
new file mode 100644
index 00000000000..4ccdd2917ae
--- /dev/null
+++ b/jstests/core/js_jit.js
@@ -0,0 +1,40 @@
+/**
+ * Validate various native types continue to work when run in JITed code.
+ *
+ * In SERVER-30362, the JIT would not compile natives types which had custom getProperty
+ * implementations correctly. We force the JIT to kick in by using large loops.
+ */
+(function() {
+    'use strict';
+
+    function testDBCollection() {
+        const c = new DBCollection(null, null, "foo", "test.foo");
+        for (let i = 0; i < 100000; i++) {
+            if (c.toString() != "test.foo") {
+                throw i;
+            }
+        }
+    }
+
+    function testDB() {
+        const c = new DB(null, "test");
+        for (let i = 0; i < 100000; i++) {
+            if (c.toString() != "test") {
+                throw i;
+            }
+        }
+    }
+
+    function testDBQuery() {
+        const c = DBQuery('a', 'b', 'c', 'd');
+        for (let i = 0; i < 100000; i++) {
+            if (c.toString() != "DBQuery: d -> null") {
+                throw i;
+            }
+        }
+    }
+
+    testDBCollection();
+    testDB();
+    testDBQuery();
+})();
\ No newline at end of file
diff --git a/jstests/core/list_collections1.js b/jstests/core/list_collections1.js
index c8c3f92fbc9..ff1aac304ff 100644
--- a/jstests/core/list_collections1.js
+++ b/jstests/core/list_collections1.js
@@ -72,8 +72,8 @@
     //
 
     var getListCollectionsCursor = function(options, subsequentBatchSize) {
-        return new DBCommandCursor(
-            mydb.getMongo(), mydb.runCommand("listCollections", options), subsequentBatchSize);
+        var res = mydb.runCommand("listCollections", options);
+        return new DBCommandCursor(res._mongo, res, subsequentBatchSize);
     };
 
     var cursorCountMatching = function(cursor, pred) {
@@ -282,9 +282,9 @@
     assert.commandWorked(mydb.createCollection("quux"));
 
     res = mydb.runCommand("listCollections", {cursor: {batchSize: 0}});
-    cursor = new DBCommandCursor(mydb.getMongo(), res, 2);
+    cursor = new DBCommandCursor(res._mongo, res, 2);
     cursor.close();
-    cursor = new DBCommandCursor(mydb.getMongo(), res, 2);
+    cursor = new DBCommandCursor(res._mongo, res, 2);
     assert.throws(function() {
         cursor.hasNext();
     });
diff --git a/jstests/core/list_collections_filter.js b/jstests/core/list_collections_filter.js
index e0d18f055d0..0516d57a417 100644
--- a/jstests/core/list_collections_filter.js
+++ b/jstests/core/list_collections_filter.js
@@ -19,8 +19,8 @@
             filter = {};
         }
 
-        var cursor = new DBCommandCursor(mydb.getMongo(),
-                                         mydb.runCommand("listCollections", {filter: filter}));
+        var res = mydb.runCommand("listCollections", {filter: filter});
+        var cursor = new DBCommandCursor(res._mongo, res);
         function stripToName(result) {
             return result.name;
         }
diff --git a/jstests/core/list_indexes.js b/jstests/core/list_indexes.js
index de0f4473980..6ec9a7a13e1 100644
--- a/jstests/core/list_indexes.js
+++ b/jstests/core/list_indexes.js
@@ -27,8 +27,8 @@
     //
 
     var getListIndexesCursor = function(coll, options, subsequentBatchSize) {
-        return new DBCommandCursor(
-            coll.getDB().getMongo(), coll.runCommand("listIndexes", options), subsequentBatchSize);
+        var res = coll.runCommand("listIndexes", options);
+        return new DBCommandCursor(res._mongo, res, subsequentBatchSize);
     };
 
     var cursorGetIndexSpecs = function(cursor) {
@@ -163,9 +163,9 @@
     assert.commandWorked(coll.ensureIndex({c: 1}, {unique: true}));
 
     res = coll.runCommand("listIndexes", {cursor: {batchSize: 0}});
-    cursor = new DBCommandCursor(coll.getDB().getMongo(), res, 2);
+    cursor = new DBCommandCursor(res._mongo, res, 2);
     cursor.close();
-    cursor = new DBCommandCursor(coll.getDB().getMongo(), res, 2);
+    cursor = new DBCommandCursor(res._mongo, res, 2);
     assert.throws(function() {
         cursor.hasNext();
     });
diff --git a/jstests/core/list_indexes_invalidation.js b/jstests/core/list_indexes_invalidation.js
index b8cbe5eb134..9fe94de5efc 100644
--- a/jstests/core/list_indexes_invalidation.js
+++ b/jstests/core/list_indexes_invalidation.js
@@ -19,7 +19,7 @@
         printjson(res);
 
         // Ensure the cursor has data, rename or drop the collection, and exhaust the cursor.
-        let cursor = new DBCommandCursor(db.getMongo(), res);
+        let cursor = new DBCommandCursor(res._mongo, res);
         let errMsg =
             'expected more data from command ' + tojson(cmd) + ', with result ' + tojson(res);
         assert(cursor.hasNext(), errMsg);
diff --git a/jstests/core/list_namespaces_invalidation.js b/jstests/core/list_namespaces_invalidation.js
index 6f8033b5fe4..85eb26510bb 100644
--- a/jstests/core/list_namespaces_invalidation.js
+++ b/jstests/core/list_namespaces_invalidation.js
@@ -1,4 +1,6 @@
 // SERVER-27996/SERVER-28022 Missing invalidation for system.namespaces writes
+//
+// @tags: [requires_collmod_command]
 (function() {
     'use strict';
     let dbInvalidName = 'system_namespaces_invalidations';
diff --git a/jstests/core/max_time_ms.js b/jstests/core/max_time_ms.js
index 0442ffcba68..efe3dabcc6d 100644
--- a/jstests/core/max_time_ms.js
+++ b/jstests/core/max_time_ms.js
@@ -1,4 +1,12 @@
 // Tests query/command option $maxTimeMS.
+//
+// @tags: [
+//   # This test attempts to perform read operations after having enabled the maxTimeAlwaysTimeOut
+//   # failpoint. The former operations may be routed to a secondary in the replica set, whereas the
+//   # latter must be routed to the primary.
+//   assumes_read_preference_unchanged,
+//   requires_collmod_command,
+// ]
 
 var t = db.max_time_ms;
 var exceededTimeLimit = 50;  // ErrorCodes::ExceededTimeLimit
diff --git a/jstests/core/mr4.js b/jstests/core/mr4.js
index 58ea303f7e8..683583acecd 100644
--- a/jstests/core/mr4.js
+++ b/jstests/core/mr4.js
@@ -9,7 +9,7 @@ t.save({x: 4, tags: ["b", "c"]});
 
 m = function() {
     this.tags.forEach(function(z) {
-        emit(z, {count: xx});
+        emit(z, {count: xx.val});
     });
 };
 
@@ -21,7 +21,7 @@ r = function(key, values) {
     return {count: total};
 };
 
-res = t.mapReduce(m, r, {out: "mr4_out", scope: {xx: 1}});
+res = t.mapReduce(m, r, {out: "mr4_out", scope: {xx: {val: 1}}});
 z = res.convertToSingleObject();
 
 assert.eq(3, Object.keySet(z).length, "A1");
@@ -31,7 +31,7 @@ assert.eq(3, z.c.count, "A4");
 
 res.drop();
 
-res = t.mapReduce(m, r, {scope: {xx: 2}, out: "mr4_out"});
+res = t.mapReduce(m, r, {scope: {xx: {val: 2}}, out: "mr4_out"});
 z = res.convertToSingleObject();
 
 assert.eq(3, Object.keySet(z).length, "A1");
diff --git a/jstests/core/mr_killop.js b/jstests/core/mr_killop.js
index 78e98f0bcaa..e85986e2fc0 100644
--- a/jstests/core/mr_killop.js
+++ b/jstests/core/mr_killop.js
@@ -4,7 +4,7 @@ t = db.jstests_mr_killop;
 t.drop();
 t2 = db.jstests_mr_killop_out;
 t2.drop();
-
+db.adminCommand({"configureFailPoint": 'mr_killop_test_fp', "mode": 'alwaysOn'});
 function debug(x) {
     //        printjson( x );
 }
@@ -171,3 +171,4 @@ var loop = function() {
 };
 runMRTests(loop, false);
 runFinalizeTests(loop, false);
+db.adminCommand({"configureFailPoint": 'mr_killop_test_fp', "mode": 'off'});
diff --git a/jstests/core/mr_optim.js b/jstests/core/mr_optim.js
index 7437753ca67..1c525ae3de3 100644
--- a/jstests/core/mr_optim.js
+++ b/jstests/core/mr_optim.js
@@ -3,8 +3,17 @@
 t = db.mr_optim;
 t.drop();
 
+// We drop the output collection to ensure the test can be run multiple times successfully. We
+// explicitly avoid using the DBCollection#drop() shell helper to avoid implicitly sharding the
+// collection during the sharded_collections_jscore_passthrough.yml test suite when reading the
+// results from the output collection in the reformat() function.
+var res = db.runCommand({drop: "mr_optim_out"});
+if (res.ok !== 1) {
+    assert.commandFailedWithCode(res, ErrorCodes.NamespaceNotFound);
+}
+
 for (var i = 0; i < 1000; ++i) {
-    t.save({a: Math.random(1000), b: Math.random(10000)});
+    assert.writeOK(t.save({a: Math.random(1000), b: Math.random(10000)}));
 }
 
 function m() {
@@ -21,7 +30,7 @@ function reformat(r) {
     if (r.results)
         cursor = r.results;
     else
-        cursor = r.find();
+        cursor = r.find().sort({_id: 1});
     cursor.forEach(function(z) {
         x[z._id] = z.value;
     });
@@ -43,4 +52,4 @@ res.drop();
 
 assert.eq(x, x2, "object from inline and collection are not equal");
 
-t.drop();
\ No newline at end of file
+t.drop();
diff --git a/jstests/core/no_db_created.js b/jstests/core/no_db_created.js
index 3491914d470..b67b193494c 100644
--- a/jstests/core/no_db_created.js
+++ b/jstests/core/no_db_created.js
@@ -1,4 +1,6 @@
 // checks that operations do not create a database
+//
+// @tags: [requires_collmod_command]
 
 (function() {
     "use strict";
@@ -32,4 +34,4 @@
     noDB(mydb);
     assert.writeOK(coll.insert({}));
     mydb.dropDatabase();
-}());
\ No newline at end of file
+}());
diff --git a/jstests/core/notablescan.js b/jstests/core/notablescan.js
index 80306c08cf2..bb4c170a603 100644
--- a/jstests/core/notablescan.js
+++ b/jstests/core/notablescan.js
@@ -1,4 +1,11 @@
 // check notablescan mode
+//
+// @tags: [
+//   # This test attempts to perform read operations after having enabled the notablescan server
+//   # parameter. The former operations may be routed to a secondary in the replica set, whereas the
+//   # latter must be routed to the primary.
+//   assumes_read_preference_unchanged,
+// ]
 
 t = db.test_notablescan;
 t.drop();
diff --git a/jstests/core/operation_latency_histogram.js b/jstests/core/operation_latency_histogram.js
index 1e3f1a59b95..947a8be6520 100644
--- a/jstests/core/operation_latency_histogram.js
+++ b/jstests/core/operation_latency_histogram.js
@@ -1,4 +1,10 @@
 // Checks that histogram counters for collections are updated as we expect.
+//
+// This test attempts to perform write operations and get latency statistics using the $collStats
+// stage. The former operation must be routed to the primary in a replica set, whereas the latter
+// may be routed to a secondary.
+//
+// @tags: [assumes_read_preference_unchanged]
 
 (function() {
     "use strict";
diff --git a/jstests/core/plan_cache_clear.js b/jstests/core/plan_cache_clear.js
index 8f9cf0ea302..778239616b5 100644
--- a/jstests/core/plan_cache_clear.js
+++ b/jstests/core/plan_cache_clear.js
@@ -1,5 +1,12 @@
 // Test clearing of the plan cache, either manually through the planCacheClear command,
 // or due to system events such as an index build.
+//
+// @tags: [
+//   # This test attempts to perform queries and introspect/manipulate the server's plan cache
+//   # entries. The former operation may be routed to a secondary in the replica set, whereas the
+//   # latter must be routed to the primary.
+//   assumes_read_preference_unchanged,
+// ]
 
 var t = db.jstests_plan_cache_clear;
 t.drop();
diff --git a/jstests/core/plan_cache_list_plans.js b/jstests/core/plan_cache_list_plans.js
index 7ca599483ff..3359980ab07 100644
--- a/jstests/core/plan_cache_list_plans.js
+++ b/jstests/core/plan_cache_list_plans.js
@@ -1,4 +1,11 @@
 // Test the planCacheListPlans command.
+//
+// @tags: [
+//   # This test attempts to perform queries and introspect the server's plan cache entries. The
+//   # former operation may be routed to a secondary in the replica set, whereas the latter must be
+//   # routed to the primary.
+//   assumes_read_preference_unchanged,
+// ]
 
 var t = db.jstests_plan_cache_list_plans;
 t.drop();
diff --git a/jstests/core/plan_cache_list_shapes.js b/jstests/core/plan_cache_list_shapes.js
index 1c9ecdf9e1b..61c3111cd8a 100644
--- a/jstests/core/plan_cache_list_shapes.js
+++ b/jstests/core/plan_cache_list_shapes.js
@@ -1,5 +1,12 @@
 // Test the planCacheListQueryShapes command, which returns a list of query shapes
 // for the queries currently cached in the collection.
+//
+// @tags: [
+//   # This test attempts to perform queries with plan cache filters set up. The former operation
+//   # may be routed to a secondary in the replica set, whereas the latter must be routed to the
+//   # primary.
+//   assumes_read_preference_unchanged,
+// ]
 
 var t = db.jstests_plan_cache_list_shapes;
 t.drop();
diff --git a/jstests/core/plan_cache_shell_helpers.js b/jstests/core/plan_cache_shell_helpers.js
index dc990b19dcc..6c4ff185014 100644
--- a/jstests/core/plan_cache_shell_helpers.js
+++ b/jstests/core/plan_cache_shell_helpers.js
@@ -1,4 +1,11 @@
 // Test the shell helpers which wrap the plan cache commands.
+//
+// @tags: [
+//   # This test attempts to perform queries and introspect the server's plan cache entries. The
+//   # former operation may be routed to a secondary in the replica set, whereas the latter must be
+//   # routed to the primary.
+//   assumes_read_preference_unchanged,
+// ]
 
 var t = db.jstests_plan_cache_shell_helpers;
 t.drop();
diff --git a/jstests/core/profile2.js b/jstests/core/profile2.js
index bb1605abd1e..9da6410d2ac 100644
--- a/jstests/core/profile2.js
+++ b/jstests/core/profile2.js
@@ -24,7 +24,7 @@ assert(result.hasOwnProperty('millis'));
 assert(result.hasOwnProperty('query'));
 assert.eq('string', typeof(result.query));
 // String value is truncated.
-assert(result.query.match(/filter: { a: "a+\.\.\." } }$/));
+assert(result.query.match(/filter: { a: "a+\.\.\." }/));
 
 assert.commandWorked(coll.getDB().runCommand({profile: 0}));
 coll.getDB().system.profile.drop();
diff --git a/jstests/core/profile_getmore.js b/jstests/core/profile_getmore.js
index a9272567b1a..3f6f492597e 100644
--- a/jstests/core/profile_getmore.js
+++ b/jstests/core/profile_getmore.js
@@ -24,12 +24,13 @@
     var cursor = coll.find({a: {$gt: 0}}).sort({a: 1}).batchSize(2);
     cursor.next();  // Perform initial query and consume first of 2 docs returned.
 
-    var cursorId = getLatestProfilerEntry(testDB).cursorid;  // Save cursorid from find.
+    var cursorId =
+        getLatestProfilerEntry(testDB, {op: "query"}).cursorid;  // Save cursorid from find.
 
     cursor.next();  // Consume second of 2 docs from initial query.
     cursor.next();  // getMore performed, leaving open cursor.
 
-    var profileObj = getLatestProfilerEntry(testDB);
+    var profileObj = getLatestProfilerEntry(testDB, {op: "getmore"});
 
     assert.eq(profileObj.ns, coll.getFullName(), tojson(profileObj));
     assert.eq(profileObj.op, "getmore", tojson(profileObj));
@@ -67,7 +68,7 @@
     cursor.next();  // Consume second of 2 docs from initial query.
     cursor.next();  // getMore performed, leaving open cursor.
 
-    profileObj = getLatestProfilerEntry(testDB);
+    profileObj = getLatestProfilerEntry(testDB, {op: "getmore"});
 
     assert.eq(profileObj.hasSortStage, true, tojson(profileObj));
 
@@ -83,7 +84,7 @@
     cursor.next();     // Perform initial query and consume first of 3 docs returned.
     cursor.itcount();  // Exhaust the cursor.
 
-    profileObj = getLatestProfilerEntry(testDB);
+    profileObj = getLatestProfilerEntry(testDB, {op: "getmore"});
 
     assert(profileObj.hasOwnProperty("cursorid"),
            tojson(profileObj));  // cursorid should always be present on getMore.
@@ -101,12 +102,12 @@
     assert.commandWorked(coll.createIndex({a: 1}));
 
     var cursor = coll.aggregate([{$match: {a: {$gte: 0}}}], {cursor: {batchSize: 0}});
-    var cursorId = getLatestProfilerEntry(testDB).cursorid;
+    var cursorId = getLatestProfilerEntry(testDB, {"command.aggregate": coll.getName()}).cursorid;
     assert.neq(0, cursorId);
 
     cursor.next();  // Consume the result set.
 
-    profileObj = getLatestProfilerEntry(testDB);
+    profileObj = getLatestProfilerEntry(testDB, {op: "getmore"});
 
     assert.eq(profileObj.ns, coll.getFullName(), tojson(profileObj));
     assert.eq(profileObj.op, "getmore", tojson(profileObj));
diff --git a/jstests/core/profile_insert.js b/jstests/core/profile_insert.js
index 3994c896bda..836ec2ecff2 100644
--- a/jstests/core/profile_insert.js
+++ b/jstests/core/profile_insert.js
@@ -1,4 +1,6 @@
 // Confirms that profiled insert execution contains all expected metrics with proper values.
+//
+// @tags: [assumes_write_concern_unchanged]
 
 (function() {
     "use strict";
diff --git a/jstests/core/recursion.js b/jstests/core/recursion.js
index 926250be20d..1db491c8ae9 100644
--- a/jstests/core/recursion.js
+++ b/jstests/core/recursion.js
@@ -1,5 +1,7 @@
-// Basic tests for a form of stack recursion that's been shown to cause C++
-// side stack overflows in the past. See SERVER-19614.
+// Basic tests for a form of stack recursion that's been shown to cause C++ side stack overflows in
+// the past. See SERVER-19614.
+//
+// @tags: [requires_eval_command]
 
 (function() {
     "use strict";
diff --git a/jstests/core/regex_not_id.js b/jstests/core/regex_not_id.js
index 1f15250f240..35b2c858867 100644
--- a/jstests/core/regex_not_id.js
+++ b/jstests/core/regex_not_id.js
@@ -3,10 +3,10 @@
 var testColl = db.regex_not_id;
 testColl.drop();
 
-assert.writeOK(testColl.insert({_id: "ABCDEF1"}, {writeConcern: {w: 1}}));
+assert.writeOK(testColl.insert({_id: "ABCDEF1"}));
 
 // Should be an error.
-assert.writeError(testColl.insert({_id: /^A/}, {writeConcern: {w: 1}}));
+assert.writeError(testColl.insert({_id: /^A/}));
 
 // _id doesn't have to be first; still disallowed
-assert.writeError(testColl.insert({xxx: "ABCDEF", _id: /ABCDEF/}, {writeConcern: {w: 1}}));
\ No newline at end of file
+assert.writeError(testColl.insert({xxx: "ABCDEF", _id: /ABCDEF/}));
diff --git a/jstests/core/remove8.js b/jstests/core/remove8.js
index 563e4708cf9..3c9fd6a11a1 100644
--- a/jstests/core/remove8.js
+++ b/jstests/core/remove8.js
@@ -1,3 +1,4 @@
+// @tags: [requires_eval_command]
 
 t = db.remove8;
 t.drop();
diff --git a/jstests/core/rename4.js b/jstests/core/rename4.js
index 185193deaa9..756918db5f6 100644
--- a/jstests/core/rename4.js
+++ b/jstests/core/rename4.js
@@ -1,3 +1,5 @@
+// @tags: [requires_eval_command]
+
 t = db.jstests_rename4;
 t.drop();
 
diff --git a/jstests/core/rename_stayTemp.js b/jstests/core/rename_stayTemp.js
index d8451af2d2d..8dee8f89fea 100644
--- a/jstests/core/rename_stayTemp.js
+++ b/jstests/core/rename_stayTemp.js
@@ -11,7 +11,7 @@ function ns(coll) {
 function istemp(name) {
     var result = db.runCommand("listCollections", {filter: {name: name}});
     assert(result.ok);
-    var collections = new DBCommandCursor(db.getMongo(), result).toArray();
+    var collections = new DBCommandCursor(result._mongo, result).toArray();
     assert.eq(1, collections.length);
     return collections[0].options.temp ? true : false;
 }
diff --git a/jstests/core/shell_connection_strings.js b/jstests/core/shell_connection_strings.js
new file mode 100644
index 00000000000..22861e6ce25
--- /dev/null
+++ b/jstests/core/shell_connection_strings.js
@@ -0,0 +1,33 @@
+// Test mongo shell connect strings.
+(function() {
+    'use strict';
+
+    const mongod = new MongoURI(db.getMongo().host).servers[0];
+    const host = mongod.host;
+    const port = mongod.port;
+
+    function testConnect(ok, ...args) {
+        const exitCode = runMongoProgram('mongo', '--eval', ';', ...args);
+        if (ok) {
+            assert.eq(exitCode, 0, "failed to connect with `" + args.join(' ') + "`");
+        } else {
+            assert.neq(
+                exitCode, 0, "unexpectedly succeeded connecting with `" + args.join(' ') + "`");
+        }
+    }
+
+    testConnect(true, `${host}:${port}`);
+    testConnect(true, `${host}:${port}/test`);
+    testConnect(true, `${host}:${port}/admin`);
+    testConnect(true, host, '--port', port);
+    testConnect(true, '--host', host, '--port', port, 'test');
+    testConnect(true, '--host', host, '--port', port, 'admin');
+    testConnect(true, `mongodb://${host}:${port}/test`);
+    testConnect(true, `mongodb://${host}:${port}/test?connectTimeoutMS=10000`);
+
+    // if a full URI is provided, you cannot also specify host or port
+    testConnect(false, `${host}/test`, '--port', port);
+    testConnect(false, `mongodb://${host}:${port}/test`, '--port', port);
+    testConnect(false, `mongodb://${host}:${port}/test`, '--host', host);
+    testConnect(false, `mongodb://${host}:${port}/test`, '--host', host, '--port', port);
+})();
diff --git a/jstests/core/shell_writeconcern.js b/jstests/core/shell_writeconcern.js
index f3f190061cf..e3e7a23a9aa 100644
--- a/jstests/core/shell_writeconcern.js
+++ b/jstests/core/shell_writeconcern.js
@@ -1,7 +1,10 @@
 "use strict";
+
 // check that shell writeconcern work correctly
 // 1.) tests that it can be set on each level and is inherited
 // 2.) tests that each operation (update/insert/remove/save) take and ensure a write concern
+//
+// @tags: [assumes_write_concern_unchanged]
 
 var collA = db.shell_wc_a;
 var collB = db.shell_wc_b;
diff --git a/jstests/core/stages_delete.js b/jstests/core/stages_delete.js
index f8e7380c75a..98a270cedf3 100644
--- a/jstests/core/stages_delete.js
+++ b/jstests/core/stages_delete.js
@@ -1,3 +1,9 @@
+// @tags: [
+//   # This test attempts to remove documents using the stageDebug command, which doesn't support
+//   # specifying a writeConcern.
+//   assumes_write_concern_unchanged,
+// ]
+
 // Test basic delete stage functionality.
 var coll = db.stages_delete;
 var collScanStage = {cscan: {args: {direction: 1}, filter: {deleteMe: true}}};
diff --git a/jstests/core/startup_log.js b/jstests/core/startup_log.js
index 3b0cbe3464d..c73013d1744 100644
--- a/jstests/core/startup_log.js
+++ b/jstests/core/startup_log.js
@@ -1,101 +1,108 @@
-load('jstests/aggregation/extras/utils.js');

-

-(function() {

-    'use strict';

-

-    // Check that smallArray is entirely contained by largeArray

-    // returns false if a member of smallArray is not in largeArray

-    function arrayIsSubset(smallArray, largeArray) {

-        for (var i = 0; i < smallArray.length; i++) {

-            if (!Array.contains(largeArray, smallArray[i])) {

-                print("Could not find " + smallArray[i] + " in largeArray");

-                return false;

-            }

-        }

-

-        return true;

-    }

-

-    // Test startup_log

-    var stats = db.getSisterDB("local").startup_log.stats();

-    assert(stats.capped);

-

-    var latestStartUpLog =

-        db.getSisterDB("local").startup_log.find().sort({$natural: -1}).limit(1).next();

-    var serverStatus = db._adminCommand("serverStatus");

-    var cmdLine = db._adminCommand("getCmdLineOpts").parsed;

-

-    // Test that the startup log has the expected keys

-    var verbose = false;

-    var expectedKeys =

-        ["_id", "hostname", "startTime", "startTimeLocal", "cmdLine", "pid", "buildinfo"];

-    var keys = Object.keySet(latestStartUpLog);

-    assert(arrayEq(expectedKeys, keys, verbose), 'startup_log keys failed');

-

-    // Tests _id implicitly - should be comprised of host-timestamp

-    // Setup expected startTime and startTimeLocal from the supplied timestamp

-    var _id = latestStartUpLog._id.split('-');  // _id should consist of host-timestamp

-    var _idUptime = _id.pop();

-    var _idHost = _id.join('-');

-    var uptimeSinceEpochRounded = Math.floor(_idUptime / 1000) * 1000;

-    var startTime = new Date(uptimeSinceEpochRounded);  // Expected startTime

-

-    assert.eq(_idHost, latestStartUpLog.hostname, "Hostname doesn't match one from _id");

-    assert.eq(serverStatus.host.split(':')[0],

-              latestStartUpLog.hostname,

-              "Hostname doesn't match one in server status");

-    assert.closeWithinMS(startTime,

-                         latestStartUpLog.startTime,

-                         "StartTime doesn't match one from _id",

-                         2000);  // Expect less than 2 sec delta

-    assert.eq(cmdLine, latestStartUpLog.cmdLine, "cmdLine doesn't match that from getCmdLineOpts");

-    assert.eq(serverStatus.pid, latestStartUpLog.pid, "pid doesn't match that from serverStatus");

-

-    // Test buildinfo

-    var buildinfo = db.runCommand("buildinfo");

-    delete buildinfo.ok;  // Delete extra meta info not in startup_log

-    var isMaster = db._adminCommand("ismaster");

-

-    // Test buildinfo has the expected keys

-    var expectedKeys = [

-        "version",

-        "gitVersion",

-        "allocator",

-        "versionArray",

-        "javascriptEngine",

-        "openssl",

-        "buildEnvironment",

-        "debug",

-        "maxBsonObjectSize",

-        "bits",

-        "modules"

-    ];

-

-    var keys = Object.keySet(latestStartUpLog.buildinfo);

-    // Disabled to check

-    assert(arrayIsSubset(expectedKeys, keys),

-           "buildinfo keys failed! \n expected:\t" + expectedKeys + "\n actual:\t" + keys);

-    assert.eq(buildinfo,

-              latestStartUpLog.buildinfo,

-              "buildinfo doesn't match that from buildinfo command");

-

-    // Test version and version Array

-    var version = latestStartUpLog.buildinfo.version.split('-')[0];

-    var versionArray = latestStartUpLog.buildinfo.versionArray;

-    var versionArrayCleaned = versionArray.slice(0, 3);

-    if (versionArray[3] == -100) {

-        versionArrayCleaned[2] -= 1;

-    }

-

-    assert.eq(serverStatus.version,

-              latestStartUpLog.buildinfo.version,

-              "Mongo version doesn't match that from ServerStatus");

-    assert.eq(

-        version, versionArrayCleaned.join('.'), "version doesn't match that from the versionArray");

-    var jsEngine = latestStartUpLog.buildinfo.javascriptEngine;

-    assert((jsEngine == "none") || jsEngine.startsWith("mozjs"));

-    assert.eq(isMaster.maxBsonObjectSize,

-              latestStartUpLog.buildinfo.maxBsonObjectSize,

-              "maxBsonObjectSize doesn't match one from ismaster");

-

-})();

+/**
+ * This test attempts to read from the "local.startup_log" collection and assert that it has an
+ * entry matching the server's response from the "getCmdLineOpts" command. The former operation may
+ * be routed to a secondary in the replica set, whereas the latter must be routed to the primary.
+ *
+ * @tags: [assumes_read_preference_unchanged]
+ */
+load('jstests/aggregation/extras/utils.js');
+
+(function() {
+    'use strict';
+
+    // Check that smallArray is entirely contained by largeArray
+    // returns false if a member of smallArray is not in largeArray
+    function arrayIsSubset(smallArray, largeArray) {
+        for (var i = 0; i < smallArray.length; i++) {
+            if (!Array.contains(largeArray, smallArray[i])) {
+                print("Could not find " + smallArray[i] + " in largeArray");
+                return false;
+            }
+        }
+
+        return true;
+    }
+
+    // Test startup_log
+    var stats = db.getSisterDB("local").startup_log.stats();
+    assert(stats.capped);
+
+    var latestStartUpLog =
+        db.getSisterDB("local").startup_log.find().sort({$natural: -1}).limit(1).next();
+    var serverStatus = db._adminCommand("serverStatus");
+    var cmdLine = db._adminCommand("getCmdLineOpts").parsed;
+
+    // Test that the startup log has the expected keys
+    var verbose = false;
+    var expectedKeys =
+        ["_id", "hostname", "startTime", "startTimeLocal", "cmdLine", "pid", "buildinfo"];
+    var keys = Object.keySet(latestStartUpLog);
+    assert(arrayEq(expectedKeys, keys, verbose), 'startup_log keys failed');
+
+    // Tests _id implicitly - should be comprised of host-timestamp
+    // Setup expected startTime and startTimeLocal from the supplied timestamp
+    var _id = latestStartUpLog._id.split('-');  // _id should consist of host-timestamp
+    var _idUptime = _id.pop();
+    var _idHost = _id.join('-');
+    var uptimeSinceEpochRounded = Math.floor(_idUptime / 1000) * 1000;
+    var startTime = new Date(uptimeSinceEpochRounded);  // Expected startTime
+
+    assert.eq(_idHost, latestStartUpLog.hostname, "Hostname doesn't match one from _id");
+    assert.eq(serverStatus.host.split(':')[0],
+              latestStartUpLog.hostname,
+              "Hostname doesn't match one in server status");
+    assert.closeWithinMS(startTime,
+                         latestStartUpLog.startTime,
+                         "StartTime doesn't match one from _id",
+                         2000);  // Expect less than 2 sec delta
+    assert.eq(cmdLine, latestStartUpLog.cmdLine, "cmdLine doesn't match that from getCmdLineOpts");
+    assert.eq(serverStatus.pid, latestStartUpLog.pid, "pid doesn't match that from serverStatus");
+
+    // Test buildinfo
+    var buildinfo = db.runCommand("buildinfo");
+    delete buildinfo.ok;  // Delete extra meta info not in startup_log
+    var isMaster = db._adminCommand("ismaster");
+
+    // Test buildinfo has the expected keys
+    var expectedKeys = [
+        "version",
+        "gitVersion",
+        "allocator",
+        "versionArray",
+        "javascriptEngine",
+        "openssl",
+        "buildEnvironment",
+        "debug",
+        "maxBsonObjectSize",
+        "bits",
+        "modules"
+    ];
+
+    var keys = Object.keySet(latestStartUpLog.buildinfo);
+    // Disabled to check
+    assert(arrayIsSubset(expectedKeys, keys),
+           "buildinfo keys failed! \n expected:\t" + expectedKeys + "\n actual:\t" + keys);
+    assert.eq(buildinfo,
+              latestStartUpLog.buildinfo,
+              "buildinfo doesn't match that from buildinfo command");
+
+    // Test version and version Array
+    var version = latestStartUpLog.buildinfo.version.split('-')[0];
+    var versionArray = latestStartUpLog.buildinfo.versionArray;
+    var versionArrayCleaned = versionArray.slice(0, 3);
+    if (versionArray[3] == -100) {
+        versionArrayCleaned[2] -= 1;
+    }
+
+    assert.eq(serverStatus.version,
+              latestStartUpLog.buildinfo.version,
+              "Mongo version doesn't match that from ServerStatus");
+    assert.eq(
+        version, versionArrayCleaned.join('.'), "version doesn't match that from the versionArray");
+    var jsEngine = latestStartUpLog.buildinfo.javascriptEngine;
+    assert((jsEngine == "none") || jsEngine.startsWith("mozjs"));
+    assert.eq(isMaster.maxBsonObjectSize,
+              latestStartUpLog.buildinfo.maxBsonObjectSize,
+              "maxBsonObjectSize doesn't match one from ismaster");
+
+})();
diff --git a/jstests/core/storefunc.js b/jstests/core/storefunc.js
index 8598e9cc62b..15abc56421e 100644
--- a/jstests/core/storefunc.js
+++ b/jstests/core/storefunc.js
@@ -1,3 +1,5 @@
+// @tags: [requires_eval_command]
+
 // Use a private sister database to avoid conflicts with other tests that use system.js
 var testdb = db.getSisterDB("storefunc");
 var res;
diff --git a/jstests/core/top.js b/jstests/core/top.js
index 819b41b0981..3d98f5a7b2d 100644
--- a/jstests/core/top.js
+++ b/jstests/core/top.js
@@ -1,5 +1,11 @@
 /**
  * 1. check top numbers are correct
+ *
+ * This test attempts to perform read operations and get statistics using the top command. The
+ * former operation may be routed to a secondary in the replica set, whereas the latter must be
+ * routed to the primary.
+ *
+ * @tags: [assumes_read_preference_unchanged]
  */
 (function() {
     load("jstests/libs/stats.js");
diff --git a/jstests/core/update_affects_indexes.js b/jstests/core/update_affects_indexes.js
new file mode 100644
index 00000000000..91db3ebe565
--- /dev/null
+++ b/jstests/core/update_affects_indexes.js
@@ -0,0 +1,100 @@
+// This is a regression test for SERVER-32048. It checks that index keys are correctly updated when
+// an update modifier implicitly creates a new array element.
+(function() {
+    "use strict";
+
+    let coll = db.update_affects_indexes;
+    coll.drop();
+    let indexKeyPattern = {"a.b": 1};
+    assert.commandWorked(coll.createIndex(indexKeyPattern));
+
+    // Tests that the document 'docId' has all the index keys in 'expectedKeys' and none of the
+    // index keys in 'unexpectedKeys'.
+    function assertExpectedIndexKeys(docId, expectedKeys, unexpectedKeys) {
+        for (let key of expectedKeys) {
+            let res = coll.find(docId).hint(indexKeyPattern).min(key).returnKey().toArray();
+            assert.eq(1, res.length, tojson(res));
+            assert.eq(key, res[0]);
+        }
+
+        for (let key of unexpectedKeys) {
+            let res = coll.find(docId).hint(indexKeyPattern).min(key).returnKey().toArray();
+            if (res.length > 0) {
+                assert.eq(1, res.length, tojson(res));
+                assert.neq(key, res[0]);
+            }
+        }
+    }
+
+    // $set implicitly creates array element at end of array.
+    assert.writeOK(coll.insert({_id: 0, a: [{b: 0}]}));
+    assertExpectedIndexKeys({_id: 0}, [{"a.b": 0}], [{"a.b": null}]);
+    assert.writeOK(coll.update({_id: 0}, {$set: {"a.1.c": 0}}));
+    assertExpectedIndexKeys({_id: 0}, [{"a.b": 0}, {"a.b": null}], []);
+
+    // $set implicitly creates array element beyond end of array.
+    assert.writeOK(coll.insert({_id: 1, a: [{b: 0}]}));
+    assertExpectedIndexKeys({_id: 1}, [{"a.b": 0}], [{"a.b": null}]);
+    assert.writeOK(coll.update({_id: 1}, {$set: {"a.3.c": 0}}));
+    assertExpectedIndexKeys({_id: 1}, [{"a.b": 0}, {"a.b": null}], []);
+
+    // $set implicitly creates array element in empty array (no index key changes needed).
+    assert.writeOK(coll.insert({_id: 2, a: []}));
+    assertExpectedIndexKeys({_id: 2}, [{"a.b": null}], []);
+    assert.writeOK(coll.update({_id: 2}, {$set: {"a.0.c": 0}}));
+    assertExpectedIndexKeys({_id: 2}, [{"a.b": null}], []);
+
+    // $inc implicitly creates array element at end of array.
+    assert.writeOK(coll.insert({_id: 3, a: [{b: 0}]}));
+    assertExpectedIndexKeys({_id: 3}, [{"a.b": 0}], [{"a.b": null}]);
+    assert.writeOK(coll.update({_id: 3}, {$inc: {"a.1.c": 0}}));
+    assertExpectedIndexKeys({_id: 3}, [{"a.b": 0}, {"a.b": null}], []);
+
+    // $mul implicitly creates array element at end of array.
+    assert.writeOK(coll.insert({_id: 4, a: [{b: 0}]}));
+    assertExpectedIndexKeys({_id: 4}, [{"a.b": 0}], [{"a.b": null}]);
+    assert.writeOK(coll.update({_id: 4}, {$mul: {"a.1.c": 0}}));
+    assertExpectedIndexKeys({_id: 4}, [{"a.b": 0}, {"a.b": null}], []);
+
+    // $addToSet implicitly creates array element at end of array.
+    assert.writeOK(coll.insert({_id: 5, a: [{b: 0}]}));
+    assertExpectedIndexKeys({_id: 5}, [{"a.b": 0}], [{"a.b": null}]);
+    assert.writeOK(coll.update({_id: 5}, {$addToSet: {"a.1.c": 0}}));
+    assertExpectedIndexKeys({_id: 5}, [{"a.b": 0}, {"a.b": null}], []);
+
+    // $bit implicitly creates array element at end of array.
+    assert.writeOK(coll.insert({_id: 6, a: [{b: 0}]}));
+    assertExpectedIndexKeys({_id: 6}, [{"a.b": 0}], [{"a.b": null}]);
+    assert.writeOK(coll.update({_id: 6}, {$bit: {"a.1.c": {and: NumberInt(1)}}}));
+    assertExpectedIndexKeys({_id: 6}, [{"a.b": 0}, {"a.b": null}], []);
+
+    // $min implicitly creates array element at end of array.
+    assert.writeOK(coll.insert({_id: 7, a: [{b: 0}]}));
+    assertExpectedIndexKeys({_id: 7}, [{"a.b": 0}], [{"a.b": null}]);
+    assert.writeOK(coll.update({_id: 7}, {$min: {"a.1.c": 0}}));
+    assertExpectedIndexKeys({_id: 7}, [{"a.b": 0}, {"a.b": null}], []);
+
+    // $max implicitly creates array element at end of array.
+    assert.writeOK(coll.insert({_id: 8, a: [{b: 0}]}));
+    assertExpectedIndexKeys({_id: 8}, [{"a.b": 0}], [{"a.b": null}]);
+    assert.writeOK(coll.update({_id: 8}, {$max: {"a.1.c": 0}}));
+    assertExpectedIndexKeys({_id: 8}, [{"a.b": 0}, {"a.b": null}], []);
+
+    // $currentDate implicitly creates array element at end of array.
+    assert.writeOK(coll.insert({_id: 9, a: [{b: 0}]}));
+    assertExpectedIndexKeys({_id: 9}, [{"a.b": 0}], [{"a.b": null}]);
+    assert.writeOK(coll.update({_id: 9}, {$currentDate: {"a.1.c": true}}));
+    assertExpectedIndexKeys({_id: 9}, [{"a.b": 0}, {"a.b": null}], []);
+
+    // $push implicitly creates array element at end of array.
+    assert.writeOK(coll.insert({_id: 10, a: [{b: 0}]}));
+    assertExpectedIndexKeys({_id: 10}, [{"a.b": 0}], [{"a.b": null}]);
+    assert.writeOK(coll.update({_id: 10}, {$push: {"a.1.c": 0}}));
+    assertExpectedIndexKeys({_id: 10}, [{"a.b": 0}, {"a.b": null}], []);
+
+    // $pushAll implicitly creates array element at end of array.
+    assert.writeOK(coll.insert({_id: 11, a: [{b: 0}]}));
+    assertExpectedIndexKeys({_id: 11}, [{"a.b": 0}], [{"a.b": null}]);
+    assert.writeOK(coll.update({_id: 11}, {$pushAll: {"a.1.c": [0]}}));
+    assertExpectedIndexKeys({_id: 11}, [{"a.b": 0}, {"a.b": null}], []);
+}());
diff --git a/jstests/core/update_multi5.js b/jstests/core/update_multi5.js
index e610462a620..9e9550bc554 100644
--- a/jstests/core/update_multi5.js
+++ b/jstests/core/update_multi5.js
@@ -7,8 +7,8 @@
     assert.writeOK(t.insert({path: 'r1', subscribers: [1, 2]}));
     assert.writeOK(t.insert({path: 'r2', subscribers: [3, 4]}));
 
-    var res = assert.writeOK(t.update(
-        {}, {$addToSet: {subscribers: 5}}, {upsert: false, multi: true, writeConcern: {w: 1}}));
+    var res =
+        assert.writeOK(t.update({}, {$addToSet: {subscribers: 5}}, {upsert: false, multi: true}));
 
     assert.eq(res.nMatched, 2, tojson(res));
 
diff --git a/jstests/core/views/invalid_system_views.js b/jstests/core/views/invalid_system_views.js
index 3ba282d2ca1..49ac600ba8b 100644
--- a/jstests/core/views/invalid_system_views.js
+++ b/jstests/core/views/invalid_system_views.js
@@ -1,6 +1,8 @@
 /**
  * Tests that invalid view definitions in system.views do not impact valid commands on existing
  * collections.
+ *
+ * @tags: [requires_collmod_command]
  */
 (function() {
     "use strict";
diff --git a/jstests/core/views/views_all_commands.js b/jstests/core/views/views_all_commands.js
index 0115672dcda..e1798803a1b 100644
--- a/jstests/core/views/views_all_commands.js
+++ b/jstests/core/views/views_all_commands.js
@@ -48,6 +48,8 @@
  *
  *  skipStandalone
  *      If true, do not run this command on a standalone mongod.
+ *
+ * @tags: [requires_collmod_command]
  */
 
 (function() {
diff --git a/jstests/core/views/views_basic.js b/jstests/core/views/views_basic.js
index d6032b105df..b4a2f177982 100644
--- a/jstests/core/views/views_basic.js
+++ b/jstests/core/views/views_basic.js
@@ -13,7 +13,7 @@
         let res = viewsDB.runCommand(cmd);
         assert.commandWorked(res);
 
-        let cursor = new DBCommandCursor(db.getMongo(), res, 5);
+        let cursor = new DBCommandCursor(res._mongo, res, 5);
         let actual = cursor.toArray();
         assert(arrayEq(actual, expected),
                "actual: " + tojson(cursor.toArray()) + ", expected:" + tojson(expected));
diff --git a/jstests/core/views/views_change.js b/jstests/core/views/views_change.js
index 002284095c5..a62e7e7f04f 100644
--- a/jstests/core/views/views_change.js
+++ b/jstests/core/views/views_change.js
@@ -1,6 +1,7 @@
 /**
  * Tests the behavior of views when the backing view or collection is changed.
- * @tags: [requires_find_command]
+ *
+ * @tags: [requires_collmod_command, requires_find_command]
  */
 (function() {
     "use strict";
diff --git a/jstests/core/views/views_collation.js b/jstests/core/views/views_collation.js
index 9e4ed7feb30..38557358ce9 100644
--- a/jstests/core/views/views_collation.js
+++ b/jstests/core/views/views_collation.js
@@ -1,9 +1,13 @@
 /**
  * Tests the behavior of operations when interacting with a view's default collation.
+ *
+ * @tags: [requires_collmod_command]
  */
 (function() {
     "use strict";
 
+    load("jstests/libs/analyze_plan.js");
+
     let viewsDB = db.getSiblingDB("views_collation");
     assert.commandWorked(viewsDB.dropDatabase());
     assert.commandWorked(viewsDB.runCommand({create: "simpleCollection"}));
@@ -58,6 +62,15 @@
     assert.commandWorked(viewsDB.runCommand({count: "filView"}));
     assert.commandWorked(viewsDB.runCommand({distinct: "filView", key: "x"}));
 
+    // Explain of operations that do not specify a collation succeed.
+    assert.commandWorked(viewsDB.runCommand({aggregate: "filView", pipeline: [], explain: true}));
+    assert.commandWorked(
+        viewsDB.runCommand({explain: {find: "filView"}, verbosity: "allPlansExecution"}));
+    assert.commandWorked(
+        viewsDB.runCommand({explain: {count: "filView"}, verbosity: "allPlansExecution"}));
+    assert.commandWorked(viewsDB.runCommand(
+        {explain: {distinct: "filView", key: "x"}, verbosity: "allPlansExecution"}));
+
     // Operations with a matching collation succeed.
     assert.commandWorked(
         viewsDB.runCommand({aggregate: "filView", pipeline: [], collation: {locale: "fil"}}));
@@ -66,6 +79,18 @@
     assert.commandWorked(
         viewsDB.runCommand({distinct: "filView", key: "x", collation: {locale: "fil"}}));
 
+    // Explain of operations with a matching collation succeed.
+    assert.commandWorked(viewsDB.runCommand(
+        {aggregate: "filView", pipeline: [], explain: true, collation: {locale: "fil"}}));
+    assert.commandWorked(viewsDB.runCommand(
+        {explain: {find: "filView", collation: {locale: "fil"}}, verbosity: "allPlansExecution"}));
+    assert.commandWorked(viewsDB.runCommand(
+        {explain: {count: "filView", collation: {locale: "fil"}}, verbosity: "allPlansExecution"}));
+    assert.commandWorked(viewsDB.runCommand({
+        explain: {distinct: "filView", key: "x", collation: {locale: "fil"}},
+        verbosity: "allPlansExecution"
+    }));
+
     // Attempting to override the non-simple default collation of a view fails.
     assert.commandFailedWithCode(
         viewsDB.runCommand({aggregate: "filView", pipeline: [], collation: {locale: "en"}}),
@@ -90,6 +115,46 @@
         viewsDB.runCommand({distinct: "filView", key: "x", collation: {locale: "simple"}}),
         ErrorCodes.OptionNotSupportedOnView);
 
+    // Attempting to override the default collation of a view with explain fails.
+    assert.commandFailedWithCode(
+        viewsDB.runCommand(
+            {aggregate: "filView", pipeline: [], explain: true, collation: {locale: "en"}}),
+        ErrorCodes.OptionNotSupportedOnView);
+    assert.commandFailedWithCode(
+        viewsDB.runCommand(
+            {aggregate: "filView", pipeline: [], explain: true, collation: {locale: "simple"}}),
+        ErrorCodes.OptionNotSupportedOnView);
+    assert.commandFailedWithCode(viewsDB.runCommand({
+        explain: {find: "filView", collation: {locale: "fr"}},
+        verbosity: "allPlansExecution"
+    }),
+                                 ErrorCodes.OptionNotSupportedOnView);
+    assert.commandFailedWithCode(viewsDB.runCommand({
+        explain: {find: "filView", collation: {locale: "simple"}},
+        verbosity: "allPlansExecution"
+    }),
+                                 ErrorCodes.OptionNotSupportedOnView);
+    assert.commandFailedWithCode(viewsDB.runCommand({
+        explain: {count: "filView", collation: {locale: "zh"}},
+        verbosity: "allPlansExecution"
+    }),
+                                 ErrorCodes.OptionNotSupportedOnView);
+    assert.commandFailedWithCode(viewsDB.runCommand({
+        explain: {count: "filView", collation: {locale: "simple"}},
+        verbosity: "allPlansExecution"
+    }),
+                                 ErrorCodes.OptionNotSupportedOnView);
+    assert.commandFailedWithCode(viewsDB.runCommand({
+        explain: {distinct: "filView", key: "x", collation: {locale: "es"}},
+        verbosity: "allPlansExecution"
+    }),
+                                 ErrorCodes.OptionNotSupportedOnView);
+    assert.commandFailedWithCode(viewsDB.runCommand({
+        explain: {distinct: "filView", key: "x", collation: {locale: "simple"}},
+        verbosity: "allPlansExecution"
+    }),
+                                 ErrorCodes.OptionNotSupportedOnView);
+
     const lookupSimpleView = {
         $lookup: {from: "simpleView", localField: "x", foreignField: "x", as: "result"}
     };
@@ -253,4 +318,56 @@
         viewsDB.runCommand(
             {collMod: "esView", viewOn: "simpleCollection", pipeline: [graphLookupFilView]}),
         ErrorCodes.OptionNotSupportedOnView);
+
+    // Make sure that when an operation does not specify the collation, it correctly uses the
+    // default collation associated with the view. For this, we set up a new backing collection with
+    // a case-insensitive view.
+    assert.commandWorked(viewsDB.runCommand({create: "case_sensitive_coll"}));
+    assert.commandWorked(viewsDB.runCommand({
+        create: "case_insensitive_view",
+        viewOn: "case_sensitive_coll",
+        collation: {locale: "en", strength: 1}
+    }));
+
+    assert.writeOK(viewsDB.case_sensitive_coll.insert({f: "case"}));
+    assert.writeOK(viewsDB.case_sensitive_coll.insert({f: "Case"}));
+    assert.writeOK(viewsDB.case_sensitive_coll.insert({f: "CASE"}));
+
+    let explain, cursorStage;
+
+    // Test that aggregate against a view with a default collation correctly uses the collation.
+    assert.eq(1, viewsDB.case_sensitive_coll.aggregate([{$match: {f: "case"}}]).itcount());
+    assert.eq(3, viewsDB.case_insensitive_view.aggregate([{$match: {f: "case"}}]).itcount());
+    explain = viewsDB.case_insensitive_view.explain().aggregate([{$match: {f: "case"}}]);
+    cursorStage = getAggPlanStage(explain, "$cursor");
+    assert.neq(null, cursorStage, tojson(explain));
+    assert.eq(1, cursorStage.$cursor.queryPlanner.collation.strength, tojson(cursorStage));
+
+    // Test that count against a view with a default collation correctly uses the collation.
+    assert.eq(1, viewsDB.case_sensitive_coll.count({f: "case"}));
+    assert.eq(3, viewsDB.case_insensitive_view.count({f: "case"}));
+    explain = viewsDB.case_insensitive_view.explain().count({f: "case"});
+    cursorStage = getAggPlanStage(explain, "$cursor");
+    assert.neq(null, cursorStage, tojson(explain));
+    assert.eq(1, cursorStage.$cursor.queryPlanner.collation.strength, tojson(cursorStage));
+
+    // Test that distinct against a view with a default collation correctly uses the collation.
+    assert.eq(3, viewsDB.case_sensitive_coll.distinct("f").length);
+    assert.eq(1, viewsDB.case_insensitive_view.distinct("f").length);
+    explain = viewsDB.case_insensitive_view.explain().distinct("f");
+    cursorStage = getAggPlanStage(explain, "$cursor");
+    assert.neq(null, cursorStage, tojson(explain));
+    assert.eq(1, cursorStage.$cursor.queryPlanner.collation.strength, tojson(cursorStage));
+
+    // Test that find against a view with a default collation correctly uses the collation.
+    let findRes = viewsDB.runCommand({find: "case_sensitive_coll", filter: {f: "case"}});
+    assert.commandWorked(findRes);
+    assert.eq(1, findRes.cursor.firstBatch.length);
+    findRes = viewsDB.runCommand({find: "case_insensitive_view", filter: {f: "case"}});
+    assert.commandWorked(findRes);
+    assert.eq(3, findRes.cursor.firstBatch.length);
+    explain = viewsDB.runCommand({explain: {find: "case_insensitive_view", filter: {f: "case"}}});
+    cursorStage = getAggPlanStage(explain, "$cursor");
+    assert.neq(null, cursorStage, tojson(explain));
+    assert.eq(1, cursorStage.$cursor.queryPlanner.collation.strength, tojson(cursorStage));
 }());
diff --git a/jstests/core/views/views_find.js b/jstests/core/views/views_find.js
index c196e980ce5..53c9d4e8b98 100644
--- a/jstests/core/views/views_find.js
+++ b/jstests/core/views/views_find.js
@@ -15,7 +15,7 @@
     let assertFindResultEq = function(cmd, expected, ordered) {
         let res = viewsDB.runCommand(cmd);
         assert.commandWorked(res);
-        let arr = new DBCommandCursor(db.getMongo(), res, 5).toArray();
+        let arr = new DBCommandCursor(res._mongo, res, 5).toArray();
         let errmsg = tojson({expected: expected, got: arr});
 
         if (typeof(ordered) === "undefined" || !ordered)
diff --git a/jstests/core/views/views_rename.js b/jstests/core/views/views_rename.js
new file mode 100644
index 00000000000..ad656671ea1
--- /dev/null
+++ b/jstests/core/views/views_rename.js
@@ -0,0 +1,17 @@
+(function() {
+    // SERVER-30406 Test that renaming system.views correctly invalidates the view catalog
+    'use strict';
+    db.view.drop();
+    db.coll.drop();
+    assert.commandWorked(db.createView("view", "coll", []));
+    assert.writeOK(db.coll.insert({_id: 1}));
+    assert.eq(db.view.find().count(), 1, "couldn't find document in view");
+    assert.commandWorked(db.system.views.renameCollection("views", /*dropTarget*/ true));
+    assert.eq(db.view.find().count(),
+              0,
+              "find on view should have returned no results after renaming away system.views");
+    assert.commandWorked(db.views.renameCollection("system.views"));
+    assert.eq(db.view.find().count(),
+              1,
+              "find on view should have worked again after renaming system.views back in place");
+})();
diff --git a/jstests/core/views/views_stats.js b/jstests/core/views/views_stats.js
index 75feb857c9a..22261e9fa81 100644
--- a/jstests/core/views/views_stats.js
+++ b/jstests/core/views/views_stats.js
@@ -1,4 +1,10 @@
 // Test that top and latency histogram statistics are recorded for views.
+//
+// This test attempts to perform write operations and get latency statistics using the $collStats
+// stage. The former operation must be routed to the primary in a replica set, whereas the latter
+// may be routed to a secondary.
+//
+// @tags: [assumes_read_preference_unchanged]
 
 (function() {
     "use strict";
diff --git a/jstests/core/views/views_validation.js b/jstests/core/views/views_validation.js
index 84c7f1d3510..b32750a082f 100644
--- a/jstests/core/views/views_validation.js
+++ b/jstests/core/views/views_validation.js
@@ -1,3 +1,4 @@
+// @tags: [requires_collmod_command]
 (function() {
     "use strict";
     let viewsDb = db.getSiblingDB("views_validation");
diff --git a/jstests/core/write_result.js b/jstests/core/write_result.js
index 86486089c68..453be4ca1c9 100644
--- a/jstests/core/write_result.js
+++ b/jstests/core/write_result.js
@@ -1,6 +1,8 @@
 //
 // Tests the behavior of single writes using write commands
 //
+// @tags: [assumes_write_concern_unchanged]
+//
 
 var coll = db.write_result;
 coll.drop();
diff --git a/jstests/hooks/validate_collections.js b/jstests/hooks/validate_collections.js
index aeb38a98bf5..e64bb39d305 100644
--- a/jstests/hooks/validate_collections.js
+++ b/jstests/hooks/validate_collections.js
@@ -71,6 +71,21 @@ function validateCollections(db, obj) {
         filter = {$or: [filter, {type: {$exists: false}}]};
     }
 
+    // Optionally skip collections.
+    if (Array.isArray(jsTest.options().skipValidationNamespaces) &&
+        jsTest.options().skipValidationNamespaces.length > 0) {
+        let skippedCollections = [];
+        for (let ns of jsTest.options().skipValidationNamespaces) {
+            // Strip off the database name from 'ns' to extract the collName.
+            const collName = ns.replace(new RegExp('^' + db.getName() + '\.'), '');
+            // Skip the collection 'collName' if the db name was removed from 'ns'.
+            if (collName !== ns) {
+                skippedCollections.push({name: {$ne: collName}});
+            }
+        }
+        filter = {$and: [filter, ...skippedCollections]};
+    }
+
     let collInfo = db.getCollectionInfos(filter);
     for (var collDocument of collInfo) {
         var coll = db.getCollection(collDocument["name"]);
diff --git a/jstests/libs/analyze_plan.js b/jstests/libs/analyze_plan.js
index 62511ae1fac..5e52ad6aba3 100644
--- a/jstests/libs/analyze_plan.js
+++ b/jstests/libs/analyze_plan.js
@@ -115,3 +115,74 @@ function getChunkSkips(root) {
 
     return 0;
 }
+
+/**
+ * Given the root stage of agg explain's JSON representation of a query plan ('root'), returns all
+ * subdocuments whose stage is 'stage'. This can either be an agg stage name like "$cursor" or
+ * "$sort", or a query stage name like "IXSCAN" or "SORT".
+ *
+ * Returns an empty array if the plan does not have the requested stage. Asserts that agg explain
+ * structure matches expected format.
+ */
+function getAggPlanStages(root, stage) {
+    let results = [];
+
+    function getDocumentSources(docSourceArray) {
+        let results = [];
+        for (let i = 0; i < docSourceArray.length; i++) {
+            let properties = Object.getOwnPropertyNames(docSourceArray[i]);
+            assert.eq(1, properties.length);
+            if (properties[0] === stage) {
+                results.push(docSourceArray[i]);
+            }
+        }
+        return results;
+    }
+
+    if (root.hasOwnProperty("stages")) {
+        assert(root.stages.constructor === Array);
+
+        results = results.concat(getDocumentSources(root.stages));
+
+        assert(root.stages[0].hasOwnProperty("$cursor"));
+        assert(root.stages[0].$cursor.hasOwnProperty("queryPlanner"));
+        assert(root.stages[0].$cursor.queryPlanner.hasOwnProperty("winningPlan"));
+        results =
+            results.concat(getPlanStages(root.stages[0].$cursor.queryPlanner.winningPlan, stage));
+    }
+
+    if (root.hasOwnProperty("shards")) {
+        for (let elem in root.shards) {
+            assert(root.shards[elem].stages.constructor === Array);
+
+            results = results.concat(getDocumentSources(root.shards[elem].stages));
+
+            assert(root.shards[elem].stages[0].hasOwnProperty("$cursor"));
+            assert(root.shards[elem].stages[0].$cursor.hasOwnProperty("queryPlanner"));
+            assert(root.shards[elem].stages[0].$cursor.queryPlanner.hasOwnProperty("winningPlan"));
+            results = results.concat(
+                getPlanStages(root.shards[elem].stages[0].$cursor.queryPlanner.winningPlan, stage));
+        }
+    }
+
+    return results;
+}
+
+/**
+ * Given the root stage of agg explain's JSON representation of a query plan ('root'), returns the
+ * subdocument with its stage as 'stage'. Returns null if the plan does not have such a stage.
+ * Asserts that no more than one stage is a match.
+ */
+function getAggPlanStage(root, stage) {
+    let planStageList = getAggPlanStages(root, stage);
+
+    if (planStageList.length === 0) {
+        return null;
+    } else {
+        assert.eq(1,
+                  planStageList.length,
+                  "getAggPlanStage expects to find 0 or 1 matching stages. planStageList: " +
+                      tojson(planStageList));
+        return planStageList[0];
+    }
+}
diff --git a/jstests/libs/ftdc.js b/jstests/libs/ftdc.js
new file mode 100644
index 00000000000..7d327d39852
--- /dev/null
+++ b/jstests/libs/ftdc.js
@@ -0,0 +1,102 @@
+/**
+ * Utility test functions for FTDC
+ */
+'use strict';
+
+/**
+ * Verify that getDiagnosticData is working correctly.
+ */
+function verifyGetDiagnosticData(adminDb) {
+    // We need to retry a few times if run this test immediately after mongod is started as FTDC may
+    // not have run yet.
+    var foundGoodDocument = false;
+
+    for (var i = 0; i < 60 && foundGoodDocument == false; ++i) {
+        var result = adminDb.runCommand("getDiagnosticData");
+        assert.commandWorked(result);
+
+        var data = result.data;
+
+        if (!data.hasOwnProperty("start")) {
+            // Wait a little longer for FTDC to start
+            jsTestLog("Running getDiagnosticData: " + tojson(result));
+
+            sleep(500);
+        } else {
+            // Check for a few common properties to ensure we got data
+            assert(data.hasOwnProperty("serverStatus"),
+                   "does not have 'serverStatus' in '" + tojson(data) + "'");
+            assert(data.hasOwnProperty("end"), "does not have 'end' in '" + tojson(data) + "'");
+            foundGoodDocument = true;
+
+            jsTestLog("Got good getDiagnosticData: " + tojson(result));
+        }
+    }
+
+    assert(foundGoodDocument,
+           "getDiagnosticData failed to return a non-empty command, is FTDC running?");
+}
+
+/**
+ * Validate all the common FTDC parameters are set correctly and can be manipulated.
+ */
+function verifyCommonFTDCParameters(adminDb, isEnabled) {
+    // Are we running against MongoS?
+    var isMongos = ("isdbgrid" == adminDb.runCommand("ismaster").msg);
+
+    // Check the defaults are correct
+    //
+    function getparam(field) {
+        var q = {getParameter: 1};
+        q[field] = 1;
+
+        var ret = adminDb.runCommand(q);
+        return ret[field];
+    }
+
+    // Verify the defaults are as we documented them
+    assert.eq(getparam("diagnosticDataCollectionEnabled"), isEnabled);
+    assert.eq(getparam("diagnosticDataCollectionPeriodMillis"), 1000);
+    assert.eq(getparam("diagnosticDataCollectionDirectorySizeMB"), 200);
+    assert.eq(getparam("diagnosticDataCollectionFileSizeMB"), 10);
+    assert.eq(getparam("diagnosticDataCollectionSamplesPerChunk"), 300);
+    assert.eq(getparam("diagnosticDataCollectionSamplesPerInterimUpdate"), 10);
+
+    function setparam(obj) {
+        var ret = adminDb.runCommand(Object.extend({setParameter: 1}, obj));
+        return ret;
+    }
+
+    if (!isMongos) {
+        // The MongoS specific behavior for diagnosticDataCollectionEnabled is tested in
+        // ftdc_setdirectory.js.
+        assert.commandWorked(setparam({"diagnosticDataCollectionEnabled": 1}));
+    }
+    assert.commandWorked(setparam({"diagnosticDataCollectionPeriodMillis": 100}));
+    assert.commandWorked(setparam({"diagnosticDataCollectionDirectorySizeMB": 10}));
+    assert.commandWorked(setparam({"diagnosticDataCollectionFileSizeMB": 1}));
+    assert.commandWorked(setparam({"diagnosticDataCollectionSamplesPerChunk": 2}));
+    assert.commandWorked(setparam({"diagnosticDataCollectionSamplesPerInterimUpdate": 2}));
+
+    // Negative tests - set values below minimums
+    assert.commandFailed(setparam({"diagnosticDataCollectionPeriodMillis": 1}));
+    assert.commandFailed(setparam({"diagnosticDataCollectionDirectorySizeMB": 1}));
+    assert.commandFailed(setparam({"diagnosticDataCollectionSamplesPerChunk": 1}));
+    assert.commandFailed(setparam({"diagnosticDataCollectionSamplesPerInterimUpdate": 1}));
+
+    // Negative test - set file size bigger then directory size
+    assert.commandWorked(setparam({"diagnosticDataCollectionDirectorySizeMB": 10}));
+    assert.commandFailed(setparam({"diagnosticDataCollectionFileSizeMB": 100}));
+
+    // Negative test - set directory size less then file size
+    assert.commandWorked(setparam({"diagnosticDataCollectionDirectorySizeMB": 100}));
+    assert.commandWorked(setparam({"diagnosticDataCollectionFileSizeMB": 50}));
+    assert.commandFailed(setparam({"diagnosticDataCollectionDirectorySizeMB": 10}));
+
+    // Reset
+    assert.commandWorked(setparam({"diagnosticDataCollectionFileSizeMB": 10}));
+    assert.commandWorked(setparam({"diagnosticDataCollectionDirectorySizeMB": 200}));
+    assert.commandWorked(setparam({"diagnosticDataCollectionPeriodMillis": 1000}));
+    assert.commandWorked(setparam({"diagnosticDataCollectionSamplesPerChunk": 300}));
+    assert.commandWorked(setparam({"diagnosticDataCollectionSamplesPerInterimUpdate": 10}));
+}
\ No newline at end of file
diff --git a/jstests/libs/override_methods/override_helpers.js b/jstests/libs/override_methods/override_helpers.js
new file mode 100644
index 00000000000..12bd4b54739
--- /dev/null
+++ b/jstests/libs/override_methods/override_helpers.js
@@ -0,0 +1,117 @@
+/**
+ * The OverrideHelpers object defines convenience methods for overriding commands and functions in
+ * the mongo shell.
+ */
+var OverrideHelpers = (function() {
+    "use strict";
+
+    function isAggregationWithOutStage(commandName, commandObj) {
+        if (commandName !== "aggregate" || typeof commandObj !== "object" || commandObj === null) {
+            return false;
+        }
+
+        if (!Array.isArray(commandObj.pipeline) || commandObj.pipeline.length === 0) {
+            return false;
+        }
+
+        const lastStage = commandObj.pipeline[commandObj.pipeline.length - 1];
+        if (typeof lastStage !== "object" || lastStage === null) {
+            return false;
+        }
+
+        return Object.keys(lastStage)[0] === "$out";
+    }
+
+    function isMapReduceWithInlineOutput(commandName, commandObj) {
+        if ((commandName !== "mapReduce" && commandName !== "mapreduce") ||
+            typeof commandObj !== "object" || commandObj === null) {
+            return false;
+        }
+
+        if (typeof commandObj.out !== "object") {
+            return false;
+        }
+
+        return commandObj.out.hasOwnProperty("inline");
+    }
+
+    function prependOverrideInParallelShell(overrideFile) {
+        const startParallelShellOriginal = startParallelShell;
+
+        startParallelShell = function(jsCode, port, noConnect) {
+            let newCode;
+            if (typeof jsCode === "function") {
+                // Load the override file and immediately invoke the supplied function.
+                newCode = `load("${overrideFile}"); (${jsCode})();`;
+            } else {
+                newCode = `load("${overrideFile}"); ${jsCode};`;
+            }
+
+            return startParallelShellOriginal(newCode, port, noConnect);
+        };
+    }
+
+    function overrideRunCommand(overrideFunc) {
+        const DBQueryOriginal = DBQuery;
+        const mongoRunCommandOriginal = Mongo.prototype.runCommand;
+        const mongoRunCommandWithMetadataOriginal = Mongo.prototype.runCommandWithMetadata;
+
+        DBQuery = function(
+            mongo, db, collection, ns, query, fields, limit, skip, batchSize, options) {
+            // If the query isn't being run against the "$cmd" or "$cmd.sys" namespaces, then it
+            // represents an OP_QUERY find on that collection. We skip calling overrideFunc() in
+            // this case because the operation doesn't represent a command.
+            if (!(collection instanceof DBCollection &&
+                  (collection.getName() === "$cmd" || collection.getName().startsWith("$cmd.")))) {
+                return DBQueryOriginal.apply(this, arguments);
+            }
+
+            // Due to the function signatures of Mongo.prototype.runCommand() and
+            // Mongo.prototype.runCommandWithMetadata(), the overrideFunc() function expects that
+            // the Mongo connection object is passed as the first argument and also represents the
+            // 'this' parameter. As a workaround, we bind the appropriate 'this' value to the
+            // DBQueryOriginal constructor ahead of time.
+            const commandName = Object.keys(query)[0];
+            return overrideFunc(
+                mongo,
+                db.getName(),
+                commandName,
+                query,
+                DBQueryOriginal.bind(this),
+                (query) =>
+                    [mongo, db, collection, ns, query, fields, limit, skip, batchSize, options]);
+        };
+
+        // Copy any properties (e.g. DBQuery.Option) that are set on DBQueryOriginal.
+        Object.keys(DBQueryOriginal).forEach(function(key) {
+            DBQuery[key] = DBQueryOriginal[key];
+        });
+
+        Mongo.prototype.runCommand = function(dbName, commandObj, options) {
+            const commandName = Object.keys(commandObj)[0];
+            return overrideFunc(this,
+                                dbName,
+                                commandName,
+                                commandObj,
+                                mongoRunCommandOriginal,
+                                (commandObj) => [dbName, commandObj, options]);
+        };
+
+        Mongo.prototype.runCommandWithMetadata = function(dbName, metadata, commandArgs) {
+            const commandName = Object.keys(commandArgs)[0];
+            return overrideFunc(this,
+                                dbName,
+                                commandName,
+                                commandArgs,
+                                mongoRunCommandWithMetadataOriginal,
+                                (commandArgs) => [dbName, metadata, commandArgs]);
+        };
+    }
+
+    return {
+        isAggregationWithOutStage: isAggregationWithOutStage,
+        isMapReduceWithInlineOutput: isMapReduceWithInlineOutput,
+        prependOverrideInParallelShell: prependOverrideInParallelShell,
+        overrideRunCommand: overrideRunCommand,
+    };
+})();
diff --git a/jstests/libs/override_methods/set_majority_read_and_write_concerns.js b/jstests/libs/override_methods/set_majority_read_and_write_concerns.js
deleted file mode 100644
index d3bb4449ed4..00000000000
--- a/jstests/libs/override_methods/set_majority_read_and_write_concerns.js
+++ /dev/null
@@ -1,178 +0,0 @@
-/**
- * Use prototype overrides to set a read concern of "majority" and a write concern of "majority"
- * while running core tests.
- */
-(function() {
-    "use strict";
-    var defaultWriteConcern = {
-        w: "majority",
-        // Use a "signature" value that won't typically match a value assigned in normal use.
-        wtimeout: 60321
-    };
-    var defaultReadConcern = {level: "majority"};
-
-    var originalDBQuery = DBQuery;
-
-    DBQuery = function(mongo, db, collection, ns, query, fields, limit, skip, batchSize, options) {
-        if (ns.endsWith("$cmd")) {
-            if (query.hasOwnProperty("writeConcern") &&
-                bsonWoCompare(query.writeConcern, defaultWriteConcern) !== 0) {
-                jsTestLog("Warning: DBQuery overriding existing writeConcern of: " +
-                          tojson(query.writeConcern));
-                query.writeConcern = defaultWriteConcern;
-            }
-        }
-
-        return originalDBQuery.apply(this, arguments);
-    };
-
-    DBQuery.Option = originalDBQuery.Option;
-
-    var originalStartParallelShell = startParallelShell;
-    startParallelShell = function(jsCode, port, noConnect) {
-        var newCode;
-        var overridesFile = "jstests/libs/override_methods/set_majority_read_and_write_concerns.js";
-
-        if (typeof(jsCode) === "function") {
-            // Load the override file and immediately invoke the supplied function.
-            newCode = `load("${overridesFile}"); (${jsCode})();`;
-        } else {
-            newCode = `load("${overridesFile}"); ${jsCode};`;
-        }
-
-        return originalStartParallelShell(newCode, port, noConnect);
-    };
-
-    DB.prototype._runCommandImpl = function(dbName, obj, options) {
-        var cmdName = "";
-        for (var fieldName in obj) {
-            cmdName = fieldName;
-            break;
-        }
-
-        // These commands directly support a writeConcern argument.
-        var commandsToForceWriteConcern = [
-            "_mergeAuthzCollections",
-            "appendOplogNote",
-            "applyOps",
-            "authSchemaUpgrade",
-            "captrunc",
-            "cleanupOrphaned",
-            "clone",
-            "cloneCollection",
-            "cloneCollectionAsCapped",
-            // "collMod", SERVER-25196 - not supported
-            "convertToCapped",
-            "copydb",
-            "create",
-            "createIndexes",
-            "createRole",
-            "createUser",
-            "delete",
-            "drop",
-            "dropDatabase",
-            "dropAllRolesFromDatabase",
-            "dropAllUsersFromDatabase",
-            "dropDatabase",
-            "dropIndexes",
-            "dropRole",
-            "dropUser",
-            "emptycapped",
-            "findAndModify",
-            "findandmodify",
-            "godinsert",
-            "grantPrivilegesToRole",
-            "grantRolesToRole",
-            "grantRolesToUser",
-            "insert",
-            "mapReduceFinish",
-            "mergeAuthzCollections",
-            "moveChunk",
-            "movePrimary",
-            "reIndex",
-            "remove",
-            "renameCollection",
-            "resvChunkStart",
-            "revokePriviligesFromRole",
-            "revokeRolesFromRole",
-            "revokeRolesFromUser",
-            "update",
-            "updateRole",
-            "updateUser",
-        ];
-
-        // These are reading commands that support majority readConcern.
-        var commandsToForceReadConcern = [
-            "count",
-            "distinct",
-            "find",
-            "geoNear",
-            "geoSearch",
-            "group",
-        ];
-
-        var forceWriteConcern = Array.contains(commandsToForceWriteConcern, cmdName);
-        var forceReadConcern = Array.contains(commandsToForceReadConcern, cmdName);
-
-        if (cmdName === "aggregate") {
-            // Aggregate can be either a read or a write depending on whether it has a $out stage.
-            // $out is required to be the last stage of the pipeline.
-            var stages = obj.pipeline;
-            const lastStage = stages && Array.isArray(stages) && (stages.length !== 0)
-                ? stages[stages.length - 1]
-                : undefined;
-            const hasOut =
-                lastStage && (typeof lastStage === 'object') && lastStage.hasOwnProperty('$out');
-            if (hasOut) {
-                forceWriteConcern = true;
-            } else {
-                forceReadConcern = true;
-            }
-        }
-
-        else if (cmdName === "mapReduce") {
-            var stages = obj.pipeline;
-            const lastStage = stages && Array.isArray(stages) && (stages.length !== 0)
-                ? stages[stages.length - 1]
-                : undefined;
-            const hasOut =
-                lastStage && (typeof lastStage === 'object') && lastStage.hasOwnProperty('$out');
-            if (hasOut) {
-                forceWriteConcern = true;
-            }
-        }
-
-        if (forceWriteConcern) {
-            if (obj.hasOwnProperty("writeConcern")) {
-                if (bsonWoCompare(obj.writeConcern, defaultWriteConcern) !== 0) {
-                    jsTestLog("Warning: _runCommandImpl overriding existing writeConcern of: " +
-                              tojson(obj.writeConcern));
-                    obj.writeConcern = defaultWriteConcern;
-                }
-            } else {
-                obj.writeConcern = defaultWriteConcern;
-            }
-
-        } else if (forceReadConcern) {
-            if (obj.hasOwnProperty("readConcern")) {
-                if (bsonWoCompare(obj.readConcern, defaultReadConcern) !== 0) {
-                    jsTestLog("Warning: _runCommandImpl overriding existing readConcern of: " +
-                              tojson(obj.readConcern));
-                    obj.readConcern = defaultReadConcern;
-                }
-            } else {
-                obj.readConcern = defaultReadConcern;
-            }
-        }
-
-        var res = this.getMongo().runCommand(dbName, obj, options);
-
-        return res;
-    };
-
-    // Use a majority write concern if the operation does not specify one.
-    DBCollection.prototype.getWriteConcern = function() {
-        return new WriteConcern(defaultWriteConcern);
-    };
-
-})();
diff --git a/jstests/libs/override_methods/set_read_and_write_concerns.js b/jstests/libs/override_methods/set_read_and_write_concerns.js
new file mode 100644
index 00000000000..f371644c8f7
--- /dev/null
+++ b/jstests/libs/override_methods/set_read_and_write_concerns.js
@@ -0,0 +1,216 @@
+/**
+ * Use prototype overrides to set read concern and write concern while running tests.
+ */
+(function() {
+    "use strict";
+
+    load("jstests/libs/override_methods/override_helpers.js");
+
+    if (typeof TestData === "undefined" || !TestData.hasOwnProperty("defaultReadConcernLevel")) {
+        throw new Error(
+            "The readConcern level to use must be set as the 'defaultReadConcernLevel'" +
+            " property on the global TestData object");
+    }
+
+    const kDefaultReadConcern = {level: TestData.defaultReadConcernLevel};
+    const kDefaultWriteConcern =
+        (TestData.hasOwnProperty("defaultWriteConcern")) ? TestData.defaultWriteConcern : {
+            w: "majority",
+            // Use a "signature" value that won't typically match a value assigned in normal use.
+            // This way the wtimeout set by this override is distinguishable in the server logs.
+            wtimeout: 5 * 60 * 1000 + 321,  // 300321ms
+        };
+
+    const kCommandsSupportingReadConcern = new Set([
+        "aggregate",
+        "count",
+        "distinct",
+        "find",
+        "geoNear",
+        "geoSearch",
+        "group",
+        "parallelCollectionScan",
+    ]);
+
+    const kCommandsSupportingWriteConcern = new Set([
+        "_configsvrAddShard",
+        "_configsvrAddShardToZone",
+        "_configsvrCommitChunkMerge",
+        "_configsvrCommitChunkMigration",
+        "_configsvrCommitChunkSplit",
+        "_configsvrCreateDatabase",
+        "_configsvrEnableSharding",
+        "_configsvrMoveChunk",
+        "_configsvrMovePrimary",
+        "_configsvrRemoveShard",
+        "_configsvrRemoveShardFromZone",
+        "_configsvrShardCollection",
+        "_configsvrUpdateZoneKeyRange",
+        "_mergeAuthzCollections",
+        "_recvChunkStart",
+        "appendOplogNote",
+        "applyOps",
+        "authSchemaUpgrade",
+        "aggregate",
+        "captrunc",
+        "cleanupOrphaned",
+        "clone",
+        "cloneCollection",
+        "cloneCollectionAsCapped",
+        // "collMod", SERVER-25196 - not supported
+        "convertToCapped",
+        "copydb",
+        "create",
+        "createIndexes",
+        "createRole",
+        "createUser",
+        "delete",
+        "deleteIndexes",
+        "drop",
+        "dropAllRolesFromDatabase",
+        "dropAllUsersFromDatabase",
+        "dropDatabase",
+        "dropIndexes",
+        "dropRole",
+        "dropUser",
+        "emptycapped",
+        "findAndModify",
+        "findandmodify",
+        "godinsert",
+        "grantPrivilegesToRole",
+        "grantRolesToRole",
+        "grantRolesToUser",
+        "insert",
+        "mapReduce",
+        "mapreduce",
+        "mapreduce.shardedfinish",
+        "moveChunk",
+        "renameCollection",
+        "revokePrivilegesFromRole",
+        "revokeRolesFromRole",
+        "revokeRolesFromUser",
+        "setFeatureCompatibilityVersion",
+        "update",
+        "updateRole",
+        "updateUser",
+    ]);
+
+    function runCommandWithReadAndWriteConcerns(
+        conn, dbName, commandName, commandObj, func, makeFuncArgs) {
+        if (typeof commandObj !== "object" || commandObj === null) {
+            return func.apply(conn, makeFuncArgs(commandObj));
+        }
+
+        // If the command is in a wrapped form, then we look for the actual command object inside
+        // the query/$query object.
+        let commandObjUnwrapped = commandObj;
+        if (commandName === "query" || commandName === "$query") {
+            commandObjUnwrapped = commandObj[commandName];
+            commandName = Object.keys(commandObjUnwrapped)[0];
+        }
+
+        if (commandName === "collMod" || commandName === "eval" || commandName === "$eval") {
+            throw new Error("Cowardly refusing to run test with overridden write concern when it" +
+                            " uses a command that can only perform w=1 writes: " +
+                            tojson(commandObj));
+        }
+
+        let shouldForceReadConcern = kCommandsSupportingReadConcern.has(commandName);
+        let shouldForceWriteConcern = kCommandsSupportingWriteConcern.has(commandName);
+
+        if (commandName === "aggregate") {
+            if (OverrideHelpers.isAggregationWithOutStage(commandName, commandObjUnwrapped)) {
+                // The $out stage can only be used with readConcern={level: "local"}.
+                shouldForceReadConcern = false;
+            } else {
+                // A writeConcern can only be used with a $out stage.
+                shouldForceWriteConcern = false;
+            }
+
+            if (commandObjUnwrapped.explain) {
+                // Attempting to specify a readConcern while explaining an aggregation would always
+                // return an error prior to SERVER-30582 and it otherwise only compatible with
+                // readConcern={level: "local"}.
+                shouldForceReadConcern = false;
+            }
+        } else if (OverrideHelpers.isMapReduceWithInlineOutput(commandName, commandObjUnwrapped)) {
+            // A writeConcern can only be used with non-inline output.
+            shouldForceWriteConcern = false;
+        } else if (commandObj[commandName] === "system.profile") {
+            // Writes to the "system.profile" collection aren't guaranteed to be visible in the same
+            // majority-committed snapshot as the command they originated from. We don't override
+            // the readConcern for operations on the "system.profile" collection so that tests which
+            // assert on its contents continue to succeed.
+            shouldForceReadConcern = false;
+        }
+
+        const inWrappedForm = commandObj !== commandObjUnwrapped;
+
+        if (shouldForceReadConcern) {
+            // We create a copy of 'commandObj' to avoid mutating the parameter the caller
+            // specified.
+            commandObj = Object.assign({}, commandObj);
+            if (inWrappedForm) {
+                commandObjUnwrapped = Object.assign({}, commandObjUnwrapped);
+                commandObj[Object.keys(commandObj)[0]] = commandObjUnwrapped;
+            } else {
+                commandObjUnwrapped = commandObj;
+            }
+
+            if (commandObjUnwrapped.hasOwnProperty("readConcern")) {
+                let readConcern = commandObjUnwrapped.readConcern;
+
+                if (typeof readConcern !== "object" || readConcern === null ||
+                    (readConcern.hasOwnProperty("level") &&
+                     bsonWoCompare({_: readConcern.level}, {_: kDefaultReadConcern.level}) !== 0)) {
+                    throw new Error("Cowardly refusing to override read concern of command: " +
+                                    tojson(commandObj));
+                }
+
+                // We create a copy of the readConcern object to avoid mutating the parameter the
+                // caller specified.
+                readConcern = Object.assign({}, readConcern, kDefaultReadConcern);
+                commandObjUnwrapped.readConcern = readConcern;
+            } else {
+                commandObjUnwrapped.readConcern = kDefaultReadConcern;
+            }
+        }
+
+        if (shouldForceWriteConcern) {
+            // We create a copy of 'commandObj' to avoid mutating the parameter the caller
+            // specified.
+            commandObj = Object.assign({}, commandObj);
+            if (inWrappedForm) {
+                commandObjUnwrapped = Object.assign({}, commandObjUnwrapped);
+                commandObj[Object.keys(commandObj)[0]] = commandObjUnwrapped;
+            } else {
+                commandObjUnwrapped = commandObj;
+            }
+
+            if (commandObjUnwrapped.hasOwnProperty("writeConcern")) {
+                let writeConcern = commandObjUnwrapped.writeConcern;
+
+                if (typeof writeConcern !== "object" || writeConcern === null ||
+                    (writeConcern.hasOwnProperty("w") &&
+                     bsonWoCompare({_: writeConcern.w}, {_: kDefaultWriteConcern.w}) !== 0)) {
+                    throw new Error("Cowardly refusing to override write concern of command: " +
+                                    tojson(commandObj));
+                }
+
+                // We create a copy of the writeConcern object to avoid mutating the parameter the
+                // caller specified.
+                writeConcern = Object.assign({}, writeConcern, kDefaultWriteConcern);
+                commandObjUnwrapped.writeConcern = writeConcern;
+            } else {
+                commandObjUnwrapped.writeConcern = kDefaultWriteConcern;
+            }
+        }
+
+        return func.apply(conn, makeFuncArgs(commandObj));
+    }
+
+    OverrideHelpers.prependOverrideInParallelShell(
+        "jstests/libs/override_methods/set_read_and_write_concerns.js");
+
+    OverrideHelpers.overrideRunCommand(runCommandWithReadAndWriteConcerns);
+})();
diff --git a/jstests/libs/override_methods/set_read_preference_secondary.js b/jstests/libs/override_methods/set_read_preference_secondary.js
new file mode 100644
index 00000000000..d1d26433c5c
--- /dev/null
+++ b/jstests/libs/override_methods/set_read_preference_secondary.js
@@ -0,0 +1,162 @@
+/**
+ * Use prototype overrides to set read preference to "secondary" when running tests.
+ */
+(function() {
+    "use strict";
+
+    load("jstests/libs/override_methods/override_helpers.js");
+
+    const kReadPreferenceSecondary = {mode: "secondary"};
+    const kCommandsSupportingReadPreference = new Set([
+        "aggregate",
+        "collStats",
+        "count",
+        "dbStats",
+        "distinct",
+        "find",
+        "geoNear",
+        "geoSearch",
+        "group",
+        "mapReduce",
+        "mapreduce",
+        "parallelCollectionScan",
+    ]);
+
+    // This list of cursor-generating commands is incomplete. For example, "listCollections",
+    // "listIndexes", "parallelCollectionScan", and "repairCursor" are all missing from this list.
+    // If we ever add tests that attempt to run getMore or killCursors on cursors generated from
+    // those commands, then we should update the contents of this list and also handle any
+    // differences in the server's response format.
+    const kCursorGeneratingCommands = new Set(["aggregate", "find"]);
+
+    const CursorTracker = (function() {
+        const kNoCursor = new NumberLong(0);
+
+        const connectionsByCursorId = {};
+
+        return {
+            getConnectionUsedForCursor: function getConnectionUsedForCursor(cursorId) {
+                return (cursorId instanceof NumberLong) ? connectionsByCursorId[cursorId]
+                                                        : undefined;
+            },
+
+            setConnectionUsedForCursor: function setConnectionUsedForCursor(cursorId, cursorConn) {
+                if (cursorId instanceof NumberLong &&
+                    !bsonBinaryEqual({_: cursorId}, {_: kNoCursor})) {
+                    connectionsByCursorId[cursorId] = cursorConn;
+                }
+            },
+        };
+    })();
+
+    function runCommandWithReadPreferenceSecondary(
+        conn, dbName, commandName, commandObj, func, makeFuncArgs) {
+        if (typeof commandObj !== "object" || commandObj === null) {
+            return func.apply(conn, makeFuncArgs(commandObj));
+        }
+
+        // If the command is in a wrapped form, then we look for the actual command object inside
+        // the query/$query object.
+        let commandObjUnwrapped = commandObj;
+        if (commandName === "query" || commandName === "$query") {
+            commandObjUnwrapped = commandObj[commandName];
+            commandName = Object.keys(commandObjUnwrapped)[0];
+        }
+
+        if (commandObj[commandName] === "system.profile") {
+            throw new Error("Cowardly refusing to run test with overridden read preference" +
+                            " when it reads from a non-replicated collection: " +
+                            tojson(commandObj));
+        }
+
+        if (conn.isReplicaSetConnection()) {
+            // When a "getMore" or "killCursors" command is issued on a replica set connection, we
+            // attempt to automatically route the command to the server the cursor(s) were
+            // originally established on. This makes it possible to use the
+            // set_read_preference_secondary.js override without needing to update calls of
+            // DB#runCommand() to explicitly track the connection that was used. If the connection
+            // is actually a direct connection to a mongod or mongos process, or if the cursor id
+            // cannot be found in the CursorTracker, then we'll fall back to using DBClientRS's
+            // server selection and send the operation to the current primary. It is possible that
+            // the test is trying to exercise the behavior around when an unknown cursor id is sent
+            // to the server.
+            if (commandName === "getMore") {
+                const cursorId = commandObjUnwrapped[commandName];
+                const cursorConn = CursorTracker.getConnectionUsedForCursor(cursorId);
+                if (cursorConn !== undefined) {
+                    return func.apply(cursorConn, makeFuncArgs(commandObj));
+                }
+            } else if (commandName === "killCursors") {
+                const cursorIds = commandObjUnwrapped.cursors;
+                if (Array.isArray(cursorIds)) {
+                    let cursorConn;
+
+                    for (let cursorId of cursorIds) {
+                        const otherCursorConn = CursorTracker.getConnectionUsedForCursor(cursorId);
+                        if (cursorConn === undefined) {
+                            cursorConn = otherCursorConn;
+                        } else if (otherCursorConn !== undefined) {
+                            // We set 'cursorConn' back to undefined and break out of the loop so
+                            // that we don't attempt to automatically route the "killCursors"
+                            // command when there are cursors from different servers.
+                            cursorConn = undefined;
+                            break;
+                        }
+                    }
+
+                    if (cursorConn !== undefined) {
+                        return func.apply(cursorConn, makeFuncArgs(commandObj));
+                    }
+                }
+            }
+        }
+
+        let shouldForceReadPreference = kCommandsSupportingReadPreference.has(commandName);
+        if (OverrideHelpers.isAggregationWithOutStage(commandName, commandObjUnwrapped)) {
+            // An aggregation with a $out stage must be sent to the primary.
+            shouldForceReadPreference = false;
+        } else if ((commandName === "mapReduce" || commandName === "mapreduce") &&
+                   !OverrideHelpers.isMapReduceWithInlineOutput(commandName, commandObjUnwrapped)) {
+            // A map-reduce operation with non-inline output must be sent to the primary.
+            shouldForceReadPreference = false;
+        }
+
+        if (shouldForceReadPreference) {
+            if (commandObj === commandObjUnwrapped) {
+                // We wrap the command object using a "query" field rather than a "$query" field to
+                // match the implementation of DB.prototype._attachReadPreferenceToCommand().
+                commandObj = {query: commandObj};
+            } else {
+                // We create a copy of 'commandObj' to avoid mutating the parameter the caller
+                // specified.
+                commandObj = Object.assign({}, commandObj);
+            }
+
+            if (commandObj.hasOwnProperty("$readPreference") &&
+                !bsonBinaryEqual({_: commandObj.$readPreference}, {_: kReadPreferenceSecondary})) {
+                throw new Error("Cowardly refusing to override read preference of command: " +
+                                tojson(commandObj));
+            }
+
+            commandObj.$readPreference = kReadPreferenceSecondary;
+        }
+
+        const serverResponse = func.apply(conn, makeFuncArgs(commandObj));
+
+        if (conn.isReplicaSetConnection() && kCursorGeneratingCommands.has(commandName) &&
+            serverResponse.ok === 1 && serverResponse.hasOwnProperty("cursor")) {
+            // We associate the cursor id returned by the server with the connection that was used
+            // to establish it so that we can attempt to automatically route subsequent "getMore"
+            // and "killCursors" commands.
+            CursorTracker.setConnectionUsedForCursor(serverResponse.cursor.id,
+                                                     serverResponse._mongo);
+        }
+
+        return serverResponse;
+    }
+
+    OverrideHelpers.prependOverrideInParallelShell(
+        "jstests/libs/override_methods/set_read_preference_secondary.js");
+
+    OverrideHelpers.overrideRunCommand(runCommandWithReadPreferenceSecondary);
+})();
diff --git a/jstests/multiVersion/initial_sync_last_stable_from_latest.js b/jstests/multiVersion/initial_sync_last_stable_from_latest.js
new file mode 100644
index 00000000000..3f036c17fe9
--- /dev/null
+++ b/jstests/multiVersion/initial_sync_last_stable_from_latest.js
@@ -0,0 +1,15 @@
+/**
+ * Multiversion initial sync test. Tests that initial sync succeeds when a 'last-stable' version
+ * secondary syncs from a 'latest' version replica set.
+ */
+
+'use strict';
+
+load("./jstests/multiVersion/libs/initial_sync.js");
+
+var testName = "multiversion_initial_sync_last_stable_from_latest";
+let replSetVersion = "latest";
+let newSecondaryVersion = "last-stable";
+let fcv = "3.2";
+
+multversionInitialSyncTest(testName, replSetVersion, newSecondaryVersion, {}, fcv);
diff --git a/jstests/multiVersion/initial_sync_latest_from_last_stable.js b/jstests/multiVersion/initial_sync_latest_from_last_stable.js
new file mode 100644
index 00000000000..fd92b64e87a
--- /dev/null
+++ b/jstests/multiVersion/initial_sync_latest_from_last_stable.js
@@ -0,0 +1,14 @@
+/**
+ * Multiversion initial sync test. Tests that initial sync succeeds when a 'latest' version
+ * secondary syncs from a 'last-stable' version replica set.
+ */
+
+'use strict';
+
+load("./jstests/multiVersion/libs/initial_sync.js");
+
+var testName = "multiversion_initial_sync_latest_from_last_stable";
+let replSetVersion = "last-stable";
+let newSecondaryVersion = "latest";
+
+multversionInitialSyncTest(testName, replSetVersion, newSecondaryVersion, {});
diff --git a/jstests/multiVersion/initialsync.js b/jstests/multiVersion/initialsync.js
deleted file mode 100644
index e9a424fd05c..00000000000
--- a/jstests/multiVersion/initialsync.js
+++ /dev/null
@@ -1,58 +0,0 @@
-// Multiversion initial sync test.
-load("./jstests/multiVersion/libs/multi_rs.js");
-load("./jstests/replsets/rslib.js");
-
-var oldVersion = "last-stable";
-var newVersion = "latest";
-
-var name = "multiversioninitsync";
-
-var multitest = function(replSetVersion, newNodeVersion) {
-    var nodes = {n1: {binVersion: replSetVersion}, n2: {binVersion: replSetVersion}};
-
-    print("Start up a two-node " + replSetVersion + " replica set.");
-    var rst = new ReplSetTest({name: name, nodes: nodes});
-    rst.startSet();
-    var config = rst.getReplSetConfig();
-    // Set protocol version to 0 for 3.2 replset.
-    if (replSetVersion == newVersion) {
-        config.protocolVersion = 0;
-    }
-    rst.initiate(config);
-
-    // Wait for a primary node.
-    var primary = rst.getPrimary();
-
-    // Insert some data and wait for replication.
-    for (var i = 0; i < 25; i++) {
-        primary.getDB("foo").foo.insert({_id: i});
-    }
-    rst.awaitReplication();
-
-    print("Bring up a new node with version " + newNodeVersion + " and add to set.");
-    rst.add({binVersion: newNodeVersion});
-    rst.reInitiate();
-
-    // Wait for a primary node.
-    var primary = rst.getPrimary();
-    var secondaries = rst.getSecondaries();
-
-    print("Wait for new node to be synced.");
-    rst.awaitReplication();
-
-    rst.stopSet();
-};
-
-// *****************************************
-// Test A:
-// "Latest" version secondary is synced from
-// an old ReplSet.
-// *****************************************
-multitest(oldVersion, newVersion);
-
-// *****************************************
-// Test B:
-// Old Secondary is synced from a "latest"
-// version ReplSet.
-// *****************************************
-multitest(newVersion, oldVersion);
diff --git a/jstests/multiVersion/libs/initial_sync.js b/jstests/multiVersion/libs/initial_sync.js
new file mode 100644
index 00000000000..866f825c8b9
--- /dev/null
+++ b/jstests/multiVersion/libs/initial_sync.js
@@ -0,0 +1,50 @@
+
+'use strict';
+
+load("./jstests/multiVersion/libs/multi_rs.js");
+load("./jstests/replsets/rslib.js");
+
+/**
+ * Test that starts up a replica set with 2 nodes of version 'replSetVersion', inserts some data,
+ * then adds a new node to the replica set with version 'newNodeVersion' and waits for initial sync
+ * to complete. If the 'fcv' argument is given, sets the feature compatibility version of the
+ * replica set to 'fcv' before adding the third node.
+ */
+var multversionInitialSyncTest = function(
+    name, replSetVersion, newNodeVersion, configSettings, fcv) {
+
+    var nodes = {n1: {binVersion: replSetVersion}, n2: {binVersion: replSetVersion}};
+
+    jsTestLog("Starting up a two-node '" + replSetVersion + "' version replica set.");
+    var rst = new ReplSetTest({name: name, nodes: nodes});
+    rst.startSet();
+
+    var conf = rst.getReplSetConfig();
+    conf.settings = configSettings;
+    rst.initiate(conf);
+
+    // Wait for a primary node.
+    var primary = rst.getPrimary();
+
+    // Set 'featureCompatibilityVersion' if given.
+    if (fcv) {
+        jsTestLog("Setting FCV to '" + fcv + "' on the primary.");
+        assert.commandWorked(primary.adminCommand({setFeatureCompatibilityVersion: fcv}));
+    }
+
+    // Insert some data and wait for replication.
+    for (var i = 0; i < 25; i++) {
+        primary.getDB("foo").foo.insert({_id: i});
+    }
+    rst.awaitReplication();
+
+    jsTestLog("Bringing up a new node with version '" + newNodeVersion + "' and adding to set.");
+    rst.add({binVersion: newNodeVersion});
+    rst.reInitiate();
+
+    jsTestLog("Waiting for new node to be synced.");
+    rst.awaitReplication();
+    rst.awaitSecondaryNodes();
+
+    rst.stopSet();
+};
\ No newline at end of file
diff --git a/jstests/multiVersion/libs/multi_rs.js b/jstests/multiVersion/libs/multi_rs.js
index da976c7f4dc..87d5995ef48 100644
--- a/jstests/multiVersion/libs/multi_rs.js
+++ b/jstests/multiVersion/libs/multi_rs.js
@@ -25,6 +25,7 @@ ReplSetTest.prototype.upgradeSet = function(options, user, pwd) {
         var node = nodesToUpgrade[i];
         if (node == primary) {
             node = this.stepdown(node);
+            this.waitForState(node, ReplSetTest.State.SECONDARY);
             primary = this.getPrimary();
         }
 
diff --git a/jstests/multiVersion/set_feature_compatibility_version.js b/jstests/multiVersion/set_feature_compatibility_version.js
index d5c73f26efb..73a5785aec9 100644
--- a/jstests/multiVersion/set_feature_compatibility_version.js
+++ b/jstests/multiVersion/set_feature_compatibility_version.js
@@ -54,6 +54,23 @@
     // featureCompatibilityVersion cannot be set via setParameter.
     assert.commandFailed(adminDB.runCommand({setParameter: 1, featureCompatibilityVersion: "3.2"}));
 
+    // setFeatureCompatibilityVersion fails to downgrade to FCV=3.2 if the write fails.
+    assert.commandWorked(adminDB.runCommand({
+        configureFailPoint: "failCollectionUpdates",
+        data: {collectionNS: "admin.system.version"},
+        mode: "alwaysOn"
+    }));
+    assert.commandFailed(adminDB.runCommand({setFeatureCompatibilityVersion: "3.2"}));
+    res = adminDB.runCommand({getParameter: 1, featureCompatibilityVersion: 1});
+    assert.commandWorked(res);
+    assert.eq(res.featureCompatibilityVersion, "3.4");
+    assert.eq(adminDB.system.version.findOne({_id: "featureCompatibilityVersion"}).version, "3.4");
+    assert.commandWorked(adminDB.runCommand({
+        configureFailPoint: "failCollectionUpdates",
+        data: {collectionNS: "admin.system.version"},
+        mode: "off"
+    }));
+
     // featureCompatibilityVersion can be set to 3.2.
     assert.commandWorked(adminDB.runCommand({setFeatureCompatibilityVersion: "3.2"}));
     res = adminDB.runCommand({getParameter: 1, featureCompatibilityVersion: 1});
@@ -70,6 +87,23 @@
               "Expected index with name 'incompatible_with_version_32' to have been removed: " +
                   tojson(allIndexes));
 
+    // setFeatureCompatibilityVersion fails to upgrade to FCV=3.4 if the write fails.
+    assert.commandWorked(adminDB.runCommand({
+        configureFailPoint: "failCollectionUpdates",
+        data: {collectionNS: "admin.system.version"},
+        mode: "alwaysOn"
+    }));
+    assert.commandFailed(adminDB.runCommand({setFeatureCompatibilityVersion: "3.4"}));
+    res = adminDB.runCommand({getParameter: 1, featureCompatibilityVersion: 1});
+    assert.commandWorked(res);
+    assert.eq(res.featureCompatibilityVersion, "3.2");
+    assert.eq(adminDB.system.version.findOne({_id: "featureCompatibilityVersion"}).version, "3.2");
+    assert.commandWorked(adminDB.runCommand({
+        configureFailPoint: "failCollectionUpdates",
+        data: {collectionNS: "admin.system.version"},
+        mode: "off"
+    }));
+
     // featureCompatibilityVersion can be set to 3.4.
     assert.commandWorked(adminDB.runCommand({setFeatureCompatibilityVersion: "3.4"}));
     res = adminDB.runCommand({getParameter: 1, featureCompatibilityVersion: 1});
diff --git a/jstests/noPassthrough/ftdc_setdirectory.js b/jstests/noPassthrough/ftdc_setdirectory.js
new file mode 100644
index 00000000000..c1173055e2a
--- /dev/null
+++ b/jstests/noPassthrough/ftdc_setdirectory.js
@@ -0,0 +1,118 @@
+/**
+ * Test that verifies FTDC works in mongos.
+ */
+load('jstests/libs/ftdc.js');
+
+(function() {
+    'use strict';
+    let testPath1 = MongoRunner.toRealPath('ftdc_setdir1');
+    let testPath2 = MongoRunner.toRealPath('ftdc_setdir2');
+    let testPath3 = MongoRunner.toRealPath('ftdc_setdir3');
+    let testLog3 = testPath3 + "mongos_ftdc.log";
+
+    // Make the log file directory for mongos.
+    mkdir(testPath3);
+
+    // Startup 3 mongos:
+    // 1. Normal MongoS with no log file to verify FTDC can be startup at runtime with a path.
+    // 2. MongoS with explict diagnosticDataCollectionDirectoryPath setParameter at startup.
+    // 3. MongoS with log file to verify automatic FTDC path computation works.
+    let st = new ShardingTest({
+        shards: 1,
+        mongos: {
+            s0: {verbose: 0},
+            s1: {setParameter: {diagnosticDataCollectionDirectoryPath: testPath2}},
+            s2: {logpath: testLog3}
+        }
+    });
+
+    let admin1 = st.s0.getDB('admin');
+    let admin2 = st.s1.getDB('admin');
+    let admin3 = st.s2.getDB('admin');
+
+    function setParam(admin, obj) {
+        var ret = admin.runCommand(Object.extend({setParameter: 1}, obj));
+        return ret;
+    }
+
+    function getParam(admin, field) {
+        var q = {getParameter: 1};
+        q[field] = 1;
+
+        var ret = admin.runCommand(q);
+        assert.commandWorked(ret);
+        return ret[field];
+    }
+
+    // Verify FTDC can be started at runtime.
+    function verifyFTDCDisabledOnStartup() {
+        jsTestLog("Running verifyFTDCDisabledOnStartup");
+        verifyCommonFTDCParameters(admin1, false);
+
+        // 1. Try to enable and fail
+        assert.commandFailed(setParam(admin1, {"diagnosticDataCollectionEnabled": 1}));
+
+        // 2. Set path and succeed
+        assert.commandWorked(
+            setParam(admin1, {"diagnosticDataCollectionDirectoryPath": testPath2}));
+
+        // 3. Set path again and fail
+        assert.commandFailed(
+            setParam(admin1, {"diagnosticDataCollectionDirectoryPath": testPath2}));
+
+        // 4. Enable successfully
+        assert.commandWorked(setParam(admin1, {"diagnosticDataCollectionEnabled": 1}));
+
+        // 5. Validate getDiagnosticData returns FTDC data now
+        jsTestLog("Verifying FTDC getDiagnosticData");
+        verifyGetDiagnosticData(admin1);
+    }
+
+    // Verify FTDC is already running if there was a path set at startup.
+    function verifyFTDCStartsWithPath() {
+        jsTestLog("Running verifyFTDCStartsWithPath");
+        verifyCommonFTDCParameters(admin2, true);
+
+        // 1. Set path fail
+        assert.commandFailed(
+            setParam(admin2, {"diagnosticDataCollectionDirectoryPath": testPath2}));
+
+        // 2. Enable successfully
+        assert.commandWorked(setParam(admin2, {"diagnosticDataCollectionEnabled": 1}));
+
+        // 3. Validate getDiagnosticData returns FTDC data now
+        jsTestLog("Verifying FTDC getDiagnosticData");
+        verifyGetDiagnosticData(admin2);
+    }
+
+    function normpath(path) {
+        return path.replace(/\\/g, "/");
+    }
+
+    // Verify FTDC is already running if there was a path set at startup.
+    function verifyFTDCStartsWithLogFile() {
+        jsTestLog("Running verifyFTDCStartsWithLogFile");
+        verifyCommonFTDCParameters(admin3, true);
+
+        // 1. Verify that path is computed correctly.
+        let computedPath = getParam(admin3, "diagnosticDataCollectionDirectoryPath");
+        assert.eq(normpath(computedPath), normpath(testPath3 + "mongos_ftdc.diagnostic.data"));
+
+        // 2. Set path fail
+        assert.commandFailed(
+            setParam(admin3, {"diagnosticDataCollectionDirectoryPath": testPath2}));
+
+        // 3. Enable successfully
+        assert.commandWorked(setParam(admin3, {"diagnosticDataCollectionEnabled": 1}));
+
+        // 4. Validate getDiagnosticData returns FTDC data now
+        jsTestLog("Verifying FTDC getDiagnosticData");
+        verifyGetDiagnosticData(admin3);
+    }
+
+    verifyFTDCDisabledOnStartup();
+    verifyFTDCStartsWithPath();
+    verifyFTDCStartsWithLogFile();
+
+    st.stop();
+})();
diff --git a/jstests/noPassthrough/geo_near_fcv32.js b/jstests/noPassthrough/geo_near_fcv32.js
new file mode 100644
index 00000000000..10b331a411d
--- /dev/null
+++ b/jstests/noPassthrough/geo_near_fcv32.js
@@ -0,0 +1,85 @@
+/**
+ * Confirms that $geoNear aggregation and geoNear command succeed when FCV is 3.2.
+ */
+(function() {
+    "use strict";
+
+    const conn = MongoRunner.runMongod({});
+    assert.neq(null, conn, "mongod was unable to start up");
+
+    const testDB = conn.getDB("geo_near_fcv32");
+    testDB.test32.drop();
+    assert.commandWorked(testDB.adminCommand({setFeatureCompatibilityVersion: "3.2"}));
+
+    // Create 2dsphere index.
+    assert.commandWorked(testDB.test32.createIndex({loc: "2dsphere"}));
+
+    // Assert that $geoNear aggregate command does not fail due to collation errors when FCV is 3.2.
+    assert.eq(0,
+              testDB.test32
+                  .aggregate([{
+                      $geoNear: {
+                          near: {type: "Point", coordinates: [1.23, 1.23]},
+                          distanceField: "distance",
+                          spherical: true
+                      }
+                  }])
+                  .itcount());
+
+    // Assert that specifying the simple collation in $geoNear aggregate fails with FCV 3.2.
+    assert.throws(function() {
+        testDB.test32.aggregate([{
+            $geoNear: {
+                near: {type: "Point", coordinates: [1.23, 1.23]},
+                distanceField: "distance",
+                spherical: true,
+                collation: {locale: "simple"},
+            }
+        }]);
+    });
+
+    // Assert that specifying the simple collation in geoNear command fails with FCV 3.2.
+    assert.commandFailed(testDB.runCommand({
+        geoNear: "test32",
+        near: {type: "Point", coordinates: [1.23, 1.23]},
+        spherical: true,
+        collation: {locale: "simple"}
+    }));
+
+    assert.commandWorked(testDB.adminCommand({setFeatureCompatibilityVersion: "3.4"}));
+
+    // Create collection with case sensitive collation.
+    assert.commandWorked(
+        testDB.createCollection("test34", {collation: {locale: "en_US", strength: 2}}));
+    assert.commandWorked(testDB.test34.createIndex({loc: "2dsphere"}));
+    assert.writeOK(testDB.test34.insert({loc: [1.23, 1.23], str: "A"}));
+
+    // Assert that after upgrading FCV to 3.4 $geoNear aggregate inherits the collection's default
+    // collation.
+    assert.eq(1,
+              testDB.test34
+                  .aggregate([{
+                      $geoNear: {
+                          near: {type: "Point", coordinates: [1.23, 1.23]},
+                          distanceField: "distance",
+                          spherical: true,
+                          query: {str: "a"},
+                      }
+                  }])
+                  .itcount());
+
+    // Assert that the $geoNear aggregate accepts a specific collation and overrides the default
+    // collation in FCV 3.4.
+    assert.eq(1,
+              testDB.test34
+                  .aggregate([{
+                                $geoNear: {
+                                    near: {type: "Point", coordinates: [1.23, 1.23]},
+                                    distanceField: "distance",
+                                    spherical: true,
+                                    query: {str: "à"},
+                                }
+                             }],
+                             {collation: {locale: "en_US", strength: 1}})
+                  .itcount());
+})();
diff --git a/jstests/noPassthrough/libs/backup_restore.js b/jstests/noPassthrough/libs/backup_restore.js
index b4e02415081..e66f0f8af97 100644
--- a/jstests/noPassthrough/libs/backup_restore.js
+++ b/jstests/noPassthrough/libs/backup_restore.js
@@ -352,15 +352,31 @@ var BackupRestoreTest = function(options) {
         stopMongoProgramByPid(fsmPid);
 
         // Wait up to 5 minutes until the new hidden node is in state SECONDARY.
+        jsTestLog('CRUD and FSM clients stopped. Waiting for hidden node ' + hiddenHost +
+                  ' to become SECONDARY');
         rst.waitForState(hiddenNode, ReplSetTest.State.SECONDARY);
 
         // Wait for secondaries to finish catching up before shutting down.
+        jsTestLog(
+            'Hidden node ' + hiddenHost +
+            ' is now SECONDARY. Waiting for CRUD and FSM operations to be applied on all nodes.');
+        rst.awaitReplication();
+
+        jsTestLog('CRUD and FSM operations successfully applied on all nodes. ' +
+                  'Waiting for all nodes to agree on the primary.');
         rst.awaitNodesAgreeOnPrimary();
+
+        jsTestLog('All nodes agree on the primary. Getting current primary.');
         primary = rst.getPrimary();
-        assert.writeOK(primary.getDB("test").foo.insert(
-            {}, {writeConcern: {w: rst.nodes.length, wtimeout: 10 * 60 * 1000}}));
+
+        jsTestLog('Inserting single document into primary ' + primary.host +
+                  ' with writeConcern w:' + rst.nodes.length);
+        var writeResult = assert.writeOK(primary.getDB("test").foo.insert(
+            {}, {writeConcern: {w: rst.nodes.length, wtimeout: ReplSetTest.kDefaultTimeoutMS}}));
 
         // Stop set.
+        jsTestLog('Insert operation successful: ' + tojson(writeResult) +
+                  '. Stopping replica set.');
         rst.stopSet();
 
         // Cleanup the files from the test
diff --git a/jstests/noPassthrough/non_atomic_apply_ops_logging.js b/jstests/noPassthrough/non_atomic_apply_ops_logging.js
new file mode 100644
index 00000000000..f93e9d38189
--- /dev/null
+++ b/jstests/noPassthrough/non_atomic_apply_ops_logging.js
@@ -0,0 +1,62 @@
+// SERVER-28594 Ensure non-atomic ops are individually logged in applyOps
+// and atomic ops are collectively logged in applyOps.
+(function() {
+    "use strict";
+
+    let rst = new ReplSetTest({nodes: 1});
+    rst.startSet();
+    rst.initiate();
+
+    let primary = rst.getPrimary();
+    let testDB = primary.getDB("test");
+    let oplogColl = primary.getDB("local").oplog.rs;
+    let testCollName = "testColl";
+    let rerenamedCollName = "rerenamedColl";
+
+    testDB.runCommand({drop: testCollName});
+    testDB.runCommand({drop: rerenamedCollName});
+    assert.commandWorked(testDB.runCommand({create: testCollName}));
+    let testColl = testDB[testCollName];
+
+    // Ensure atomic apply ops logging only produces one oplog entry
+    // per call to apply ops and does not log individual operations
+    // separately.
+    assert.commandWorked(testDB.runCommand({
+        applyOps: [
+            {op: "i", ns: testColl.getFullName(), o: {_id: 1, a: "foo"}},
+            {op: "i", ns: testColl.getFullName(), o: {_id: 2, a: "bar"}}
+        ]
+    }));
+    assert.eq(oplogColl.find({"o.applyOps": {"$exists": true}}).count(), 1);
+    assert.eq(oplogColl.find({"op": "i"}).count(), 0);
+    // Ensure non-atomic apply ops logging produces an oplog entry for
+    // each operation in the apply ops call and no record of applyOps
+    // appears for these operations.
+    assert.commandWorked(testDB.runCommand({
+        applyOps: [
+            {
+              op: "c",
+              ns: "test.$cmd",
+              o: {
+                  renameCollection: "test.testColl",
+                  to: "test.renamedColl",
+                  stayTemp: false,
+                  dropTarget: false
+              }
+            },
+            {
+              op: "c",
+              ns: "test.$cmd",
+              o: {
+                  renameCollection: "test.renamedColl",
+                  to: "test." + rerenamedCollName,
+                  stayTemp: false,
+                  dropTarget: false
+              }
+            }
+        ]
+    }));
+    assert.eq(oplogColl.find({"o.renameCollection": {"$exists": true}}).count(), 2);
+    assert.eq(oplogColl.find({"o.applyOps": {"$exists": true}}).count(), 1);
+    rst.stopSet();
+})();
diff --git a/jstests/noPassthrough/partial_unique_indexes.js b/jstests/noPassthrough/partial_unique_indexes.js
new file mode 100644
index 00000000000..7c376872ade
--- /dev/null
+++ b/jstests/noPassthrough/partial_unique_indexes.js
@@ -0,0 +1,47 @@
+/**
+ * SERVER-32001: Test that indexing paths for non-unique, partial, unique, partial&unique
+ * crud operations correctly handle WriteConflictExceptions.
+ */
+(function() {
+    "strict";
+
+    let conn = MongoRunner.runMongod();
+    let testDB = conn.getDB("test");
+
+    let t = testDB.jstests_parallel_allops;
+    t.drop();
+
+    t.createIndex({x: 1, _id: 1}, {partialFilterExpression: {_id: {$lt: 500}}, unique: true});
+    t.createIndex({y: -1, _id: 1}, {unique: true});
+    t.createIndex({x: -1}, {partialFilterExpression: {_id: {$gte: 500}}, unique: false});
+    t.createIndex({y: 1}, {unique: false});
+
+    let _id = {"#RAND_INT": [0, 1000]};
+    let ops = [
+        {op: "remove", ns: t.getFullName(), query: {_id}},
+        {op: "update", ns: t.getFullName(), query: {_id}, update: {$inc: {x: 1}}, upsert: true},
+        {op: "update", ns: t.getFullName(), query: {_id}, update: {$inc: {y: 1}}, upsert: true},
+    ];
+
+    let seconds = 5;
+    let parallel = 5;
+    let host = testDB.getMongo().host;
+
+    let benchArgs = {ops, seconds, parallel, host};
+
+    assert.commandWorked(testDB.adminCommand({
+        configureFailPoint: 'WTWriteConflictExceptionForReads',
+        mode: {activationProbability: 0.01}
+    }));
+    assert.commandWorked(testDB.adminCommand(
+        {configureFailPoint: 'WTWriteConflictException', mode: {activationProbability: 0.01}}));
+    res = benchRun(benchArgs);
+    printjson({res});
+
+    assert.commandWorked(
+        testDB.adminCommand({configureFailPoint: 'WTWriteConflictException', mode: "off"}));
+    assert.commandWorked(
+        testDB.adminCommand({configureFailPoint: 'WTWriteConflictExceptionForReads', mode: "off"}));
+    res = t.validate();
+    assert(res.valid, tojson(res));
+})();
diff --git a/jstests/noPassthrough/skip_sharding_configuration_checks.js b/jstests/noPassthrough/skip_sharding_configuration_checks.js
new file mode 100644
index 00000000000..ff3e95440eb
--- /dev/null
+++ b/jstests/noPassthrough/skip_sharding_configuration_checks.js
@@ -0,0 +1,53 @@
+/**
+ *  Starts standalone RS with skipShardingConfigurationChecks.
+ *  @tags: [requires_persistence]
+ */
+(function() {
+    'use strict';
+
+    function expectState(rst, state) {
+        assert.soon(function() {
+            var status = rst.status();
+            if (status.myState != state) {
+                print("Waiting for state " + state + " in replSetGetStatus output: " +
+                      tojson(status));
+            }
+            return status.myState == state;
+        });
+    }
+
+    let configSvr = MongoRunner.runMongod(
+        {configsvr: "", setParameter: 'skipShardingConfigurationChecks=true'});
+    assert.eq(configSvr, null);
+
+    let shardSvr =
+        MongoRunner.runMongod({shardsvr: "", setParameter: 'skipShardingConfigurationChecks=true'});
+    assert.eq(shardSvr, null);
+
+    var st = new ShardingTest({name: "skipConfig", shards: {rs0: {nodes: 1}}});
+    var configRS = st.configRS;
+    var shardRS = st.rs0;
+
+    st.stopAllMongos();
+    shardRS.stopSet(15, true);
+    configRS.stopSet(undefined, true);
+
+    jsTestLog("Restarting configRS as a standalone ReplicaSet");
+
+    for (let i = 0; i < configRS.nodes.length; i++) {
+        delete configRS.nodes[i].fullOptions.configsvr;
+        configRS.nodes[i].fullOptions.setParameter = 'skipShardingConfigurationChecks=true';
+    }
+    configRS.startSet({}, true);
+    expectState(configRS, ReplSetTest.State.PRIMARY);
+    configRS.stopSet();
+
+    jsTestLog("Restarting shardRS as a standalone ReplicaSet");
+    for (let i = 0; i < shardRS.nodes.length; i++) {
+        delete shardRS.nodes[i].fullOptions.shardsvr;
+        shardRS.nodes[i].fullOptions.setParameter = 'skipShardingConfigurationChecks=true';
+    }
+    shardRS.startSet({}, true);
+    expectState(shardRS, ReplSetTest.State.PRIMARY);
+    shardRS.stopSet();
+})();
diff --git a/jstests/noPassthroughWithMongod/apply_ops_index_collation.js b/jstests/noPassthroughWithMongod/apply_ops_index_collation.js
new file mode 100644
index 00000000000..6248bc63b0f
--- /dev/null
+++ b/jstests/noPassthroughWithMongod/apply_ops_index_collation.js
@@ -0,0 +1,80 @@
+// Cannot implicitly shard accessed collections because of collection existing when none
+// expected.
+// @tags: [assumes_no_implicit_collection_creation_after_drop]
+
+// Tests creation of indexes using applyOps for collections with a non-simple default collation.
+// Indexes created through applyOps should be built exactly according to their index spec, without
+// inheriting the collection default collation, since this is how the oplog entries are replicated.
+// TODO SERVER-31435: Move this test into core once applyOps with createIndexes replicates
+// correctly.
+(function() {
+    "use strict";
+
+    load("jstests/libs/get_index_helpers.js");
+
+    const coll = db.apply_ops_index_collation;
+    coll.drop();
+    assert.commandWorked(db.createCollection(coll.getName(), {collation: {locale: "fr_CA"}}));
+
+    // An index created using an insert-style oplog entry with a non-simple collation does not
+    // inherit the collection default collation.
+    let res = assert.commandWorked(db.adminCommand({
+        applyOps: [{
+            op: "i",
+            ns: db.system.indexes.getFullName(),
+            o: {
+                v: 2,
+                key: {c: 1},
+                name: "c_1_en",
+                ns: coll.getFullName(),
+                collation: {
+                    locale: "en_US",
+                    caseLevel: false,
+                    caseFirst: "off",
+                    strength: 3,
+                    numericOrdering: false,
+                    alternate: "non-ignorable",
+                    maxVariable: "punct",
+                    normalization: false,
+                    backwards: false,
+                    version: "57.1"
+                }
+            }
+        }]
+    }));
+    let allIndexes = coll.getIndexes();
+    let spec = GetIndexHelpers.findByName(allIndexes, "c_1_en");
+    assert.neq(null, spec, "Index 'c_1_en' not found: " + tojson(allIndexes));
+    assert.eq(2, spec.v, tojson(spec));
+    assert.eq("en_US", spec.collation.locale, tojson(spec));
+
+    // An index created using an insert-style oplog entry with a simple collation does not inherit
+    // the collection default collation.
+    res = assert.commandWorked(db.adminCommand({
+        applyOps: [{
+            op: "i",
+            ns: db.system.indexes.getFullName(),
+            o: {v: 2, key: {c: 1}, name: "c_1", ns: coll.getFullName()}
+        }]
+    }));
+    allIndexes = coll.getIndexes();
+    spec = GetIndexHelpers.findByName(allIndexes, "c_1");
+    assert.neq(null, spec, "Index 'c_1' not found: " + tojson(allIndexes));
+    assert.eq(2, spec.v, tojson(spec));
+    assert(!spec.hasOwnProperty("collation"), tojson(spec));
+
+    // A v=1 index created using an insert-style oplog entry does not inherit the collection default
+    // collation.
+    res = assert.commandWorked(db.adminCommand({
+        applyOps: [{
+            op: "i",
+            ns: db.system.indexes.getFullName(),
+            o: {v: 1, key: {d: 1}, name: "d_1", ns: coll.getFullName()}
+        }]
+    }));
+    allIndexes = coll.getIndexes();
+    spec = GetIndexHelpers.findByName(allIndexes, "d_1");
+    assert.neq(null, spec, "Index 'd_1' not found: " + tojson(allIndexes));
+    assert.eq(1, spec.v, tojson(spec));
+    assert(!spec.hasOwnProperty("collation"), tojson(spec));
+})();
diff --git a/jstests/noPassthroughWithMongod/ftdc_params.js b/jstests/noPassthroughWithMongod/ftdc_params.js
index ced6c1b5675..08714040fcb 100644
--- a/jstests/noPassthroughWithMongod/ftdc_params.js
+++ b/jstests/noPassthroughWithMongod/ftdc_params.js
@@ -1,58 +1,10 @@
 // FTDC test cases
 //
+load('jstests/libs/ftdc.js');
+
 (function() {
     'use strict';
     var admin = db.getSiblingDB("admin");
 
-    // Check the defaults are correct
-    //
-    function getparam(field) {
-        var q = {getParameter: 1};
-        q[field] = 1;
-
-        var ret = admin.runCommand(q);
-        return ret[field];
-    }
-
-    // Verify the defaults are as we documented them
-    assert.eq(getparam("diagnosticDataCollectionEnabled"), true);
-    assert.eq(getparam("diagnosticDataCollectionPeriodMillis"), 1000);
-    assert.eq(getparam("diagnosticDataCollectionDirectorySizeMB"), 200);
-    assert.eq(getparam("diagnosticDataCollectionFileSizeMB"), 10);
-    assert.eq(getparam("diagnosticDataCollectionSamplesPerChunk"), 300);
-    assert.eq(getparam("diagnosticDataCollectionSamplesPerInterimUpdate"), 10);
-
-    function setparam(obj) {
-        var ret = admin.runCommand(Object.extend({setParameter: 1}, obj));
-        return ret;
-    }
-
-    assert.commandWorked(setparam({"diagnosticDataCollectionEnabled": 1}));
-    assert.commandWorked(setparam({"diagnosticDataCollectionPeriodMillis": 100}));
-    assert.commandWorked(setparam({"diagnosticDataCollectionDirectorySizeMB": 10}));
-    assert.commandWorked(setparam({"diagnosticDataCollectionFileSizeMB": 1}));
-    assert.commandWorked(setparam({"diagnosticDataCollectionSamplesPerChunk": 2}));
-    assert.commandWorked(setparam({"diagnosticDataCollectionSamplesPerInterimUpdate": 2}));
-
-    // Negative tests - set values below minimums
-    assert.commandFailed(setparam({"diagnosticDataCollectionPeriodMillis": 1}));
-    assert.commandFailed(setparam({"diagnosticDataCollectionDirectorySizeMB": 1}));
-    assert.commandFailed(setparam({"diagnosticDataCollectionSamplesPerChunk": 1}));
-    assert.commandFailed(setparam({"diagnosticDataCollectionSamplesPerInterimUpdate": 1}));
-
-    // Negative test - set file size bigger then directory size
-    assert.commandWorked(setparam({"diagnosticDataCollectionDirectorySizeMB": 10}));
-    assert.commandFailed(setparam({"diagnosticDataCollectionFileSizeMB": 100}));
-
-    // Negative test - set directory size less then file size
-    assert.commandWorked(setparam({"diagnosticDataCollectionDirectorySizeMB": 100}));
-    assert.commandWorked(setparam({"diagnosticDataCollectionFileSizeMB": 50}));
-    assert.commandFailed(setparam({"diagnosticDataCollectionDirectorySizeMB": 10}));
-
-    // Reset
-    assert.commandWorked(setparam({"diagnosticDataCollectionFileSizeMB": 10}));
-    assert.commandWorked(setparam({"diagnosticDataCollectionDirectorySizeMB": 200}));
-    assert.commandWorked(setparam({"diagnosticDataCollectionPeriodMillis": 1000}));
-    assert.commandWorked(setparam({"diagnosticDataCollectionSamplesPerChunk": 300}));
-    assert.commandWorked(setparam({"diagnosticDataCollectionSamplesPerInterimUpdate": 10}));
+    verifyCommonFTDCParameters(admin, true);
 })();
diff --git a/jstests/noPassthroughWithMongod/host_connection_string_validation.js b/jstests/noPassthroughWithMongod/host_connection_string_validation.js
index a252ef39230..f7bac51db88 100644
--- a/jstests/noPassthroughWithMongod/host_connection_string_validation.js
+++ b/jstests/noPassthroughWithMongod/host_connection_string_validation.js
@@ -89,8 +89,8 @@
         print("Testing " + (isGood ? "good" : "bad") + " connection string " + i + "...");
         print("    * testing " + connectionString);
         testHost(connectionString, isGood);
-        print("    * testing mongodb://" + connectionString);
-        testHost("mongodb://" + connectionString, isGood);
+        print("    * testing mongodb://" + encodeURIComponent(connectionString));
+        testHost("mongodb://" + encodeURIComponent(connectionString), isGood);
     }
 
     var i;
diff --git a/jstests/noPassthroughWithMongod/initial_sync_oplog_rollover.js b/jstests/noPassthroughWithMongod/initial_sync_oplog_rollover.js
index 3a47e14447c..d186a8f59ec 100644
--- a/jstests/noPassthroughWithMongod/initial_sync_oplog_rollover.js
+++ b/jstests/noPassthroughWithMongod/initial_sync_oplog_rollover.js
@@ -34,7 +34,7 @@
     assert.writeOK(coll.insert({a: 1}));
 
     function getFirstOplogEntry(conn) {
-        return conn.getDB('local').oplog.rs.find().sort({ts: 1}).limit(1)[0];
+        return conn.getDB('local').oplog.rs.find().sort({$natural: 1}).limit(1)[0];
     }
 
     var firstOplogEntry = getFirstOplogEntry(primary);
diff --git a/jstests/noPassthroughWithMongod/replset_host_connection_validation.js b/jstests/noPassthroughWithMongod/replset_host_connection_validation.js
index 7e48de363e2..e65387dd705 100644
--- a/jstests/noPassthroughWithMongod/replset_host_connection_validation.js
+++ b/jstests/noPassthroughWithMongod/replset_host_connection_validation.js
@@ -1,5 +1,6 @@
 // Test --host with a replica set.
 (function() {
+    'use strict';
 
     const replSetName = 'hostTestReplSetName';
 
@@ -26,35 +27,56 @@
 
         // Pass the inner test's exit code back as the outer test's exit code
         if (exitCode != 0) {
-            doassert("inner test failed with exit code " + exitcode);
+            doassert("inner test failed with exit code " + exitCode);
         }
         return;
     }
 
-    const testHost = function(host) {
-        const exitCode = runMongoProgram('mongo', '--eval', ';', '--host', host);
-        if (exitCode !== 0) {
-            doassert("failed to connect with `--host " + host +
-                     "`, but expected success. Exit code: " + exitCode);
+    function testHost(host, uri, ok) {
+        const exitCode = runMongoProgram('mongo', '--eval', ';', '--host', host, uri);
+        if (ok) {
+            assert.eq(exitCode, 0, "failed to connect with `--host " + host + "`");
+        } else {
+            assert.neq(exitCode, 0, "unexpectedly succeeded to connect with `--host " + host + "`");
         }
-    };
-
-    const connStrings = [
-        `localhost:${port}`,
-        `${replSetName}/localhost:${port}`,
-        `mongodb://localhost:${port}/admin?replicaSet=${replSetName}`,
-        `mongodb://localhost:${port}`,
-    ];
-
-    function runConnectionStringTestFor(i, connectionString) {
-        print("Testing connection string " + i + "...");
-        print("    * testing " + connectionString);
-        testHost(connectionString);
     }
 
-    for (let i = 0; i < connStrings.length; ++i) {
-        runConnectionStringTestFor(i, connStrings[i]);
+    function runConnectionStringTestFor(connectionString, uri, ok) {
+        print("* Testing: --host " + connectionString + " " + uri);
+        if (!ok) {
+            print("  This should fail");
+        }
+        testHost(connectionString, uri, ok);
+    }
+
+    function expSuccess(str) {
+        runConnectionStringTestFor(str, '', true);
+        if (!str.startsWith('mongodb://')) {
+            runConnectionStringTestFor(str, 'dbname', true);
+        }
     }
 
+    function expFailure(str) {
+        runConnectionStringTestFor(str, '', false);
+    }
+
+    expSuccess(`localhost:${port}`);
+    expSuccess(`${replSetName}/localhost:${port}`);
+    expSuccess(`${replSetName}/localhost:${port},[::1]:${port}`);
+    expSuccess(`${replSetName}/localhost:${port},`);
+    expSuccess(`${replSetName}/localhost:${port},,`);
+    expSuccess(`mongodb://localhost:${port}/admin?replicaSet=${replSetName}`);
+    expSuccess(`mongodb://localhost:${port}`);
+
+    expFailure(',');
+    expFailure(',,');
+    expFailure(`${replSetName}/`);
+    expFailure(`${replSetName}/,`);
+    expFailure(`${replSetName}/,,`);
+    expFailure(`${replSetName}//not/a/socket`);
+    expFailure(`mongodb://localhost:${port}/admin?replicaSet=`);
+    expFailure('mongodb://localhost:');
+    expFailure(`mongodb://:${port}`);
+
     jsTest.log("SUCCESSFUL test completion");
 })();
diff --git a/jstests/replsets/apply_batches_totalMillis.js b/jstests/replsets/apply_batches_totalMillis.js
new file mode 100644
index 00000000000..9e093211cb6
--- /dev/null
+++ b/jstests/replsets/apply_batches_totalMillis.js
@@ -0,0 +1,63 @@
+/**
+ * serverStatus.metrics.repl.apply.batches.totalMillis is a cumulative measure of how much time a
+ * node spends applying batches. This test checks that it includes the time spent waiting for
+ * batches to finish, by comparing the time recorded after replicating a small and a large load.
+ */
+
+(function() {
+    "use strict";
+
+    // Gets the value of metrics.repl.apply.batches.totalMillis.
+    function getTotalMillis(node) {
+        return assert.commandWorked(node.adminCommand({serverStatus: 1}))
+            .metrics.repl.apply.batches.totalMillis;
+    }
+
+    // Do a bulk insert of documents as: {{key: 0}, {key: 1}, {key: 2}, ... , {key: num-1}}
+    function performBulkInsert(coll, key, num) {
+        let bulk = coll.initializeUnorderedBulkOp();
+        for (let i = 0; i < num; i++) {
+            let doc = {};
+            doc[key] = i;
+            bulk.insert(doc);
+        }
+        assert.writeOK(bulk.execute());
+        rst.awaitReplication();
+    }
+
+    let name = "apply_batches_totalMillis";
+    let rst = new ReplSetTest({name: name, nodes: 2});
+    rst.startSet();
+    rst.initiate();
+
+    let primary = rst.getPrimary();
+    let secondary = rst.getSecondary();
+    let coll = primary.getDB(name)["foo"];
+
+    // Perform an initial write on the system and ensure steady state.
+    assert.writeOK(coll.insert({init: 0}));
+    rst.awaitReplication();
+    let baseTime = getTotalMillis(secondary);
+
+    // Introduce a small load and wait for it to be replicated.
+    performBulkInsert(coll, "small", 1000);
+
+    // Record the time spent applying the small load.
+    let timeAfterSmall = getTotalMillis(secondary);
+    let deltaSmall = timeAfterSmall - baseTime;
+
+    // Insert a significantly larger load.
+    performBulkInsert(coll, "large", 20000);
+
+    // Record the time spent applying the large load.
+    let timeAfterLarge = getTotalMillis(secondary);
+    let deltaLarge = timeAfterLarge - timeAfterSmall;
+
+    jsTestLog(`Recorded deltas: {small: ${deltaSmall}ms, large: ${deltaLarge}ms}.`);
+
+    // We should have recorded at least as much time on the second load as we did on the first.
+    // This is a crude comparison that is only taken to check that the timer is used correctly.
+    assert(deltaLarge >= deltaSmall, "Expected a higher net totalMillis for the larger load.");
+    rst.stopSet();
+
+})();
\ No newline at end of file
diff --git a/jstests/replsets/apply_ops_concurrent_non_atomic_different_db.js b/jstests/replsets/apply_ops_concurrent_non_atomic_different_db.js
new file mode 100644
index 00000000000..05cb6f9e996
--- /dev/null
+++ b/jstests/replsets/apply_ops_concurrent_non_atomic_different_db.js
@@ -0,0 +1,11 @@
+(function() {
+    'use strict';
+
+    load('jstests/replsets/libs/apply_ops_concurrent_non_atomic.js');
+
+    new ApplyOpsConcurrentNonAtomicTest({
+        ns1: 'test1.coll1',
+        ns2: 'test2.coll2',
+        requiresDocumentLevelConcurrency: false,
+    }).run();
+}());
diff --git a/jstests/replsets/apply_ops_concurrent_non_atomic_same_collection.js b/jstests/replsets/apply_ops_concurrent_non_atomic_same_collection.js
new file mode 100644
index 00000000000..004eeaaa52f
--- /dev/null
+++ b/jstests/replsets/apply_ops_concurrent_non_atomic_same_collection.js
@@ -0,0 +1,11 @@
+(function() {
+    'use strict';
+
+    load('jstests/replsets/libs/apply_ops_concurrent_non_atomic.js');
+
+    new ApplyOpsConcurrentNonAtomicTest({
+        ns1: 'test.coll',
+        ns2: 'test.coll',
+        requiresDocumentLevelConcurrency: true,
+    }).run();
+}());
diff --git a/jstests/replsets/apply_ops_concurrent_non_atomic_same_db.js b/jstests/replsets/apply_ops_concurrent_non_atomic_same_db.js
new file mode 100644
index 00000000000..10f874382a5
--- /dev/null
+++ b/jstests/replsets/apply_ops_concurrent_non_atomic_same_db.js
@@ -0,0 +1,11 @@
+(function() {
+    'use strict';
+
+    load('jstests/replsets/libs/apply_ops_concurrent_non_atomic.js');
+
+    new ApplyOpsConcurrentNonAtomicTest({
+        ns1: 'test.coll1',
+        ns2: 'test.coll2',
+        requiresDocumentLevelConcurrency: false,
+    }).run();
+}());
diff --git a/jstests/replsets/clean_shutdown_oplog_state.js b/jstests/replsets/clean_shutdown_oplog_state.js
index 5ac2f72d556..91b31fc66d7 100644
--- a/jstests/replsets/clean_shutdown_oplog_state.js
+++ b/jstests/replsets/clean_shutdown_oplog_state.js
@@ -10,7 +10,7 @@
     var rst = new ReplSetTest({
         name: "name",
         nodes: 2,
-        oplogSize: 100,
+        oplogSize: 500,
     });
 
     rst.startSet();
diff --git a/jstests/replsets/index_delete.js b/jstests/replsets/index_delete.js
index f7c45e2f2e0..1c352a18581 100644
--- a/jstests/replsets/index_delete.js
+++ b/jstests/replsets/index_delete.js
@@ -60,7 +60,7 @@ try {
         } else {
             return false;
         }
-    }, "index not started on secondary", 30000, 50);
+    }, "index not started on secondary");
 } finally {
     // Turn off failpoint and let the index build resumes.
     assert.commandWorked(
diff --git a/jstests/replsets/initial_sync_rename_collection_unsafe.js b/jstests/replsets/initial_sync_rename_collection_unsafe.js
new file mode 100644
index 00000000000..a105e1c9287
--- /dev/null
+++ b/jstests/replsets/initial_sync_rename_collection_unsafe.js
@@ -0,0 +1,61 @@
+/**
+ * Tests that renameCollection commands do not abort initial sync when users specify
+ * 'allowUnsafeRenamesDuringInitialSync'.
+ */
+
+(function() {
+    'use strict';
+
+    load("jstests/libs/check_log.js");
+
+    var parameters = TestData.setParameters;
+    if (parameters && parameters.indexOf("use3dot2InitialSync=true") != -1) {
+        jsTest.log("Skipping this test because use3dot2InitialSync was provided.");
+        return;
+    }
+
+    const basename = 'initial_sync_rename_collection_unsafe';
+
+    const rst = new ReplSetTest({name: basename, nodes: 1});
+    rst.startSet();
+    rst.initiate();
+
+    const dbName = 'd';
+    const primary = rst.getPrimary();
+    const primaryDB = primary.getDB(dbName);
+
+    assert.writeOK(primaryDB['foo'].save({}));
+
+    jsTestLog('Bring up a new node');
+    const secondary = rst.add({setParameter: {allowUnsafeRenamesDuringInitialSync: true}});
+    assert.commandWorked(secondary.adminCommand(
+        {configureFailPoint: 'initialSyncHangBeforeCopyingDatabases', mode: 'alwaysOn'}));
+    rst.reInitiate();
+    assert.eq(primary, rst.getPrimary(), 'Primary changed after reconfig');
+
+    // Wait for fail point message to be logged.
+    checkLog.contains(secondary,
+                      'initial sync - initialSyncHangBeforeCopyingDatabases fail point enabled');
+
+    jsTestLog('Rename collection on the primary');
+    assert.commandWorked(primaryDB['foo'].renameCollection('renamed'));
+
+    assert.commandWorked(secondary.adminCommand(
+        {configureFailPoint: 'initialSyncHangBeforeCopyingDatabases', mode: 'off'}));
+
+    checkLog.contains(secondary, 'allowUnsafeRenamesDuringInitialSync set to true');
+
+    jsTestLog('Wait for both nodes to be up-to-date');
+    rst.awaitSecondaryNodes();
+    rst.awaitReplication();
+
+    jsTestLog('Check that all collections were renamed correctly on the secondary');
+    const secondaryDB = secondary.getDB(dbName);
+    assert.eq(secondaryDB['renamed'].find().itcount(), 1, 'renamed collection does not exist');
+    assert.eq(secondaryDB['foo'].find().itcount(), 0, 'collection `foo` exists after rename');
+
+    let res = assert.commandWorked(secondary.adminCommand({replSetGetStatus: 1, initialSync: 1}));
+    assert.eq(res.initialSyncStatus.failedInitialSyncAttempts, 0);
+
+    rst.stopSet();
+})();
diff --git a/jstests/replsets/last_vote.js b/jstests/replsets/last_vote.js
index b9e0474217c..44d349f3237 100644
--- a/jstests/replsets/last_vote.js
+++ b/jstests/replsets/last_vote.js
@@ -134,9 +134,6 @@
               "replSetRequestVotes response had the wrong term: " + tojson(response));
     assert(!response.voteGranted,
            "node granted vote in term before last vote doc: " + tojson(response));
-    assert.eq(response.reason,
-              "candidate's term is lower than mine",
-              "replSetRequestVotes response had the wrong reason: " + tojson(response));
     assertNodeHasLastVote(node0, term, rst.nodes[0]);
     assertCurrentTerm(node0, term);
 
@@ -178,9 +175,6 @@
               "replSetRequestVotes response had the wrong term: " + tojson(response));
     assert(!response.voteGranted,
            "node granted vote in term of last vote doc: " + tojson(response));
-    assert.eq(response.reason,
-              "already voted for another candidate this term",
-              "replSetRequestVotes response had the wrong reason: " + tojson(response));
     assertNodeHasLastVote(node0, term, rst.nodes[0]);
     assertCurrentTerm(node0, term);
 
diff --git a/jstests/replsets/libs/apply_ops_concurrent_non_atomic.js b/jstests/replsets/libs/apply_ops_concurrent_non_atomic.js
new file mode 100644
index 00000000000..5e170378de2
--- /dev/null
+++ b/jstests/replsets/libs/apply_ops_concurrent_non_atomic.js
@@ -0,0 +1,232 @@
+/**
+ * This test ensures that multiple non-atomic applyOps commands can run concurrently.
+ * Prior to SERVER-29802, applyOps would acquire the global lock regardless of the
+ * atomicity of the operations (as a whole) being applied.
+ *
+ * Every instance of ApplyOpsConcurrentNonAtomicTest is configured with an "options" document
+ * with the following format:
+ * {
+ *     ns1: <string>,
+ *     ns1: <string>,
+ *     requiresDocumentLevelConcurrency: <bool>,
+ * }
+ *
+ * ns1:
+ *     Fully qualified namespace of first set of CRUD operations. For simplicity, only insert
+ *     operations will be used. The set of documents generated for the inserts into ns1 will have
+ *     _id values distinct from those generated for ns2.
+ *
+ * ns2:
+ *     Fully qualified namespace of second set of CRUD operations. This may be the same namespace as
+ *     ns1. As with ns1, only insert operations will be used.
+ *
+ * requiresDocumentLevelConcurrency:
+ *     Set to true if this test case can only be run with a storage engine that supports document
+ *     level concurrency.
+ */
+var ApplyOpsConcurrentNonAtomicTest = function(options) {
+    'use strict';
+
+    load('jstests/concurrency/fsm_workload_helpers/server_types.js');
+
+    if (!(this instanceof ApplyOpsConcurrentNonAtomicTest)) {
+        return new ApplyOpsConcurrentNonAtomicTest(options);
+    }
+
+    // Capture the 'this' reference
+    var self = this;
+
+    self.options = options;
+
+    /**
+     * Logs message using test name as prefix.
+     */
+    function testLog(message) {
+        jsTestLog('ApplyOpsConcurrentNonAtomicTest: ' + message);
+    }
+
+    /**
+     * Creates an array of insert operations for applyOps into collection 'coll'.
+     */
+    function generateInsertOps(coll, numOps, id) {
+        // Explicit 'use strict' to prevent mozjs from injecting its own "use strict" directive
+        // (with incorrect indentation) when we convert this function into a string for
+        // startParallelShell().
+        'use strict';
+        const ops = Array(numOps).fill('ignored').map((unused, i) => {
+            return {op: 'i', ns: coll.getFullName(), o: {_id: (id * numOps + i), id: id}};
+        });
+        return ops;
+    }
+
+    /**
+     * Runs applyOps in non-atomic mode to insert 'numOps' documents into collection 'coll'.
+     */
+    function applyOpsInsertNonAtomic(coll, numOps, id) {
+        'use strict';
+        const ops = generateInsertOps(coll, numOps, id);
+        const mydb = coll.getDB();
+        assert.commandWorked(mydb.runCommand({applyOps: ops, allowAtomic: false}),
+                             'failed to insert documents into ' + coll.getFullName());
+    }
+
+    /**
+     * Parses 'numOps' and collection namespace from 'options' and runs applyOps to inserted
+     * generated documents.
+     *
+     * options format:
+     * {
+     *     ns: <string>,
+     *     numOps: <int>,
+     *     id: <int>,
+     * }
+     *
+     * ns:
+     *     Fully qualified namespace of collection to insert documents into.
+     *
+     * numOps:
+     *     Number of insert operations to generate for applyOps command.
+     *
+     * id:
+     *     Index of collection for applyOps. Used with 'numOps' to generate _id values that will not
+     *     collide with collections with different indexes.
+     */
+    function insertFunction(options) {
+        'use strict';
+
+        const coll = db.getMongo().getCollection(options.ns);
+        const numOps = options.numOps;
+        const id = options.id;
+
+        testLog('Starting to apply ' + numOps + ' operations in collection ' + coll.getFullName());
+        applyOpsInsertNonAtomic(coll, numOps, id);
+        testLog('Successfully applied ' + numOps + ' operations in collection ' +
+                coll.getFullName());
+    }
+
+    /**
+     * Creates a function for startParallelShell() to run that will insert documents into
+     * collection 'coll' using applyOps.
+     */
+    function createInsertFunction(coll, numOps, id) {
+        const options = {
+            ns: coll.getFullName(),
+            numOps: numOps,
+            id: id,
+        };
+        const functionName = 'insertFunction_' + coll.getFullName().replace(/\./g, '_');
+        const s =                                                                     //
+            '\n\n' +                                                                  //
+            'const testLog = ' + testLog + ';\n\n' +                                  //
+            'const generateInsertOps = ' + generateInsertOps + ';\n\n' +              //
+            'const applyOpsInsertNonAtomic = ' + applyOpsInsertNonAtomic + ';\n\n' +  //
+            'const ' + functionName + ' = ' + insertFunction + ';\n\n' +              //
+            functionName + '(' + tojson(options) + ');';                              //
+        return s;
+    }
+
+    /**
+     * Returns number of insert operations reported by serverStatus.
+     * In 3.4 'opcountersRepl', not 'opcounters' was previously the correct field. Now non-atomic
+     * ops are now replicated as they are applied and are counted toward the global op counter.
+     */
+    function getInsertOpCount(serverStatus) {
+        return serverStatus.opcounters.insert;
+    }
+
+    /**
+     * Runs the test.
+     */
+    this.run = function() {
+        const options = this.options;
+
+        assert(options.ns1, 'collection 1 namespace not provided');
+        assert(options.ns2, 'collection 2 namespace not provided');
+
+        const replTest = new ReplSetTest({nodes: 1});
+        replTest.startSet();
+        replTest.initiate();
+
+        const primary = replTest.getPrimary();
+        const adminDb = primary.getDB('admin');
+
+        if (options.requiresDocumentLevelConcurrency &&
+            !supportsDocumentLevelConcurrency(adminDb)) {
+            testLog('Skipping test because storage engine does not support document level ' +
+                    'concurrency.');
+            return;
+        }
+
+        const coll1 = primary.getCollection(options.ns1);
+        const db1 = coll1.getDB();
+        const coll2 = primary.getCollection(options.ns2);
+        const db2 = coll2.getDB();
+
+        assert.commandWorked(db1.createCollection(coll1.getName()));
+        if (coll1.getFullName() !== coll2.getFullName()) {
+            assert.commandWorked(db2.createCollection(coll2.getName()));
+        }
+
+        // Enable fail point to pause applyOps between operations.
+        assert.commandWorked(primary.adminCommand(
+            {configureFailPoint: 'applyOpsPauseBetweenOperations', mode: 'alwaysOn'}));
+
+        // This logs each operation being applied.
+        const previousLogLevel =
+            assert.commandWorked(primary.setLogLevel(3, 'replication')).was.replication.verbosity;
+
+        testLog('Applying operations in collections ' + coll1.getFullName() + ' and ' +
+                coll2.getFullName());
+
+        const numOps = 100;
+        const insertProcess1 =
+            startParallelShell(createInsertFunction(coll1, numOps, 0), replTest.getPort(0));
+        const insertProcess2 =
+            startParallelShell(createInsertFunction(coll2, numOps, 1), replTest.getPort(0));
+
+        // The fail point will prevent applyOps from advancing past the first operation in each
+        // batch of operations. If applyOps is applying both sets of operations concurrently without
+        // holding the global lock, the insert opcounter will eventually be incremented to 2.
+        try {
+            let insertOpCount = 0;
+            const expectedFinalOpCount = 2;
+            assert.soon(
+                function() {
+                    const serverStatus = adminDb.serverStatus();
+                    insertOpCount = getInsertOpCount(serverStatus);
+                    // This assertion may fail if the fail point is not implemented correctly within
+                    // applyOps. This allows us to fail fast instead of waiting for the
+                    // assert.soon() function to time out.
+                    assert.lte(insertOpCount,
+                               expectedFinalOpCount,
+                               'Expected at most ' + expectedFinalOpCount +
+                                   ' documents inserted with fail point enabled. ' +
+                                   'Most recent insert operation count = ' + insertOpCount);
+                    return insertOpCount === expectedFinalOpCount;
+                },
+                'Insert operation count did not reach ' + expectedFinalOpCount +
+                    ' as expected with fail point enabled. Most recent insert operation count = ' +
+                    insertOpCount);
+        } finally {
+            assert.commandWorked(primary.adminCommand(
+                {configureFailPoint: 'applyOpsPauseBetweenOperations', mode: 'off'}));
+        }
+
+        insertProcess1();
+        insertProcess2();
+
+        testLog('Successfully applied operations in collections ' + coll1.getFullName() + ' and ' +
+                coll2.getFullName());
+
+        // Reset log level.
+        primary.setLogLevel(previousLogLevel, 'replication');
+
+        const serverStatus = adminDb.serverStatus();
+        assert.eq(200,
+                  getInsertOpCount(serverStatus),
+                  'incorrect number of insert operations in server status after applyOps: ' +
+                      tojson(serverStatus));
+
+        replTest.stopSet();
+    };
+};
diff --git a/jstests/replsets/libs/apply_ops_insert_write_conflict.js b/jstests/replsets/libs/apply_ops_insert_write_conflict.js
index 88b299c08b6..d0a47f5b6d8 100644
--- a/jstests/replsets/libs/apply_ops_insert_write_conflict.js
+++ b/jstests/replsets/libs/apply_ops_insert_write_conflict.js
@@ -36,14 +36,6 @@ var ApplyOpsInsertWriteConflictTest = function(options) {
             return {op: 'i', ns: t.getFullName(), o: {_id: i}};
         });
 
-        if (!options.atomic) {
-            // Adding a command to the list of operations to prevent the applyOps command from
-            // applying
-            // all the operations atomically.
-            ops.push({ns: "test.$cmd", op: "c", o: {applyOps: []}});
-            numOps++;
-        }
-
         // Probabilities for WCE are chosen based on empirical testing.
         // The probability for WCE during an atomic applyOps should be much smaller than that for
         // the non-atomic case because we have to attempt to re-apply the entire batch of 'numOps'
@@ -62,7 +54,7 @@ var ApplyOpsInsertWriteConflictTest = function(options) {
         var previousLogLevel =
             assert.commandWorked(primaryDB.setLogLevel(3, 'replication')).was.replication.verbosity;
 
-        var applyOpsResult = primaryDB.adminCommand({applyOps: ops});
+        var applyOpsResult = primaryDB.adminCommand({applyOps: ops, allowAtomic: options.atomic});
 
         // Reset log level.
         primaryDB.setLogLevel(previousLogLevel, 'replication');
diff --git a/jstests/replsets/no_flapping_during_network_partition.js b/jstests/replsets/no_flapping_during_network_partition.js
index 5a289b2afdd..daab9be2bca 100644
--- a/jstests/replsets/no_flapping_during_network_partition.js
+++ b/jstests/replsets/no_flapping_during_network_partition.js
@@ -39,7 +39,7 @@
     primary.disconnect(secondary);
 
     jsTestLog("Wait long enough for the secondary to call for an election.");
-    checkLog.contains(secondary, "can see a healthy primary of equal or greater priority");
+    checkLog.contains(secondary, "can see a healthy primary");
 
     jsTestLog("Verify the primary and secondary do not change during the partition.");
     assert.eq(primary, replTest.getPrimary());
diff --git a/jstests/replsets/noop_writes_wait_for_write_concern.js b/jstests/replsets/noop_writes_wait_for_write_concern.js
new file mode 100644
index 00000000000..61ffc518df4
--- /dev/null
+++ b/jstests/replsets/noop_writes_wait_for_write_concern.js
@@ -0,0 +1,235 @@
+/**
+ * This file tests that if a user initiates a write that becomes a noop due to being a duplicate
+ * operation, that we still wait for write concern. This is because we must wait for write concern
+ * on the write that made this a noop so that we can be sure it doesn't get rolled back if we
+ * acknowledge it.
+ */
+
+(function() {
+    "use strict";
+    load('jstests/libs/write_concern_util.js');
+
+    var name = 'noop_writes_wait_for_write_concern';
+    var replTest = new ReplSetTest({
+        name: name,
+        nodes: [{}, {rsConfig: {priority: 0}}, {rsConfig: {priority: 0}}],
+    });
+    replTest.startSet();
+    replTest.initiate();
+    // Stops node 1 so that all w:3 write concerns time out. We have 3 data bearing nodes so that
+    // 'dropDatabase' can satisfy its implicit writeConcern: majority but still time out from the
+    // explicit w:3 write concern.
+    replTest.stop(1);
+
+    var primary = replTest.getPrimary();
+    assert.eq(primary, replTest.nodes[0]);
+    var dbName = 'testDB';
+    var db = primary.getDB(dbName);
+    var collName = 'testColl';
+    var coll = db[collName];
+
+    function dropTestCollection() {
+        coll.drop();
+        assert.eq(0, coll.find().itcount(), "test collection not empty");
+    }
+
+    // Each entry in this array contains a command whose noop write concern behavior needs to be
+    // tested. Entries have the following structure:
+    // {
+    //      req: <object>,                   // Command request object that will result in a noop
+    //                                       // write after the setup function is called.
+    //
+    //      setupFunc: <function()>,         // Function to run to ensure that the request is a
+    //                                       // noop.
+    //
+    //      confirmFunc: <function(res)>,    // Function to run after the command is run to ensure
+    //                                       // that it executed properly. Accepts the result of
+    //                                       // the noop request to validate it.
+    // }
+    var commands = [];
+
+    commands.push({
+        req: {applyOps: [{op: "i", ns: coll.getFullName(), o: {_id: 1}}]},
+        setupFunc: function() {
+            assert.writeOK(coll.insert({_id: 1}));
+        },
+        confirmFunc: function(res) {
+            assert.commandWorked(res);
+            assert.eq(res.applied, 1);
+            assert.eq(res.results[0], true);
+            assert.eq(coll.find().itcount(), 1);
+            assert.eq(coll.count({_id: 1}), 1);
+        }
+    });
+
+    // 'update' where the document to update does not exist.
+    commands.push({
+        req: {update: collName, updates: [{q: {a: 1}, u: {b: 2}}]},
+        setupFunc: function() {
+            assert.writeOK(coll.insert({a: 1}));
+            assert.writeOK(coll.update({a: 1}, {b: 2}));
+        },
+        confirmFunc: function(res) {
+            assert.commandWorked(res);
+            assert.eq(res.n, 0);
+            assert.eq(res.nModified, 0);
+            assert.eq(coll.find().itcount(), 1);
+            assert.eq(coll.count({b: 2}), 1);
+        }
+    });
+
+    // 'update' where the update has already been done.
+    commands.push({
+        req: {update: collName, updates: [{q: {a: 1}, u: {$set: {b: 2}}}]},
+        setupFunc: function() {
+            assert.writeOK(coll.insert({a: 1}));
+            assert.writeOK(coll.update({a: 1}, {$set: {b: 2}}));
+        },
+        confirmFunc: function(res) {
+            assert.commandWorked(res);
+            assert.eq(res.n, 1);
+            assert.eq(res.nModified, 0);
+            assert.eq(coll.find().itcount(), 1);
+            assert.eq(coll.count({a: 1, b: 2}), 1);
+        }
+    });
+
+    commands.push({
+        req: {delete: collName, deletes: [{q: {a: 1}, limit: 1}]},
+        setupFunc: function() {
+            assert.writeOK(coll.insert({a: 1}));
+            assert.writeOK(coll.remove({a: 1}));
+        },
+        confirmFunc: function(res) {
+            assert.commandWorked(res);
+            assert.eq(res.n, 0);
+            assert.eq(coll.count({a: 1}), 0);
+        }
+    });
+
+    commands.push({
+        req: {createIndexes: collName, indexes: [{key: {a: 1}, name: "a_1"}]},
+        setupFunc: function() {
+            assert.writeOK(coll.insert({a: 1}));
+            assert.commandWorked(
+                db.runCommand({createIndexes: collName, indexes: [{key: {a: 1}, name: "a_1"}]}));
+        },
+        confirmFunc: function(res) {
+            assert.commandWorked(res);
+            assert.eq(res.numIndexesBefore, res.numIndexesAfter);
+        }
+    });
+
+    // 'findAndModify' where the document to update does not exist.
+    commands.push({
+        req: {findAndModify: collName, query: {a: 1}, update: {b: 2}},
+        setupFunc: function() {
+            assert.writeOK(coll.insert({a: 1}));
+            assert.commandWorked(
+                db.runCommand({findAndModify: collName, query: {a: 1}, update: {b: 2}}));
+        },
+        confirmFunc: function(res) {
+            assert.commandWorked(res);
+            assert.eq(res.lastErrorObject.updatedExisting, false);
+            assert.eq(coll.find().itcount(), 1);
+            assert.eq(coll.count({b: 2}), 1);
+        }
+    });
+
+    // 'findAndModify' where the update has already been done.
+    commands.push({
+        req: {findAndModify: collName, query: {a: 1}, update: {$set: {b: 2}}},
+        setupFunc: function() {
+            assert.writeOK(coll.insert({a: 1}));
+            assert.commandWorked(
+                db.runCommand({findAndModify: collName, query: {a: 1}, update: {$set: {b: 2}}}));
+        },
+        confirmFunc: function(res) {
+            assert.commandWorked(res);
+            assert.eq(res.lastErrorObject.updatedExisting, true);
+            assert.eq(coll.find().itcount(), 1);
+            assert.eq(coll.count({a: 1, b: 2}), 1);
+        }
+    });
+
+    commands.push({
+        req: {dropDatabase: 1},
+        setupFunc: function() {
+            assert.writeOK(coll.insert({a: 1}));
+            assert.commandWorked(db.runCommand({dropDatabase: 1}));
+        },
+        confirmFunc: function(res) {
+            assert.commandWorked(res);
+        }
+    });
+
+    commands.push({
+        req: {drop: collName},
+        setupFunc: function() {
+            assert.writeOK(coll.insert({a: 1}));
+            assert.commandWorked(db.runCommand({drop: collName}));
+        },
+        confirmFunc: function(res) {
+            assert.commandFailedWithCode(res, ErrorCodes.NamespaceNotFound);
+        }
+    });
+
+    commands.push({
+        req: {create: collName},
+        setupFunc: function() {
+            assert.commandWorked(db.runCommand({create: collName}));
+        },
+        confirmFunc: function(res) {
+            assert.commandFailedWithCode(res, ErrorCodes.NamespaceExists);
+        }
+    });
+
+    commands.push({
+        req: {insert: collName, documents: [{_id: 1}]},
+        setupFunc: function() {
+            assert.writeOK(coll.insert({_id: 1}));
+        },
+        confirmFunc: function(res) {
+            assert.commandWorked(res);
+            assert.eq(res.n, 0);
+            assert.eq(res.writeErrors[0].code, ErrorCodes.DuplicateKey);
+            assert.eq(coll.count({_id: 1}), 1);
+        }
+    });
+
+    function testCommandWithWriteConcern(cmd) {
+        // Provide a small wtimeout that we expect to time out.
+        cmd.req.writeConcern = {w: 3, wtimeout: 1000};
+        jsTest.log("Testing " + tojson(cmd.req));
+
+        dropTestCollection();
+
+        cmd.setupFunc();
+
+        // We run the command on a different connection. If the the command were run on the
+        // same connection, then the client last op for the noop write would be set by the setup
+        // operation. By using a fresh connection the client last op begins as null.
+        // This test explicitly tests that write concern for noop writes works when the
+        // client last op has not already been set by a duplicate operation.
+        var shell2 = new Mongo(primary.host);
+
+        // We check the error code of 'res' in the 'confirmFunc'.
+        var res = shell2.getDB(dbName).runCommand(cmd.req);
+
+        try {
+            // Tests that the command receives a write concern error. If we don't wait for write
+            // concern on noop writes then we won't get a write concern error.
+            assertWriteConcernError(res);
+            cmd.confirmFunc(res);
+        } catch (e) {
+            // Make sure that we print out the response.
+            printjson(res);
+            throw e;
+        }
+    }
+
+    commands.forEach(function(cmd) {
+        testCommandWithWriteConcern(cmd);
+    });
+
+})();
\ No newline at end of file
diff --git a/jstests/replsets/read_committed_with_catalog_changes.js b/jstests/replsets/read_committed_with_catalog_changes.js
index 02d26759e26..7928b16d3ae 100644
--- a/jstests/replsets/read_committed_with_catalog_changes.js
+++ b/jstests/replsets/read_committed_with_catalog_changes.js
@@ -214,7 +214,7 @@ load("jstests/replsets/rslib.js");       // For startSetIfSupportsReadMajority.
                                      "Expected read of " + coll.getFullName() + " to block");
     }
 
-    function assertReadsSucceed(coll, timeoutMs = 10000) {
+    function assertReadsSucceed(coll, timeoutMs = 20000) {
         var res =
             coll.runCommand('find', {"readConcern": {"level": "majority"}, "maxTimeMS": timeoutMs});
         assert.commandWorked(res, 'reading from ' + coll.getFullName());
diff --git a/jstests/replsets/startParallelShell.js b/jstests/replsets/startParallelShell.js
new file mode 100644
index 00000000000..beca88d19a9
--- /dev/null
+++ b/jstests/replsets/startParallelShell.js
@@ -0,0 +1,33 @@
+// Test startParallelShell() in a replica set.
+
+var db;
+
+(function() {
+    'use strict';
+
+    const setName = 'rs0';
+    const replSet = new ReplSetTest({name: setName, nodes: 3});
+    const nodes = replSet.nodeList();
+    replSet.startSet();
+    replSet.initiate();
+
+    const url = replSet.getURL();
+    print("* Connecting to " + url);
+    const mongo = new Mongo(url);
+    db = mongo.getDB('admin');
+    assert.eq(url, mongo.host, "replSet.getURL() should match active connection string");
+
+    print("* Starting parallel shell on --host " + db.getMongo().host);
+    startParallelShell('db.coll0.insert({test: "connString only"});');
+    assert.soon(function() {
+        return db.coll0.find({test: "connString only"}).count() === 1;
+    });
+
+    const uri = new MongoURI(url);
+    const port0 = uri.servers[0].port;
+    print("* Starting parallel shell w/ --port " + port0);
+    startParallelShell('db.coll0.insert({test: "explicit port"});', port0);
+    assert.soon(function() {
+        return db.coll0.find({test: "explicit port"}).count() === 1;
+    });
+})();
diff --git a/jstests/replsets/too_stale_secondary.js b/jstests/replsets/too_stale_secondary.js
index 369662e5f16..1a0c28b3454 100644
--- a/jstests/replsets/too_stale_secondary.js
+++ b/jstests/replsets/too_stale_secondary.js
@@ -34,7 +34,7 @@
     "use strict";
 
     function getFirstOplogEntry(conn) {
-        return conn.getDB('local').oplog.rs.find().sort({ts: 1}).limit(1)[0];
+        return conn.getDB('local').oplog.rs.find().sort({$natural: 1}).limit(1)[0];
     }
 
     /**
diff --git a/jstests/sharding/auth.js b/jstests/sharding/auth.js
index 61b25f10dde..037532acaec 100644
--- a/jstests/sharding/auth.js
+++ b/jstests/sharding/auth.js
@@ -38,7 +38,7 @@
         name: "auth",
         mongos: 1,
         shards: 0,
-        other: {keyFile: "jstests/libs/key1", chunkSize: 1, enableAutoSplit: true},
+        other: {keyFile: "jstests/libs/key1", chunkSize: 1, enableAutoSplit: false},
     });
 
     if (s.getDB('admin').runCommand('buildInfo').bits < 64) {
@@ -167,6 +167,7 @@
     s.getDB("test").foo.remove({});
 
     var num = 10000;
+    assert.commandWorked(s.s.adminCommand({split: "test.foo", middle: {x: num / 2}}));
     var bulk = s.getDB("test").foo.initializeUnorderedBulkOp();
     for (i = 0; i < num; i++) {
         bulk.insert(
diff --git a/jstests/sharding/auto_rebalance_parallel.js b/jstests/sharding/auto_rebalance_parallel.js
index 955319b8c5d..c7078a6898a 100644
--- a/jstests/sharding/auto_rebalance_parallel.js
+++ b/jstests/sharding/auto_rebalance_parallel.js
@@ -1,43 +1,71 @@
 /**
  * Tests that the cluster is balanced in parallel in one balancer round (standalone).
  */
+
 (function() {
     'use strict';
 
     var st = new ShardingTest({shards: 4});
+    var config = st.s0.getDB('config');
 
     assert.commandWorked(st.s0.adminCommand({enableSharding: 'TestDB'}));
     st.ensurePrimaryShard('TestDB', st.shard0.shardName);
-    assert.commandWorked(st.s0.adminCommand({shardCollection: 'TestDB.TestColl', key: {Key: 1}}));
 
-    var coll = st.s0.getDB('TestDB').TestColl;
+    function prepareCollectionForBalance(collName) {
+        assert.commandWorked(st.s0.adminCommand({shardCollection: collName, key: {Key: 1}}));
+
+        var coll = st.s0.getCollection(collName);
+
+        // Create 4 chunks initially and ensure they get balanced within 1 balancer round
+        assert.writeOK(coll.insert({Key: 1, Value: 'Test value 1'}));
+        assert.writeOK(coll.insert({Key: 10, Value: 'Test value 10'}));
+        assert.writeOK(coll.insert({Key: 20, Value: 'Test value 20'}));
+        assert.writeOK(coll.insert({Key: 30, Value: 'Test value 30'}));
+
+        assert.commandWorked(st.splitAt(collName, {Key: 10}));
+        assert.commandWorked(st.splitAt(collName, {Key: 20}));
+        assert.commandWorked(st.splitAt(collName, {Key: 30}));
+
+        // Move two of the chunks to shard0001 so we have option to do parallel balancing
+        assert.commandWorked(st.moveChunk(collName, {Key: 20}, st.shard1.shardName));
+        assert.commandWorked(st.moveChunk(collName, {Key: 30}, st.shard1.shardName));
 
-    // Create 4 chunks initially and ensure they get balanced within 1 balancer round
-    assert.writeOK(coll.insert({Key: 1, Value: 'Test value 1'}));
-    assert.writeOK(coll.insert({Key: 10, Value: 'Test value 10'}));
-    assert.writeOK(coll.insert({Key: 20, Value: 'Test value 20'}));
-    assert.writeOK(coll.insert({Key: 30, Value: 'Test value 30'}));
+        assert.eq(2, config.chunks.find({ns: collName, shard: st.shard0.shardName}).itcount());
+        assert.eq(2, config.chunks.find({ns: collName, shard: st.shard1.shardName}).itcount());
+    }
 
-    assert.commandWorked(st.splitAt('TestDB.TestColl', {Key: 10}));
-    assert.commandWorked(st.splitAt('TestDB.TestColl', {Key: 20}));
-    assert.commandWorked(st.splitAt('TestDB.TestColl', {Key: 30}));
+    function checkCollectionBalanced(collName) {
+        assert.eq(1, config.chunks.find({ns: collName, shard: st.shard0.shardName}).itcount());
+        assert.eq(1, config.chunks.find({ns: collName, shard: st.shard1.shardName}).itcount());
+        assert.eq(1, config.chunks.find({ns: collName, shard: st.shard2.shardName}).itcount());
+        assert.eq(1, config.chunks.find({ns: collName, shard: st.shard3.shardName}).itcount());
+    }
 
-    // Move two of the chunks to shard0001 so we have option to do parallel balancing
-    assert.commandWorked(st.moveChunk('TestDB.TestColl', {Key: 20}, st.shard1.shardName));
-    assert.commandWorked(st.moveChunk('TestDB.TestColl', {Key: 30}, st.shard1.shardName));
+    function countMoves(collName) {
+        return config.changelog.find({what: 'moveChunk.start', ns: collName}).itcount();
+    }
 
-    assert.eq(2, st.s0.getDB('config').chunks.find({shard: st.shard0.shardName}).itcount());
-    assert.eq(2, st.s0.getDB('config').chunks.find({shard: st.shard1.shardName}).itcount());
+    prepareCollectionForBalance('TestDB.TestColl1');
+    prepareCollectionForBalance('TestDB.TestColl2');
+
+    // Count the moveChunk start attempts accurately and ensure that only the correct number of
+    // migrations are scheduled
+    const testColl1InitialMoves = countMoves('TestDB.TestColl1');
+    const testColl2InitialMoves = countMoves('TestDB.TestColl2');
 
-    // Do enable the balancer and wait for a single balancer round
     st.startBalancer();
     st.awaitBalancerRound();
+    st.awaitBalancerRound();
     st.stopBalancer();
 
-    assert.eq(1, st.s0.getDB('config').chunks.find({shard: st.shard0.shardName}).itcount());
-    assert.eq(1, st.s0.getDB('config').chunks.find({shard: st.shard1.shardName}).itcount());
-    assert.eq(1, st.s0.getDB('config').chunks.find({shard: st.shard2.shardName}).itcount());
-    assert.eq(1, st.s0.getDB('config').chunks.find({shard: st.shard3.shardName}).itcount());
+    checkCollectionBalanced('TestDB.TestColl1');
+    checkCollectionBalanced('TestDB.TestColl2');
+
+    assert.eq(2, countMoves('TestDB.TestColl1') - testColl1InitialMoves);
+    assert.eq(2, countMoves('TestDB.TestColl2') - testColl2InitialMoves);
+
+    // Ensure there are no migration errors reported
+    assert.eq(0, config.changelog.find({what: 'moveChunk.error'}).itcount());
 
     st.stop();
 })();
diff --git a/jstests/sharding/autosplit.js b/jstests/sharding/autosplit.js
index bb34021487f..0eba386b6a3 100644
--- a/jstests/sharding/autosplit.js
+++ b/jstests/sharding/autosplit.js
@@ -4,7 +4,12 @@
 (function() {
     'use strict';
 
-    var s = new ShardingTest({name: "auto1", shards: 2, mongos: 1, other: {enableAutoSplit: true}});
+    var s = new ShardingTest({
+        name: "auto1",
+        shards: 2,
+        mongos: 1,
+        other: {enableAutoSplit: true, chunkSize: 10},
+    });
 
     assert.commandWorked(s.s0.adminCommand({enablesharding: "test"}));
     s.ensurePrimaryShard('test', 'shard0001');
diff --git a/jstests/sharding/mapReduce_inSharded_outSharded.js b/jstests/sharding/mapReduce_inSharded_outSharded.js
index d1aba2599f0..d73eb517e98 100644
--- a/jstests/sharding/mapReduce_inSharded_outSharded.js
+++ b/jstests/sharding/mapReduce_inSharded_outSharded.js
@@ -1,60 +1,70 @@
-var verifyOutput = function(out) {
-    printjson(out);
-    assert.eq(out.counts.input, 51200, "input count is wrong");
-    assert.eq(out.counts.emit, 51200, "emit count is wrong");
-    assert.gt(out.counts.reduce, 99, "reduce count is wrong");
-    assert.eq(out.counts.output, 512, "output count is wrong");
-};
-
-var st = new ShardingTest(
-    {shards: 2, verbose: 1, mongos: 1, other: {chunkSize: 1, enableBalancer: true}});
-
-st.adminCommand({enablesharding: "mrShard"});
-st.ensurePrimaryShard('mrShard', 'shard0001');
-st.adminCommand({shardcollection: "mrShard.srcSharded", key: {"_id": 1}});
-
-var db = st.getDB("mrShard");
-
-var bulk = db.srcSharded.initializeUnorderedBulkOp();
-for (j = 0; j < 100; j++) {
-    for (i = 0; i < 512; i++) {
-        bulk.insert({j: j, i: i});
+(function() {
+    "use strict";
+
+    var verifyOutput = function(out) {
+        printjson(out);
+        assert.eq(out.counts.input, 51200, "input count is wrong");
+        assert.eq(out.counts.emit, 51200, "emit count is wrong");
+        assert.gt(out.counts.reduce, 99, "reduce count is wrong");
+        assert.eq(out.counts.output, 512, "output count is wrong");
+    };
+
+    var st = new ShardingTest(
+        {shards: 2, verbose: 1, mongos: 1, other: {chunkSize: 1, enableBalancer: true}});
+
+    var admin = st.s0.getDB('admin');
+
+    assert.commandWorked(admin.runCommand({enablesharding: "mrShard"}));
+    st.ensurePrimaryShard('mrShard', 'shard0001');
+    assert.commandWorked(
+        admin.runCommand({shardcollection: "mrShard.srcSharded", key: {"_id": 1}}));
+
+    var db = st.s0.getDB("mrShard");
+
+    var bulk = db.srcSharded.initializeUnorderedBulkOp();
+    for (var j = 0; j < 100; j++) {
+        for (var i = 0; i < 512; i++) {
+            bulk.insert({j: j, i: i});
+        }
+    }
+    assert.writeOK(bulk.execute());
+
+    function map() {
+        emit(this.i, 1);
+    }
+    function reduce(key, values) {
+        return Array.sum(values);
     }
-}
-assert.writeOK(bulk.execute());
-
-function map() {
-    emit(this.i, 1);
-}
-function reduce(key, values) {
-    return Array.sum(values);
-}
-
-// sharded src sharded dst
-var suffix = "InShardedOutSharded";
-
-var out =
-    db.srcSharded.mapReduce(map, reduce, {out: {replace: "mrReplace" + suffix, sharded: true}});
-verifyOutput(out);
-
-out = db.srcSharded.mapReduce(map, reduce, {out: {merge: "mrMerge" + suffix, sharded: true}});
-verifyOutput(out);
-
-out = db.srcSharded.mapReduce(map, reduce, {out: {reduce: "mrReduce" + suffix, sharded: true}});
-verifyOutput(out);
-
-out = db.srcSharded.mapReduce(map, reduce, {out: {inline: 1}});
-verifyOutput(out);
-assert(out.results != 'undefined', "no results for inline");
-
-out = db.srcSharded.mapReduce(
-    map, reduce, {out: {replace: "mrReplace" + suffix, db: "mrShardOtherDB", sharded: true}});
-verifyOutput(out);
-
-out = db.runCommand({
-    mapReduce: "srcSharded",  // use new name mapReduce rather than mapreduce
-    map: map,
-    reduce: reduce,
-    out: "mrBasic" + "srcSharded",
-});
-verifyOutput(out);
+
+    // sharded src sharded dst
+    var suffix = "InShardedOutSharded";
+
+    var out =
+        db.srcSharded.mapReduce(map, reduce, {out: {replace: "mrReplace" + suffix, sharded: true}});
+    verifyOutput(out);
+
+    out = db.srcSharded.mapReduce(map, reduce, {out: {merge: "mrMerge" + suffix, sharded: true}});
+    verifyOutput(out);
+
+    out = db.srcSharded.mapReduce(map, reduce, {out: {reduce: "mrReduce" + suffix, sharded: true}});
+    verifyOutput(out);
+
+    out = db.srcSharded.mapReduce(map, reduce, {out: {inline: 1}});
+    verifyOutput(out);
+    assert(out.results != 'undefined', "no results for inline");
+
+    out = db.srcSharded.mapReduce(
+        map, reduce, {out: {replace: "mrReplace" + suffix, db: "mrShardOtherDB", sharded: true}});
+    verifyOutput(out);
+
+    out = db.runCommand({
+        mapReduce: "srcSharded",  // use new name mapReduce rather than mapreduce
+        map: map,
+        reduce: reduce,
+        out: "mrBasic" + "srcSharded",
+    });
+    verifyOutput(out);
+
+    st.stop();
+
+})();
diff --git a/jstests/sharding/migrateBig_balancer.js b/jstests/sharding/migrateBig_balancer.js
index 9eb50c2168e..03e98fa5493 100644
--- a/jstests/sharding/migrateBig_balancer.js
+++ b/jstests/sharding/migrateBig_balancer.js
@@ -31,11 +31,8 @@
 
     assert.eq(40, coll.count(), "prep1");
 
-    printjson(coll.stats());
-
-    admin.printShardingStatus();
-
-    admin.runCommand({shardcollection: "" + coll, key: {_id: 1}});
+    assert.commandWorked(admin.runCommand({shardcollection: "" + coll, key: {_id: 1}}));
+    st.printShardingStatus();
 
     assert.lt(
         5, mongos.getDB("config").chunks.find({ns: "test.stuff"}).count(), "not enough chunks");
@@ -56,5 +53,4 @@
     }, "never migrated", 10 * 60 * 1000, 1000);
 
     st.stop();
-
 })();
diff --git a/jstests/sharding/movechunk_commit_changelog_stats.js b/jstests/sharding/movechunk_commit_changelog_stats.js
new file mode 100644
index 00000000000..d257bb6ed94
--- /dev/null
+++ b/jstests/sharding/movechunk_commit_changelog_stats.js
@@ -0,0 +1,41 @@
+//
+// Tests that the changelog entry for moveChunk.commit contains stats on the migration.
+//
+
+(function() {
+    'use strict';
+
+    var st = new ShardingTest({mongos: 1, shards: 2});
+    var kDbName = 'db';
+
+    var mongos = st.s0;
+    var shard0 = st.shard0.shardName;
+    var shard1 = st.shard1.shardName;
+
+    assert.commandWorked(mongos.adminCommand({enableSharding: kDbName}));
+    st.ensurePrimaryShard(kDbName, shard0);
+
+    function assertCountsInChangelog() {
+        let changeLog = st.s.getDB('config').changelog.find({what: 'moveChunk.commit'}).toArray();
+        assert.gt(changeLog.length, 0);
+        for (let i = 0; i < changeLog.length; i++) {
+            assert(changeLog[i].details.hasOwnProperty('counts') ||
+                   changeLog[i].details.hasOwnProperty('clonedBytes'));
+        }
+    }
+
+    var ns = kDbName + '.fooHashed';
+    assert.commandWorked(mongos.adminCommand({shardCollection: ns, key: {_id: 'hashed'}}));
+
+    var aChunk = mongos.getDB('config').chunks.findOne({_id: RegExp(ns), shard: shard0});
+    assert(aChunk);
+
+    // Assert counts field exists in the changelog entry for moveChunk.commit
+    assert.commandWorked(
+        mongos.adminCommand({moveChunk: ns, bounds: [aChunk.min, aChunk.max], to: shard1}));
+    assertCountsInChangelog();
+
+    mongos.getDB(kDbName).fooHashed.drop();
+
+    st.stop();
+})();
\ No newline at end of file
diff --git a/jstests/sharding/printShardingStatus.js b/jstests/sharding/printShardingStatus.js
index 85330311fd4..cdc36999e51 100644
--- a/jstests/sharding/printShardingStatus.js
+++ b/jstests/sharding/printShardingStatus.js
@@ -7,6 +7,8 @@
 
     var st = new ShardingTest({shards: 1, mongos: 2, config: 1, other: {smallfiles: true}});
 
+    var standalone = MongoRunner.runMongod();
+
     var mongos = st.s0;
     var admin = mongos.getDB("admin");
 
@@ -85,9 +87,10 @@
     testBasicVerboseOnly(outputVerbose);
 
     // Take a copy of the config db, in order to test the harder-to-setup cases below.
+    // Copy into a standalone to also test running printShardingStatus() against a config dump.
     // TODO: Replace this manual copy with copydb once SERVER-13080 is fixed.
     var config = mongos.getDB("config");
-    var configCopy = mongos.getDB("configCopy");
+    var configCopy = standalone.getDB("configCopy");
     config.getCollectionInfos().forEach(function(c) {
         // Create collection with options.
         assert.commandWorked(configCopy.createCollection(c.name, c.options));
@@ -140,11 +143,11 @@
     configCopy.mongos.remove({});
 
     var output = grabStatusOutput(configCopy, false);
-    assertPresentInOutput(output, "most recently active mongoses:\n\tnone", "no mongoses");
+    assertPresentInOutput(output, "most recently active mongoses:\n        none", "no mongoses");
 
     var output = grabStatusOutput(configCopy, true);
     assertPresentInOutput(
-        output, "most recently active mongoses:\n\tnone", "no mongoses (verbose)");
+        output, "most recently active mongoses:\n        none", "no mongoses (verbose)");
 
     assert(mongos.getDB(dbName).dropDatabase());
 
@@ -237,5 +240,7 @@
 
     assert(mongos.getDB("test").dropDatabase());
 
+    MongoRunner.stopMongod(standalone);
+
     st.stop();
 })();
diff --git a/jstests/sharding/shard_existing_coll_chunk_count.js b/jstests/sharding/shard_existing_coll_chunk_count.js
new file mode 100644
index 00000000000..60145fef712
--- /dev/null
+++ b/jstests/sharding/shard_existing_coll_chunk_count.js
@@ -0,0 +1,165 @@
+/**
+ * This test confirms that after sharding a collection with some pre-existing data,
+ * the resulting chunks aren't auto-split too aggressively.
+ */
+(function() {
+    'use strict';
+
+    var s = new ShardingTest({
+        name: "shard_existing_coll_chunk_count",
+        shards: 1,
+        mongos: 1,
+        other: {enableAutoSplit: true},
+    });
+
+    assert.commandWorked(s.s.adminCommand({enablesharding: "test"}));
+
+    var collNum = 0;
+    var overhead = Object.bsonsize({_id: ObjectId(), i: 1, pad: ""});
+
+    var getNumberChunks = function(ns) {
+        return s.configRS.getPrimary().getDB("config").getCollection("chunks").count({ns});
+    };
+
+    var runCase = function(opts) {
+        // Expected options.
+        assert.gte(opts.docSize, 0);
+        assert.gte(opts.stages.length, 2);
+
+        // Compute padding.
+        if (opts.docSize < overhead) {
+            var pad = "";
+        } else {
+            var pad = (new Array(opts.docSize - overhead + 1)).join(' ');
+        }
+
+        collNum++;
+        var db = s.getDB("test");
+        var collName = "coll" + collNum;
+        var coll = db.getCollection(collName);
+        var i = 0;
+        var limit = 0;
+        var stageNum = 0;
+        var stage = opts.stages[stageNum];
+
+        // Insert initial docs.
+        var bulk = coll.initializeUnorderedBulkOp();
+        limit += stage.numDocsToInsert;
+        for (; i < limit; i++) {
+            bulk.insert({i, pad});
+        }
+        assert.writeOK(bulk.execute());
+
+        // Create shard key index.
+        assert.commandWorked(coll.createIndex({i: 1}));
+
+        // Shard collection.
+        assert.commandWorked(s.s.adminCommand({shardcollection: coll.getFullName(), key: {i: 1}}));
+
+        // Confirm initial number of chunks.
+        var numChunks = getNumberChunks(coll.getFullName());
+        assert.eq(numChunks,
+                  stage.expectedNumChunks,
+                  'in ' + coll.getFullName() + ' expected ' + stage.expectedNumChunks +
+                      ' initial chunks, but found ' + numChunks + '\nopts: ' + tojson(opts) +
+                      '\nchunks:\n' + s.getChunksString(coll.getFullName()));
+
+        // Do the rest of the stages.
+        for (stageNum = 1; stageNum < opts.stages.length; stageNum++) {
+            stage = opts.stages[stageNum];
+
+            // Insert the later docs (one at a time, to maximise the autosplit effects).
+            limit += stage.numDocsToInsert;
+            for (; i < limit; i++) {
+                coll.insert({i, pad});
+            }
+
+            // Confirm number of chunks for this stage.
+            var numChunks = getNumberChunks(coll.getFullName());
+            assert.eq(numChunks,
+                      stage.expectedNumChunks,
+                      'in ' + coll.getFullName() + ' expected ' + stage.expectedNumChunks +
+                          ' chunks for stage ' + stageNum + ', but found ' + numChunks +
+                          '\nopts: ' + tojson(opts) + '\nchunks:\n' +
+                          s.getChunksString(coll.getFullName()));
+        }
+    };
+
+    // Original problematic case.
+    runCase({
+        docSize: 0,
+        stages: [
+            {numDocsToInsert: 20000, expectedNumChunks: 1},
+            {numDocsToInsert: 7, expectedNumChunks: 1},
+            {numDocsToInsert: 1000, expectedNumChunks: 1},
+        ],
+    });
+
+    // Original problematic case (worse).
+    runCase({
+        docSize: 0,
+        stages: [
+            {numDocsToInsert: 90000, expectedNumChunks: 1},
+            {numDocsToInsert: 7, expectedNumChunks: 1},
+            {numDocsToInsert: 1000, expectedNumChunks: 1},
+        ],
+    });
+
+    // Pathological case #1.
+    runCase({
+        docSize: 522,
+        stages: [
+            {numDocsToInsert: 8191, expectedNumChunks: 1},
+            {numDocsToInsert: 2, expectedNumChunks: 1},
+            {numDocsToInsert: 1000, expectedNumChunks: 1},
+        ],
+    });
+
+    // Pathological case #2.
+    runCase({
+        docSize: 522,
+        stages: [
+            {numDocsToInsert: 8192, expectedNumChunks: 1},
+            {numDocsToInsert: 8192, expectedNumChunks: 1},
+        ],
+    });
+
+    // Lower chunksize to 1MB, and restart the mongos for it to take.
+    assert.writeOK(
+        s.getDB("config").getCollection("settings").update({_id: "chunksize"}, {$set: {value: 1}}, {
+            upsert: true
+        }));
+    s.restartMongos(0);
+
+    // Original problematic case, scaled down to smaller chunksize.
+    runCase({
+        docSize: 0,
+        stages: [
+            {numDocsToInsert: 10000, expectedNumChunks: 1},
+            {numDocsToInsert: 10, expectedNumChunks: 1},
+            {numDocsToInsert: 20, expectedNumChunks: 1},
+            {numDocsToInsert: 40, expectedNumChunks: 1},
+            {numDocsToInsert: 1000, expectedNumChunks: 1},
+        ],
+    });
+
+    // Docs just smaller than half chunk size.
+    runCase({
+        docSize: 510 * 1024,
+        stages: [
+            {numDocsToInsert: 10, expectedNumChunks: 6},
+            {numDocsToInsert: 10, expectedNumChunks: 12},
+        ],
+    });
+
+    // Docs just larger than half chunk size.
+    runCase({
+        docSize: 514 * 1024,
+        stages: [
+            {numDocsToInsert: 10, expectedNumChunks: 10},
+            {numDocsToInsert: 10, expectedNumChunks: 20},
+        ],
+    });
+
+    s.stop();
+})();
diff --git a/jstests/sharding/shard_identity_rollback.js b/jstests/sharding/shard_identity_rollback.js
index c7b6fedaacc..37cc8b10726 100644
--- a/jstests/sharding/shard_identity_rollback.js
+++ b/jstests/sharding/shard_identity_rollback.js
@@ -89,8 +89,9 @@
             }
         },
         function() {
-            var oldPriOplog = priConn.getDB('local').oplog.rs.find().sort({ts: -1}).toArray();
-            var newPriOplog = newPriConn.getDB('local').oplog.rs.find().sort({ts: -1}).toArray();
+            var oldPriOplog = priConn.getDB('local').oplog.rs.find().sort({$natural: -1}).toArray();
+            var newPriOplog =
+                newPriConn.getDB('local').oplog.rs.find().sort({$natural: -1}).toArray();
             return "timed out waiting for original primary to shut down after rollback. " +
                 "Old primary oplog: " + tojson(oldPriOplog) + "; new primary oplog: " +
                 tojson(newPriOplog);
diff --git a/jstests/sharding/write_cmd_auto_split.js b/jstests/sharding/write_cmd_auto_split.js
index 1cf9b5ab39a..95151b1e7e9 100644
--- a/jstests/sharding/write_cmd_auto_split.js
+++ b/jstests/sharding/write_cmd_auto_split.js
@@ -40,7 +40,7 @@
 
     assert.eq(1, configDB.chunks.find().itcount());
 
-    for (var x = 0; x < 1100; x++) {
+    for (var x = 0; x < 3100; x++) {
         assert.writeOK(testDB.runCommand({
             update: 'update',
             updates: [{q: {x: x}, u: {x: x, v: doc1k}, upsert: true}],
@@ -80,7 +80,7 @@
 
     // Note: Estimated 'chunk size' tracked by mongos is initialized with a random value so
     // we are going to be conservative.
-    for (var x = 0; x < 1100; x += 400) {
+    for (var x = 0; x < 3100; x += 400) {
         var docs = [];
 
         for (var y = 0; y < 400; y++) {
@@ -101,7 +101,7 @@
 
     assert.eq(1, configDB.chunks.find().itcount());
 
-    for (var x = 0; x < 1100; x += 400) {
+    for (var x = 0; x < 3100; x += 400) {
         var docs = [];
 
         for (var y = 0; y < 400; y++) {
diff --git a/rpm/init.d-mongod b/rpm/init.d-mongod
index 56539ef4d42..4e172b9f15c 100755
--- a/rpm/init.d-mongod
+++ b/rpm/init.d-mongod
@@ -64,6 +64,7 @@ start()
   ulimit -n 64000
   ulimit -m unlimited
   ulimit -u 64000
+  ulimit -l unlimited
 
   echo -n $"Starting mongod: "
   daemon --user "$MONGO_USER" --check $mongod "$NUMACTL $mongod $OPTIONS >/dev/null 2>&1"
diff --git a/rpm/init.d-mongod.suse b/rpm/init.d-mongod.suse
index fae1fb8b3f3..2216507a009 100644
--- a/rpm/init.d-mongod.suse
+++ b/rpm/init.d-mongod.suse
@@ -65,6 +65,7 @@ start()
   ulimit -n 64000
   ulimit -m unlimited
   ulimit -u 64000
+  ulimit -l unlimited
 
   echo -n "Starting mongod: "
   $NUMACTL /sbin/start_daemon -u "$MONGO_USER" -p "$PIDFILEPATH" $mongod $OPTIONS >/dev/null 2>&1
diff --git a/rpm/mongod.service b/rpm/mongod.service
index 708bf663614..e4d48d08c1e 100644
--- a/rpm/mongod.service
+++ b/rpm/mongod.service
@@ -13,6 +13,7 @@ ExecStartPre=/usr/bin/chown mongod:mongod /var/run/mongodb
 ExecStartPre=/usr/bin/chmod 0755 /var/run/mongodb
 PermissionsStartOnly=true
 PIDFile=/var/run/mongodb/mongod.pid
+Type=forking
 # file size
 LimitFSIZE=infinity
 # cpu time
@@ -23,6 +24,8 @@ LimitAS=infinity
 LimitNOFILE=64000
 # processes/threads
 LimitNPROC=64000
+# locked memory
+LimitMEMLOCK=infinity
 # total threads (user+kernel)
 TasksMax=infinity
 TasksAccounting=false
diff --git a/site_scons/site_tools/jstoh.py b/site_scons/site_tools/jstoh.py
index 62424a6cd4f..dc90b324b28 100644
--- a/site_scons/site_tools/jstoh.py
+++ b/site_scons/site_tools/jstoh.py
@@ -8,8 +8,8 @@ def jsToHeader(target, source):
 
     h = [
         '#include "mongo/base/string_data.h"',
+        '#include "mongo/scripting/engine.h"',
         'namespace mongo {',
-        'struct JSFile{ const char* name; const StringData& source; };',
         'namespace JSFiles{',
     ]
 
@@ -21,7 +21,7 @@ def jsToHeader(target, source):
         objname = os.path.split(filename)[1].split('.')[0]
         stringname = '_jscode_raw_' + objname
 
-        h.append('const char ' + stringname + "[] = {")
+        h.append('constexpr char ' + stringname + "[] = {")
 
         with open(filename, 'r') as f:
             for line in f:
@@ -30,8 +30,8 @@ def jsToHeader(target, source):
         h.append("0};")
         # symbols aren't exported w/o this
         h.append('extern const JSFile %s;' % objname)
-        h.append('const JSFile %s = { "%s", StringData(%s) };' %
-                 (objname, filename.replace('\\', '/'), stringname))
+        h.append('const JSFile %s = { "%s", StringData(%s, sizeof(%s) - 1) };' %
+                 (objname, filename.replace('\\', '/'), stringname, stringname))
 
     h.append("} // namespace JSFiles")
     h.append("} // namespace mongo")
diff --git a/src/mongo/SConscript b/src/mongo/SConscript
index 2cfb2060e7f..1208446d70f 100644
--- a/src/mongo/SConscript
+++ b/src/mongo/SConscript
@@ -341,6 +341,7 @@ env.Install(
         LIBDEPS=[
             'db/conn_pool_options',
             'db/commands/core',
+            'db/ftdc/ftdc_mongos',
             'db/mongodandmongos',
             's/client/sharding_connection_hook',
             's/commands/cluster_commands',
@@ -409,12 +410,17 @@ if not has_option('noshell') and usemozjs:
     mongo_shell = shellEnv.Program(
         "mongo",
         ["shell/dbshell.cpp"] + env.WindowsResourceFile("shell/shell.rc"),
-        LIBDEPS=["$BUILD_DIR/third_party/shim_pcrecpp",
-                 "shell_core",
-                 "db/server_options_core",
-                 "client/clientdriver",
-                 "$BUILD_DIR/mongo/util/password",
-                 ])
+        LIBDEPS=[
+            "$BUILD_DIR/third_party/shim_pcrecpp",
+            "shell_core",
+            "db/server_options_core",
+            "client/clientdriver",
+            "$BUILD_DIR/mongo/util/password",
+        ],
+        LIBDEPS_PRIVATE=[
+            "$BUILD_DIR/mongo/client/connection_string",
+        ]
+    )
 
     shellEnv.Install( '#/', mongo_shell )
 else:
diff --git a/src/mongo/base/error_codes.err b/src/mongo/base/error_codes.err
index c8def6bd695..98d41461fce 100644
--- a/src/mongo/base/error_codes.err
+++ b/src/mongo/base/error_codes.err
@@ -202,6 +202,12 @@ error_code("CannotBuildIndexKeys", 201)
 error_code("NetworkInterfaceExceededTimeLimit", 202)
 error_code("TooManyLocks", 208)
 error_code("UpdateOperationFailed", 218)
+error_code("FTDCPathNotSet", 219)
+error_code("FTDCPathAlreadySet", 220)
+error_code("ProducerConsumerQueueBatchTooLarge", 247)
+error_code("ProducerConsumerQueueEndClosed", 248)
+
+# Error codes 4000-8999 are reserved.
 
 # Non-sequential error codes (for compatibility only)
 error_code("SocketException", 9001)
diff --git a/src/mongo/base/parse_number_test.cpp b/src/mongo/base/parse_number_test.cpp
index 0ed768a7dd3..76b1018af21 100644
--- a/src/mongo/base/parse_number_test.cpp
+++ b/src/mongo/base/parse_number_test.cpp
@@ -290,9 +290,9 @@ TEST(Double, TestParsingNormal) {
     // not parseable by the Windows SDK libc or the Solaris libc in the mode we build.
     // See SERVER-14131.
 
-    ASSERT_PARSES(double, "0xff", 0xff);
-    ASSERT_PARSES(double, "-0xff", -0xff);
-    ASSERT_PARSES(double, "0xabcab.defdefP-10", 0xabcab.defdefP-10);
+    ASSERT_PARSES(double, "0xff", 255);
+    ASSERT_PARSES(double, "-0xff", -255);
+    ASSERT_PARSES(double, "0xabcab.defdefP-10", 687.16784283419838);
 #endif
 }
 
diff --git a/src/mongo/base/validate_locale.cpp b/src/mongo/base/validate_locale.cpp
index ffb7190c804..b51eac53edb 100644
--- a/src/mongo/base/validate_locale.cpp
+++ b/src/mongo/base/validate_locale.cpp
@@ -32,6 +32,7 @@
 #include <locale>
 
 #include "mongo/base/init.h"
+#include "mongo/util/mongoutils/str.h"
 
 namespace mongo {
 
@@ -40,13 +41,15 @@ MONGO_INITIALIZER_GENERAL(ValidateLocale, MONGO_NO_PREREQUISITES, MONGO_DEFAULT_
     try {
         // Validate that boost can correctly load the user's locale
         boost::filesystem::path("/").has_root_directory();
-    } catch (const std::runtime_error&) {
-        return Status(ErrorCodes::BadValue,
-                      "Invalid or no user locale set."
+    } catch (const std::runtime_error& e) {
+        return Status(
+            ErrorCodes::BadValue,
+            str::stream()
+                << "Invalid or no user locale set. "
 #ifndef _WIN32
-                      " Please ensure LANG and/or LC_* environment variables are set correctly."
+                << " Please ensure LANG and/or LC_* environment variables are set correctly. "
 #endif
-                      );
+                << e.what());
     }
 
 #ifdef _WIN32
diff --git a/src/mongo/bson/bson_comparator_interface_base.h b/src/mongo/bson/bson_comparator_interface_base.h
index f70bec7b872..29d80916f72 100644
--- a/src/mongo/bson/bson_comparator_interface_base.h
+++ b/src/mongo/bson/bson_comparator_interface_base.h
@@ -28,9 +28,11 @@
 
 #pragma once
 
+#include <boost/container/flat_set.hpp>
 #include <initializer_list>
 #include <map>
 #include <set>
+#include <vector>
 
 #include "mongo/base/disallow_copying.h"
 #include "mongo/base/string_data_comparator_interface.h"
@@ -51,6 +53,9 @@ class BSONComparatorInterfaceBase {
     MONGO_DISALLOW_COPYING(BSONComparatorInterfaceBase);
 
 public:
+    BSONComparatorInterfaceBase(BSONComparatorInterfaceBase&& other) = default;
+    BSONComparatorInterfaceBase& operator=(BSONComparatorInterfaceBase&& other) = default;
+
     /**
      * A deferred comparison between two objects of type T, which can be converted into a boolean
      * via the evaluate() method.
@@ -122,6 +127,8 @@ public:
 
     using Set = std::set<T, LessThan>;
 
+    using FlatSet = boost::container::flat_set<T, LessThan>;
+
     using UnorderedSet = stdx::unordered_set<T, Hasher, EqualTo>;
 
     template <typename ValueType>
@@ -207,6 +214,10 @@ protected:
         return Set(init, LessThan(this));
     }
 
+    FlatSet makeFlatSet(const std::vector<T>& elements) const {
+        return FlatSet(elements.begin(), elements.end(), LessThan(this));
+    }
+
     UnorderedSet makeUnorderedSet(std::initializer_list<T> init = {}) const {
         return UnorderedSet(init, 0, Hasher(this), EqualTo(this));
     }
diff --git a/src/mongo/bson/bsonelement_comparator_interface.h b/src/mongo/bson/bsonelement_comparator_interface.h
index c6223a2d352..d69397a68db 100644
--- a/src/mongo/bson/bsonelement_comparator_interface.h
+++ b/src/mongo/bson/bsonelement_comparator_interface.h
@@ -33,6 +33,9 @@
 
 namespace mongo {
 
+typedef std::set<BSONElement, BSONElementCmpWithoutField> BSONElementSet;
+typedef std::multiset<BSONElement, BSONElementCmpWithoutField> BSONElementMultiSet;
+
 /**
  * A BSONElement::ComparatorInterface is an abstract class for comparing BSONElement objects. Usage
  * for comparing two BSON elements, 'lhs' and 'rhs', where 'comparator' is an instance of a class
@@ -61,6 +64,14 @@ public:
     }
 
     /**
+     * Constructs a BSONEltFlatSet whose equivalence classes are given by this comparator. This
+     * comparator must outlive the returned set.
+     */
+    FlatSet makeBSONEltFlatSet(const std::vector<BSONElement>& elements) const {
+        return makeFlatSet(elements);
+    }
+
+    /**
      * Constructs a BSONEltUnorderedSet whose equivalence classes are given by this
      * comparator. This comparator must outlive the returned set.
      */
@@ -91,6 +102,8 @@ public:
 
 using BSONEltSet = BSONComparatorInterfaceBase<BSONElement>::Set;
 
+using BSONEltFlatSet = BSONComparatorInterfaceBase<BSONElement>::FlatSet;
+
 using BSONEltUnorderedSet = BSONComparatorInterfaceBase<BSONElement>::UnorderedSet;
 
 template <typename ValueType>
diff --git a/src/mongo/bson/bsonobj.cpp b/src/mongo/bson/bsonobj.cpp
index d9af2c706a9..bc0045f5681 100644
--- a/src/mongo/bson/bsonobj.cpp
+++ b/src/mongo/bson/bsonobj.cpp
@@ -33,6 +33,7 @@
 
 #include "mongo/base/data_range.h"
 #include "mongo/bson/bson_validate.h"
+#include "mongo/bson/bsonelement_comparator_interface.h"
 #include "mongo/db/json.h"
 #include "mongo/util/allocator.h"
 #include "mongo/util/hex.h"
diff --git a/src/mongo/bson/bsonobj.h b/src/mongo/bson/bsonobj.h
index 22ab6ff182f..0036fea6518 100644
--- a/src/mongo/bson/bsonobj.h
+++ b/src/mongo/bson/bsonobj.h
@@ -42,7 +42,6 @@
 #include "mongo/base/string_data_comparator_interface.h"
 #include "mongo/bson/bson_comparator_interface_base.h"
 #include "mongo/bson/bsonelement.h"
-#include "mongo/bson/bsonelement_comparator_interface.h"
 #include "mongo/bson/bsontypes.h"
 #include "mongo/bson/oid.h"
 #include "mongo/bson/timestamp.h"
@@ -53,9 +52,6 @@
 
 namespace mongo {
 
-typedef std::set<BSONElement, BSONElementCmpWithoutField> BSONElementSet;
-typedef std::multiset<BSONElement, BSONElementCmpWithoutField> BSONElementMSet;
-
 /**
    C++ representation of a "BSON" object -- that is, an extended JSON-style
    object in a binary representation.
diff --git a/src/mongo/bson/util/builder.h b/src/mongo/bson/util/builder.h
index b959c59236e..ba6945e3fbc 100644
--- a/src/mongo/bson/util/builder.h
+++ b/src/mongo/bson/util/builder.h
@@ -320,15 +320,16 @@ private:
     }
     /* "slow" portion of 'grow()'  */
     void NOINLINE_DECL grow_reallocate(int minSize) {
+        if (minSize > BufferMaxSize) {
+            std::stringstream ss;
+            ss << "BufBuilder attempted to grow() to " << minSize << " bytes, past the 64MB limit.";
+            msgasserted(13548, ss.str().c_str());
+        }
+
         int a = 64;
         while (a < minSize)
             a = a * 2;
 
-        if (a > BufferMaxSize) {
-            std::stringstream ss;
-            ss << "BufBuilder attempted to grow() to " << a << " bytes, past the 64MB limit.";
-            msgasserted(13548, ss.str().c_str());
-        }
         _buf.realloc(a);
         size = a;
     }
diff --git a/src/mongo/client/connection_string.h b/src/mongo/client/connection_string.h
index e8fc8eb2637..2e888b56dad 100644
--- a/src/mongo/client/connection_string.h
+++ b/src/mongo/client/connection_string.h
@@ -28,11 +28,13 @@
 
 #pragma once
 
+#include <sstream>
 #include <string>
 #include <vector>
 
 #include "mongo/base/status_with.h"
 #include "mongo/base/string_data.h"
+#include "mongo/bson/util/builder.h"
 #include "mongo/stdx/mutex.h"
 #include "mongo/util/assert_util.h"
 #include "mongo/util/net/hostandport.h"
@@ -157,6 +159,10 @@ public:
         return _string < other._string;
     }
 
+
+    friend std::ostream& operator<<(std::ostream&, const ConnectionString&);
+    friend StringBuilder& operator<<(StringBuilder&, const ConnectionString&);
+
 private:
     /**
      * Creates a SET connection string with the specified set name and servers.
@@ -179,4 +185,15 @@ private:
     static stdx::mutex _connectHookMutex;
     static ConnectionHook* _connectHook;
 };
+
+inline std::ostream& operator<<(std::ostream& ss, const ConnectionString& cs) {
+    ss << cs._string;
+    return ss;
+}
+
+inline StringBuilder& operator<<(StringBuilder& sb, const ConnectionString& cs) {
+    sb << cs._string;
+    return sb;
+}
+
 }  // namespace mongo
diff --git a/src/mongo/client/fetcher.cpp b/src/mongo/client/fetcher.cpp
index 741efe9afab..4918bbf76c2 100644
--- a/src/mongo/client/fetcher.cpp
+++ b/src/mongo/client/fetcher.cpp
@@ -171,7 +171,8 @@ Fetcher::Fetcher(executor::TaskExecutor* executor,
                  const BSONObj& findCmdObj,
                  const CallbackFn& work,
                  const BSONObj& metadata,
-                 Milliseconds timeout,
+                 Milliseconds findNetworkTimeout,
+                 Milliseconds getMoreNetworkTimeout,
                  std::unique_ptr<RemoteCommandRetryScheduler::RetryPolicy> firstCommandRetryPolicy)
     : _executor(executor),
       _source(source),
@@ -179,10 +180,11 @@ Fetcher::Fetcher(executor::TaskExecutor* executor,
       _cmdObj(findCmdObj.getOwned()),
       _metadata(metadata.getOwned()),
       _work(work),
-      _timeout(timeout),
+      _findNetworkTimeout(findNetworkTimeout),
+      _getMoreNetworkTimeout(getMoreNetworkTimeout),
       _firstRemoteCommandScheduler(
           _executor,
-          RemoteCommandRequest(_source, _dbname, _cmdObj, _metadata, nullptr, _timeout),
+          RemoteCommandRequest(_source, _dbname, _cmdObj, _metadata, nullptr, _findNetworkTimeout),
           stdx::bind(&Fetcher::_callback, this, stdx::placeholders::_1, kFirstBatchFieldName),
           std::move(firstCommandRetryPolicy)) {
     uassert(ErrorCodes::BadValue, "callback function cannot be null", work);
@@ -204,10 +206,6 @@ BSONObj Fetcher::getMetadataObject() const {
     return _metadata;
 }
 
-Milliseconds Fetcher::getTimeout() const {
-    return _timeout;
-}
-
 std::string Fetcher::toString() const {
     return getDiagnosticString();
 }
@@ -221,7 +219,8 @@ std::string Fetcher::getDiagnosticString() const {
     output << " query: " << _cmdObj;
     output << " query metadata: " << _metadata;
     output << " active: " << _isActive_inlock();
-    output << " timeout: " << _timeout;
+    output << " findNetworkTimeout: " << _findNetworkTimeout;
+    output << " getMoreNetworkTimeout: " << _getMoreNetworkTimeout;
     output << " shutting down?: " << _isShuttingDown_inlock();
     output << " first: " << _first;
     output << " firstCommandScheduler: " << _firstRemoteCommandScheduler.toString();
@@ -316,7 +315,8 @@ Status Fetcher::_scheduleGetMore(const BSONObj& cmdObj) {
     }
     StatusWith<executor::TaskExecutor::CallbackHandle> scheduleResult =
         _executor->scheduleRemoteCommand(
-            RemoteCommandRequest(_source, _dbname, cmdObj, _metadata, nullptr, _timeout),
+            RemoteCommandRequest(
+                _source, _dbname, cmdObj, _metadata, nullptr, _getMoreNetworkTimeout),
             stdx::bind(&Fetcher::_callback, this, stdx::placeholders::_1, kNextBatchFieldName));
 
     if (!scheduleResult.isOK()) {
diff --git a/src/mongo/client/fetcher.h b/src/mongo/client/fetcher.h
index 050ac50f249..3af4c6f329d 100644
--- a/src/mongo/client/fetcher.h
+++ b/src/mongo/client/fetcher.h
@@ -130,7 +130,8 @@ public:
             const BSONObj& cmdObj,
             const CallbackFn& work,
             const BSONObj& metadata = rpc::ServerSelectionMetadata(true, boost::none).toBSON(),
-            Milliseconds timeout = RemoteCommandRequest::kNoTimeout,
+            Milliseconds findNetworkTimeout = RemoteCommandRequest::kNoTimeout,
+            Milliseconds getMoreNetworkTimeout = RemoteCommandRequest::kNoTimeout,
             std::unique_ptr<RemoteCommandRetryScheduler::RetryPolicy> firstCommandRetryPolicy =
                 RemoteCommandRetryScheduler::makeNoRetryPolicy());
 
@@ -152,11 +153,6 @@ public:
     BSONObj getMetadataObject() const;
 
     /**
-     * Returns timeout for remote commands to complete.
-     */
-    Milliseconds getTimeout() const;
-
-    /**
      * Returns diagnostic information.
      */
     std::string getDiagnosticString() const;
@@ -260,7 +256,8 @@ private:
     executor::TaskExecutor::CallbackHandle _getMoreCallbackHandle;
 
     // Socket timeout
-    Milliseconds _timeout;
+    Milliseconds _findNetworkTimeout;
+    Milliseconds _getMoreNetworkTimeout;
 
     // First remote command scheduler.
     RemoteCommandRetryScheduler _firstRemoteCommandScheduler;
diff --git a/src/mongo/client/fetcher_test.cpp b/src/mongo/client/fetcher_test.cpp
index 00dae071d3a..66fb026c6b6 100644
--- a/src/mongo/client/fetcher_test.cpp
+++ b/src/mongo/client/fetcher_test.cpp
@@ -242,6 +242,7 @@ TEST_F(FetcherTest, InvalidConstruction) {
                 unreachableCallback,
                 rpc::makeEmptyMetadata(),
                 RemoteCommandRequest::kNoTimeout,
+                RemoteCommandRequest::kNoTimeout,
                 std::unique_ptr<RemoteCommandRetryScheduler::RetryPolicy>()),
         UserException,
         ErrorCodes::BadValue,
@@ -273,7 +274,6 @@ TEST_F(FetcherTest, RemoteCommandRequestShouldContainCommandParametersPassedToCo
     ASSERT_EQUALS(source, fetcher->getSource());
     ASSERT_BSONOBJ_EQ(findCmdObj, fetcher->getCommandObject());
     ASSERT_BSONOBJ_EQ(metadataObj, fetcher->getMetadataObject());
-    ASSERT_EQUALS(timeout, fetcher->getTimeout());
 
     ASSERT_OK(fetcher->schedule());
 
@@ -284,6 +284,7 @@ TEST_F(FetcherTest, RemoteCommandRequestShouldContainCommandParametersPassedToCo
         ASSERT_TRUE(net->hasReadyRequests());
         auto noi = net->getNextReadyRequest();
         request = noi->getRequest();
+        ASSERT_EQUALS(timeout, request.timeout);
     }
 
     ASSERT_EQUALS(source, request.target);
@@ -1048,6 +1049,7 @@ TEST_F(FetcherTest, FetcherAppliesRetryPolicyToFirstCommandButNotToGetMoreReques
                                          makeCallback(),
                                          rpc::makeEmptyMetadata(),
                                          executor::RemoteCommandRequest::kNoTimeout,
+                                         executor::RemoteCommandRequest::kNoTimeout,
                                          std::move(policy));
 
     callbackHook = appendGetMoreRequest;
diff --git a/src/mongo/client/mongo_uri.cpp b/src/mongo/client/mongo_uri.cpp
index e737479d05b..e2356881290 100644
--- a/src/mongo/client/mongo_uri.cpp
+++ b/src/mongo/client/mongo_uri.cpp
@@ -32,131 +32,323 @@
 
 #include "mongo/client/mongo_uri.h"
 
+#include <utility>
+
 #include "mongo/base/status_with.h"
 #include "mongo/bson/bsonobjbuilder.h"
 #include "mongo/client/dbclientinterface.h"
 #include "mongo/client/sasl_client_authenticate.h"
+#include "mongo/db/namespace_string.h"
+#include "mongo/util/hex.h"
 #include "mongo/util/mongoutils/str.h"
-#include "mongo/util/password_digest.h"
 
 #include <boost/algorithm/string/case_conv.hpp>
 #include <boost/algorithm/string/classification.hpp>
+#include <boost/algorithm/string/find_iterator.hpp>
 #include <boost/algorithm/string/predicate.hpp>
-#include <boost/algorithm/string/split.hpp>
-#include <boost/regex.hpp>
+
+namespace {
+constexpr std::array<char, 16> hexits{
+    '0', '1', '2', '3', '4', '5', '6', '7', '8', '9', 'A', 'B', 'C', 'D', 'E', 'F'};
+const mongo::StringData kURIPrefix{"mongodb://"};
+}
+
+/**
+ * RFC 3986 Section 2.1 - Percent Encoding
+ *
+ * Encode data elements in a way which will allow them to be embedded
+ * into a mongodb:// URI safely.
+ */
+void mongo::uriEncode(std::ostream& ss, StringData toEncode, StringData passthrough) {
+    for (const auto& c : toEncode) {
+        if ((c == '-') || (c == '_') || (c == '.') || (c == '~') || isalnum(c) ||
+            (passthrough.find(c) != std::string::npos)) {
+            ss << c;
+        } else {
+            // Encoding anything not included in section 2.3 "Unreserved characters"
+            ss << '%' << hexits[(c >> 4) & 0xF] << hexits[c & 0xF];
+        }
+    }
+}
+
+mongo::StatusWith<std::string> mongo::uriDecode(StringData toDecode) {
+    StringBuilder out;
+    for (size_t i = 0; i < toDecode.size(); ++i) {
+        const auto c = toDecode[i];
+        if (c == '%') {
+            if (i + 2 > toDecode.size()) {
+                return Status(ErrorCodes::FailedToParse,
+                              "Encountered partial escape sequence at end of string");
+            }
+            out << fromHex(toDecode.substr(i + 1, 2));
+            i += 2;
+        } else {
+            out << c;
+        }
+    }
+    return out.str();
+}
 
 namespace mongo {
 
 namespace {
-const char kMongoDBURL[] =
-    // scheme: non-capturing
-    "mongodb://"
 
-    // credentials: two inner captures for user and password
-    "(?:([^:]+)(?::([^@]+))?@)?"
+/**
+ * Helper Method for MongoURI::parse() to split a string into exactly 2 pieces by a char delimeter
+ */
+std::pair<StringData, StringData> partitionForward(StringData str, const char c) {
+    const auto delim = str.find(c);
+    if (delim == std::string::npos) {
+        return {str, StringData()};
+    }
+    return {str.substr(0, delim), str.substr(delim + 1)};
+}
 
-    // servers: grabs all host:port or UNIX socket names
-    "((?:[^\\/]+|/.+\\.sock)(?:,(?:[^\\/]+|/.+\\.sock))*)"
+/**
+ * Helper method for MongoURI::parse() to split a string into exactly 2 pieces by a char delimiter
+ * searching backward from the end of the string.
+ */
+std::pair<StringData, StringData> partitionBackward(StringData str, const char c) {
+    const auto delim = str.rfind(c);
+    if (delim == std::string::npos) {
+        return {StringData(), str};
+    }
+    return {str.substr(0, delim), str.substr(delim + 1)};
+}
 
-    // database and options are grouped together
-    "(?:/"
+/**
+ * Breakout method for parsing application/x-www-form-urlencoded option pairs
+ *
+ * foo=bar&baz=qux&...
+ */
+StatusWith<MongoURI::OptionsMap> parseOptions(StringData options, StringData url) {
+    MongoURI::OptionsMap ret;
+    if (options.empty()) {
+        return ret;
+    }
 
-    // database: matches anything but the chars that cannot be part of a MongoDB database name which
-    // are (in order) - forward slash, back slash, dot, space, double-quote, dollar sign, asterisk,
-    // less than, greater than, colon, pipe, question mark.
-    "([^/\\\\\\.\\ \"\\$*<>:\\|\\?]*)?"
+    if (options.find('?') != std::string::npos) {
+        return Status(ErrorCodes::FailedToParse,
+                      str::stream()
+                          << "URI Cannot Contain multiple questions marks for mongodb:// URL: "
+                          << url);
+    }
 
-    // options
-    "(?:\\?([^&=?]+=[^&=?]+(?:&[^&=?]+=[^&=?]+)*))?"
+    const auto optionsStr = options.toString();
+    for (auto i =
+             boost::make_split_iterator(optionsStr, boost::first_finder("&", boost::is_iequal()));
+         i != std::remove_reference<decltype((i))>::type{};
+         ++i) {
+        const auto opt = boost::copy_range<std::string>(*i);
+        if (opt.empty()) {
+            return Status(ErrorCodes::FailedToParse,
+                          str::stream()
+                              << "Missing a key/value pair in the options for mongodb:// URL: "
+                              << url);
+        }
 
-    // close db/options group
-    ")?";
+        const auto kvPair = partitionForward(opt, '=');
+        const auto keyRaw = kvPair.first;
+        if (keyRaw.empty()) {
+            return Status(
+                ErrorCodes::FailedToParse,
+                str::stream()
+                    << "Missing a key for key/value pair in the options for mongodb:// URL: "
+                    << url);
+        }
+        const auto key = uriDecode(keyRaw);
+        if (!key.isOK()) {
+            return Status(
+                ErrorCodes::FailedToParse,
+                str::stream() << "Key '" << keyRaw
+                              << "' in options cannot properly be URL decoded for mongodb:// URL: "
+                              << url);
+        }
+        const auto valRaw = kvPair.second;
+        if (valRaw.empty()) {
+            return Status(ErrorCodes::FailedToParse,
+                          str::stream() << "Missing value for key '" << keyRaw
+                                        << "' in the options for mongodb:// URL: "
+                                        << url);
+        }
+        const auto val = uriDecode(valRaw);
+        if (!val.isOK()) {
+            return Status(
+                ErrorCodes::FailedToParse,
+                str::stream() << "Value '" << valRaw << "' for key '" << keyRaw
+                              << "' in options cannot properly be URL decoded for mongodb:// URL: "
+                              << url);
+        }
 
-}  // namespace
+        ret[key.getValue()] = val.getValue();
+    }
+
+    return ret;
+}
 
+}  // namespace
 
 StatusWith<MongoURI> MongoURI::parse(const std::string& url) {
-    if (!boost::algorithm::starts_with(url, "mongodb://")) {
-        auto cs_status = ConnectionString::parse(url);
+    const StringData urlSD(url);
+
+    // 1. Validate and remove the scheme prefix mongodb://
+    if (!urlSD.startsWith(kURIPrefix)) {
+        const auto cs_status = ConnectionString::parse(url);
         if (!cs_status.isOK()) {
             return cs_status.getStatus();
         }
-
         return MongoURI(cs_status.getValue());
     }
+    const auto uriWithoutPrefix = urlSD.substr(kURIPrefix.size());
+
+    // 2. Split the string by the first, unescaped / (if any), yielding:
+    // split[0]: User information and host identifers
+    // split[1]: Auth database and connection options
+    const auto userAndDb = partitionForward(uriWithoutPrefix, '/');
+    const auto userAndHostInfo = userAndDb.first;
+    const auto databaseAndOptions = userAndDb.second;
+
+    // 2.b Make sure that there are no question marks in the left side of the /
+    //     as any options after the ? must still have the / delimeter
+    if (databaseAndOptions.empty() && userAndHostInfo.find('?') != std::string::npos) {
+        return Status(
+            ErrorCodes::FailedToParse,
+            str::stream()
+                << "URI must contain slash delimeter between hosts and options for mongodb:// URL: "
+                << url);
+    }
 
-    const boost::regex mongoUrlRe(kMongoDBURL);
-
-    boost::smatch matches;
-    if (!boost::regex_match(url, matches, mongoUrlRe)) {
+    // 3. Split the user information and host identifiers string by the last, unescaped @, yielding:
+    // split[0]: User information
+    // split[1]: Host identifiers;
+    const auto userAndHost = partitionBackward(userAndHostInfo, '@');
+    const auto userInfo = userAndHost.first;
+    const auto hostIdentifiers = userAndHost.second;
+
+    // 4. Validate, split (if applicable), and URL decode the user information, yielding:
+    // split[0] = username
+    // split[1] = password
+    const auto userAndPass = partitionForward(userInfo, ':');
+    const auto usernameSD = userAndPass.first;
+    const auto passwordSD = userAndPass.second;
+
+    const auto containsColonOrAt = [](StringData str) {
+        return (str.find(':') != std::string::npos) || (str.find('@') != std::string::npos);
+    };
+
+    if (containsColonOrAt(usernameSD)) {
         return Status(ErrorCodes::FailedToParse,
-                      str::stream() << "Failed to parse mongodb:// URL: " << url);
+                      str::stream() << "Username must be URL Encoded for mongodb:// URL: " << url);
     }
-
-    // We have the whole input plus 5 top level captures (user, password, host, db, options).
-    invariant(matches.size() == 6);
-
-    if (!matches[3].matched) {
-        return Status(ErrorCodes::FailedToParse, "No server(s) specified");
+    if (containsColonOrAt(passwordSD)) {
+        return Status(ErrorCodes::FailedToParse,
+                      str::stream() << "Password must be URL Encoded for mongodb:// URL: " << url);
     }
 
-    std::map<std::string, std::string> options;
+    // Get the username and make sure it did not fail to decode
+    const auto usernameWithStatus = uriDecode(usernameSD);
+    if (!usernameWithStatus.isOK()) {
+        return Status(
+            ErrorCodes::FailedToParse,
+            str::stream() << "Username cannot properly be URL decoded for mongodb:// URL: " << url);
+    }
+    const auto username = usernameWithStatus.getValue();
+
+    // Get the password and make sure it did not fail to decode
+    const auto passwordWithStatus = uriDecode(passwordSD);
+    if (!passwordWithStatus.isOK())
+        return Status(
+            ErrorCodes::FailedToParse,
+            str::stream() << "Password cannot properly be URL decoded for mongodb:// URL: " << url);
+    const auto password = passwordWithStatus.getValue();
+
+    // 5. Validate, split, and URL decode the host identifiers.
+    const auto hostIdentifiersStr = hostIdentifiers.toString();
+    std::vector<HostAndPort> servers;
+    for (auto i = boost::make_split_iterator(hostIdentifiersStr,
+                                             boost::first_finder(",", boost::is_iequal()));
+         i != std::remove_reference<decltype((i))>::type{};
+         ++i) {
+        const auto hostWithStatus = uriDecode(boost::copy_range<std::string>(*i));
+        if (!hostWithStatus.isOK()) {
+            return Status(
+                ErrorCodes::FailedToParse,
+                str::stream() << "Host cannot properly be URL decoded for mongodb:// URL: " << url);
+        }
 
-    if (matches[5].matched) {
-        const std::string optionsMatch = matches[5].str();
+        const auto host = hostWithStatus.getValue();
+        if (host.empty()) {
+            continue;
+        }
 
-        std::vector<boost::iterator_range<std::string::const_iterator>> optionsTokens;
-        boost::algorithm::split(optionsTokens, optionsMatch, boost::algorithm::is_any_of("=&"));
+        if ((host.find('/') != std::string::npos) && !StringData(host).endsWith(".sock")) {
+            return Status(
+                ErrorCodes::FailedToParse,
+                str::stream() << "'" << host << "' in '" << url
+                              << "' appears to be a unix socket, but does not end in '.sock'");
+        }
 
-        if (optionsTokens.size() % 2 != 0) {
-            return Status(ErrorCodes::FailedToParse,
-                          str::stream()
-                              << "Missing a key or value in the options for mongodb:// URL: "
-                              << url);
-            ;
+        const auto statusHostAndPort = HostAndPort::parse(host);
+        if (!statusHostAndPort.isOK()) {
+            return statusHostAndPort.getStatus();
         }
+        servers.push_back(statusHostAndPort.getValue());
+    }
+    if (servers.empty()) {
+        return Status(ErrorCodes::FailedToParse, "No server(s) specified");
+    }
 
-        for (size_t i = 0; i != optionsTokens.size(); i = i + 2)
-            options[std::string(optionsTokens[i].begin(), optionsTokens[i].end())] =
-                std::string(optionsTokens[i + 1].begin(), optionsTokens[i + 1].end());
+    // 6. Split the auth database and connection options string by the first, unescaped ?, yielding:
+    // split[0] = auth database
+    // split[1] = connection options
+    const auto dbAndOpts = partitionForward(databaseAndOptions, '?');
+    const auto databaseSD = dbAndOpts.first;
+    const auto connectionOptions = dbAndOpts.second;
+    const auto databaseWithStatus = uriDecode(databaseSD);
+    if (!databaseWithStatus.isOK()) {
+        return Status(ErrorCodes::FailedToParse,
+                      str::stream()
+                          << "Database name cannot properly be URL decoded for mongodb:// URL: "
+                          << url);
+    }
+    const auto database = databaseWithStatus.getValue();
+
+    // 7. Validate the database contains no prohibited characters
+    // Prohibited characters:
+    // slash ("/"), backslash ("\"), space (" "), double-quote ("""), or dollar sign ("$")
+    // period (".") is also prohibited, but drivers MAY allow periods
+    if (!database.empty() &&
+        !NamespaceString::validDBName(database,
+                                      NamespaceString::DollarInDbNameBehavior::Disallow)) {
+        return Status(ErrorCodes::FailedToParse,
+                      str::stream()
+                          << "Database name cannot have reserved characters for mongodb:// URL: "
+                          << url);
     }
 
-    OptionsMap::const_iterator optIter;
+    // 8. Validate, split, and URL decode the connection options
+    const auto optsWith = parseOptions(connectionOptions, url);
+    if (!optsWith.isOK()) {
+        return optsWith.getStatus();
+    }
+    const auto options = optsWith.getValue();
 
     // If a replica set option was specified, store it in the 'setName' field.
-    bool haveSetName;
+    const auto optIter = options.find("replicaSet");
     std::string setName;
-    if ((haveSetName = ((optIter = options.find("replicaSet")) != options.end()))) {
+    if (optIter != options.end()) {
         setName = optIter->second;
+        invariant(!setName.empty());
     }
 
-    std::vector<HostAndPort> servers;
-
-    {
-        std::vector<std::string> servers_split;
-        const std::string serversStr = matches[3].str();
-        boost::algorithm::split(servers_split, serversStr, boost::is_any_of(","));
-        for (auto&& s : servers_split) {
-            auto statusHostAndPort = HostAndPort::parse(s);
-            if (!statusHostAndPort.isOK()) {
-                return statusHostAndPort.getStatus();
-            }
-
-            servers.push_back(statusHostAndPort.getValue());
-        }
-    }
-
-    const bool direct = !haveSetName && (servers.size() == 1);
-
-    if (!direct && setName.empty()) {
+    if ((servers.size() > 1) && setName.empty()) {
         return Status(ErrorCodes::FailedToParse,
                       "Cannot list multiple servers in URL without 'replicaSet' option");
     }
 
     ConnectionString cs(
-        direct ? ConnectionString::MASTER : ConnectionString::SET, servers, setName);
-    return MongoURI(
-        std::move(cs), matches[1].str(), matches[2].str(), matches[4].str(), std::move(options));
+        setName.empty() ? ConnectionString::MASTER : ConnectionString::SET, servers, setName);
+    return MongoURI(std::move(cs), username, password, database, std::move(options));
 }
 
 }  // namespace mongo
diff --git a/src/mongo/client/mongo_uri.h b/src/mongo/client/mongo_uri.h
index 7b6364761d5..f7a7c4aeb81 100644
--- a/src/mongo/client/mongo_uri.h
+++ b/src/mongo/client/mongo_uri.h
@@ -29,12 +29,14 @@
 #pragma once
 
 #include <map>
+#include <sstream>
 #include <string>
 #include <vector>
 
 #include "mongo/base/status_with.h"
 #include "mongo/base/string_data.h"
 #include "mongo/bson/bsonobj.h"
+#include "mongo/bson/util/builder.h"
 #include "mongo/client/connection_string.h"
 #include "mongo/stdx/mutex.h"
 #include "mongo/util/assert_util.h"
@@ -43,12 +45,38 @@
 namespace mongo {
 
 /**
+ * Encode a string for embedding in a URI.
+ * Replaces reserved bytes with %xx sequences.
+ *
+ * Optionally allows passthrough characters to remain unescaped.
+ */
+void uriEncode(std::ostream& ss, StringData str, StringData passthrough = ""_sd);
+inline std::string uriEncode(StringData str, StringData passthrough = ""_sd) {
+    std::ostringstream ss;
+    uriEncode(ss, str, passthrough);
+    return ss.str();
+}
+
+/**
+ * Decode a URI encoded string.
+ * Replaces + and %xx sequences with their original byte.
+ */
+StatusWith<std::string> uriDecode(StringData str);
+
+/**
  * MongoURI handles parsing of URIs for mongodb, and falls back to old-style
  * ConnectionString parsing. It's used primarily by the shell.
  * It parses URIs with the following format:
  *
  *    mongodb://[usr:pwd@]host1[:port1]...[,hostN[:portN]]][/[db][?options]]
  *
+ * While this format is generally RFC 3986 compliant, some exceptions do exist:
+ *   1. The 'host' field, as defined by section 3.2.2 is expanded in the following ways:
+ *     a. Multiple hosts may be specified as a comma separated list.
+ *     b. Hosts may take the form of absolute paths for unix domain sockets.
+ *       i. Sockets must end in the suffix '.sock'
+ *   2. The 'fragment' field, as defined by section 3.5 is not permitted.
+ *
  * For a complete list of URI string options, see
  * https://wiki.mongodb.com/display/DH/Connection+String+Format
  *
@@ -125,6 +153,9 @@ public:
 
     MongoURI() = default;
 
+    friend std::ostream& operator<<(std::ostream&, const MongoURI&);
+    friend StringBuilder& operator<<(StringBuilder&, const MongoURI&);
+
 private:
     MongoURI(ConnectionString connectString,
              const std::string& user,
@@ -146,4 +177,14 @@ private:
     OptionsMap _options;
 };
 
+inline std::ostream& operator<<(std::ostream& ss, const MongoURI& uri) {
+    ss << uri._connectString;
+    return ss;
+}
+
+inline StringBuilder& operator<<(StringBuilder& sb, const MongoURI& uri) {
+    sb << uri._connectString;
+    return sb;
+}
+
 }  // namespace mongo
diff --git a/src/mongo/client/mongo_uri_test.cpp b/src/mongo/client/mongo_uri_test.cpp
index 6a5f6c9882a..c32b1c0b5c5 100644
--- a/src/mongo/client/mongo_uri_test.cpp
+++ b/src/mongo/client/mongo_uri_test.cpp
@@ -28,11 +28,17 @@
 
 #include "mongo/platform/basic.h"
 
-#include "mongo/client/mongo_uri.h"
+#include <fstream>
 
 #include "mongo/base/string_data.h"
+#include "mongo/bson/bsonobj.h"
+#include "mongo/bson/bsontypes.h"
+#include "mongo/bson/json.h"
+#include "mongo/client/mongo_uri.h"
 #include "mongo/unittest/unittest.h"
 
+#include <boost/filesystem/operations.hpp>
+
 namespace {
 using mongo::MongoURI;
 
@@ -60,16 +66,77 @@ const URITestCase validCases[] = {
 
     {"mongodb://user@127.0.0.1", "user", "", kMaster, "", 1, 0, ""},
 
-    {"mongodb://127.0.0.1/dbName?foo=a&c=b", "", "", kMaster, "", 1, 2, "dbName"},
-
     {"mongodb://localhost/?foo=bar", "", "", kMaster, "", 1, 1, ""},
 
+    {"mongodb://localhost,/?foo=bar", "", "", kMaster, "", 1, 1, ""},
+
     {"mongodb://user:pwd@127.0.0.1:1234", "user", "pwd", kMaster, "", 1, 0, ""},
 
     {"mongodb://user@127.0.0.1:1234", "user", "", kMaster, "", 1, 0, ""},
 
     {"mongodb://127.0.0.1:1234/dbName?foo=a&c=b", "", "", kMaster, "", 1, 2, "dbName"},
 
+    {"mongodb://127.0.0.1/dbName?foo=a&c=b", "", "", kMaster, "", 1, 2, "dbName"},
+
+    {"mongodb://user:pwd@127.0.0.1,/dbName?foo=a&c=b", "user", "pwd", kMaster, "", 1, 2, "dbName"},
+
+    {"mongodb://user:pwd@127.0.0.1,127.0.0.2/dbname?a=b&replicaSet=replName",
+     "user",
+     "pwd",
+     kSet,
+     "replName",
+     2,
+     2,
+     "dbname"},
+
+    {"mongodb://needs%20encoding%25%23!%3C%3E:pwd@127.0.0.1,127.0.0.2/"
+     "dbname?a=b&replicaSet=replName",
+     "needs encoding%#!<>",
+     "pwd",
+     kSet,
+     "replName",
+     2,
+     2,
+     "dbname"},
+
+    {"mongodb://needs%20encoding%25%23!%3C%3E:pwd@127.0.0.1,127.0.0.2/"
+     "db@name?a=b&replicaSet=replName",
+     "needs encoding%#!<>",
+     "pwd",
+     kSet,
+     "replName",
+     2,
+     2,
+     "db@name"},
+
+    {"mongodb://user:needs%20encoding%25%23!%3C%3E@127.0.0.1,127.0.0.2/"
+     "dbname?a=b&replicaSet=replName",
+     "user",
+     "needs encoding%#!<>",
+     kSet,
+     "replName",
+     2,
+     2,
+     "dbname"},
+
+    {"mongodb://user:pwd@127.0.0.1,127.0.0.2/dbname?a=b&replicaSet=needs%20encoding%25%23!%3C%3E",
+     "user",
+     "pwd",
+     kSet,
+     "needs encoding%#!<>",
+     2,
+     2,
+     "dbname"},
+
+    {"mongodb://user:pwd@127.0.0.1,127.0.0.2/needsencoding%40hello?a=b&replicaSet=replName",
+     "user",
+     "pwd",
+     kSet,
+     "replName",
+     2,
+     2,
+     "needsencoding@hello"},
+
     {"mongodb://user:pwd@127.0.0.1,127.0.0.2/?replicaSet=replName",
      "user",
      "pwd",
@@ -259,21 +326,27 @@ const URITestCase validCases[] = {
      1,
      2,
      ""},
-    {"mongodb:///tmp/mongodb-27017.sock", "", "", kMaster, "", 1, 0, ""},
 
-    {"mongodb:///tmp/mongodb-27017.sock,/tmp/mongodb-27018.sock/?replicaSet=replName",
+    {"mongodb://%2Ftmp%2Fmongodb-27017.sock", "", "", kMaster, "", 1, 0, ""},
+
+    {"mongodb://%2Ftmp%2Fmongodb-27017.sock,%2Ftmp%2Fmongodb-27018.sock/?replicaSet=replName",
      "",
      "",
      kSet,
      "replName",
      2,
      1,
-     ""}};
+     ""},
+};
 
 const InvalidURITestCase invalidCases[] = {
 
     // No host.
     {"mongodb://"},
+    {"mongodb://usr:pwd@/dbname?a=b"},
+
+    // Username and password must be encoded (cannot have ':' or '@')
+    {"mongodb://usr:pwd:@127.0.0.1/dbName?foo=a&c=b"},
 
     // Needs a "/" after the hosts and before the options.
     {"mongodb://localhost:27017,localhost:27018?replicaSet=missingSlash"},
@@ -282,32 +355,109 @@ const InvalidURITestCase invalidCases[] = {
     {"mongodb://localhost:27017localhost:27018"},
 
     // Domain sockets have to end in ".sock".
-    {"mongodb:///notareal/domainsock"},
+    {"mongodb://%2Fnotareal%2Fdomainsock"},
+
+    // Database name cannot contain slash ("/"), backslash ("\"), space (" "), double-quote ("""),
+    // or dollar sign ("$")
+    {"mongodb://usr:pwd@localhost:27017/db$name?a=b"},
+    {"mongodb://usr:pwd@localhost:27017/db/name?a=b"},
+    {"mongodb://usr:pwd@localhost:27017/db\\name?a=b"},
+    {"mongodb://usr:pwd@localhost:27017/db name?a=b"},
+    {"mongodb://usr:pwd@localhost:27017/db\"name?a=b"},
+
+    // Options must have a key
+    {"mongodb://usr:pwd@localhost:27017/dbname?=b"},
+
+    // Cannot skip a key value pair
+    {"mongodb://usr:pwd@localhost:27017/dbname?a=b&&b=c"},
+
+    // Multiple Unix domain sockets and auth DB resembling a socket (relative path)
+    {"mongodb://rel%2Fmongodb-27017.sock,rel%2Fmongodb-27018.sock/admin.sock?replicaSet=replName"},
+
+    // Multiple Unix domain sockets with auth DB resembling a path (relative path)
+    {"mongodb://rel%2Fmongodb-27017.sock,rel%2Fmongodb-27018.sock/admin.shoe?replicaSet=replName"},
+
+    // Multiple Unix domain sockets and auth DB resembling a socket (absolute path)
+    {"mongodb://%2Ftmp%2Fmongodb-27017.sock,%2Ftmp%2Fmongodb-27018.sock/"
+     "admin.sock?replicaSet=replName"},
+
+    // Multiple Unix domain sockets with auth DB resembling a path (absolute path)
+    {"mongodb://%2Ftmp%2Fmongodb-27017.sock,%2Ftmp%2Fmongodb-27018.sock/"
+     "admin.shoe?replicaSet=replName"},
+
+    // Missing value in key value pair for options
+    {"mongodb://127.0.0.1:1234/dbName?foo=a&c=b&d"},
+    {"mongodb://127.0.0.1:1234/dbName?foo=a&c=b&d="},
+    {"mongodb://127.0.0.1:1234/dbName?foo=a&h=&c=b&d=6"},
+    {"mongodb://127.0.0.1:1234/dbName?foo=a&h&c=b&d=6"},
+
+    // Missing a hostname, or unparsable hostname(s)
+    {"mongodb://,/dbName"},
+    {"mongodb://user:pwd@,/dbName"},
+    {"mongodb://localhost:1234:5678/dbName"},
 
     // Options can't have multiple question marks. Only one.
     {"mongodb://localhost:27017/?foo=a?c=b&d=e?asdf=foo"},
+
+    // Missing a key in key value pair for options
+    {"mongodb://127.0.0.1:1234/dbName?foo=a&=d&c=b"},
+
+    // Missing an entire key-value pair
+    {"mongodb://127.0.0.1:1234/dbName?foo=a&&c=b"},
 };
 
+// Helper Method to take a filename for a json file and return the array of tests inside of it
+mongo::BSONObj getBsonFromJsonFile(std::string fileName) {
+    boost::filesystem::path directoryPath = boost::filesystem::current_path();
+    boost::filesystem::path filePath(directoryPath / "src" / "mongo" / "client" /
+                                     "mongo_uri_tests" / fileName);
+    std::string filename(filePath.string());
+    std::ifstream infile(filename.c_str());
+    std::string data((std::istreambuf_iterator<char>(infile)), std::istreambuf_iterator<char>());
+    if (data.empty()) {
+        // v3.4 test infra is unable to see these extra files.
+        return mongo::BSONObj();
+    }
+    mongo::BSONObj obj = mongo::fromjson(data);
+    ASSERT_TRUE(obj.valid(mongo::BSONVersion::kLatest));
+    ASSERT_TRUE(obj.hasField("tests"));
+    mongo::BSONObj arr = obj.getField("tests").embeddedObject().getOwned();
+    ASSERT_TRUE(arr.couldBeArray());
+    return arr;
+}
+
+// Helper method to take a BSONElement and either extract its string or return an empty string
+std::string returnStringFromElementOrNull(mongo::BSONElement element) {
+    ASSERT_TRUE(!element.eoo());
+    if (element.type() == mongo::jstNULL) {
+        return std::string();
+    }
+    ASSERT_EQ(element.type(), mongo::String);
+    return element.String();
+}
+
+// Helper method to take a valid test case, parse() it, and assure the output is correct
+void testValidURIFormat(URITestCase testCase) {
+    mongo::unittest::log() << "Testing URI: " << testCase.URI << '\n';
+    std::string errMsg;
+    const auto cs_status = MongoURI::parse(testCase.URI);
+    ASSERT_OK(cs_status);
+    auto result = cs_status.getValue();
+    ASSERT_EQ(testCase.uname, result.getUser());
+    ASSERT_EQ(testCase.password, result.getPassword());
+    ASSERT_EQ(testCase.type, result.type());
+    ASSERT_EQ(testCase.setname, result.getSetName());
+    ASSERT_EQ(testCase.numservers, result.getServers().size());
+    ASSERT_EQ(testCase.numOptions, result.getOptions().size());
+    ASSERT_EQ(testCase.database, result.getDatabase());
+}
+
 TEST(MongoURI, GoodTrickyURIs) {
     const size_t numCases = sizeof(validCases) / sizeof(validCases[0]);
 
     for (size_t i = 0; i != numCases; ++i) {
         const URITestCase testCase = validCases[i];
-        mongo::unittest::log() << "Testing URI: " << testCase.URI << '\n';
-        std::string errMsg;
-        auto cs_status = MongoURI::parse(testCase.URI);
-        if (!cs_status.getStatus().toString().empty()) {
-            mongo::unittest::log() << "error with uri: " << cs_status.getStatus().toString();
-        }
-        ASSERT_TRUE(cs_status.isOK());
-        auto result = cs_status.getValue();
-        ASSERT_EQ(testCase.uname, result.getUser());
-        ASSERT_EQ(testCase.password, result.getPassword());
-        ASSERT_EQ(testCase.type, result.type());
-        ASSERT_EQ(testCase.setname, result.getSetName());
-        ASSERT_EQ(testCase.numservers, result.getServers().size());
-        ASSERT_EQ(testCase.numOptions, result.getOptions().size());
-        ASSERT_EQ(testCase.database, result.getDatabase());
+        testValidURIFormat(testCase);
     }
 }
 
@@ -318,7 +468,7 @@ TEST(MongoURI, InvalidURIs) {
         const InvalidURITestCase testCase = invalidCases[i];
         mongo::unittest::log() << "Testing URI: " << testCase.URI << '\n';
         auto cs_status = MongoURI::parse(testCase.URI);
-        ASSERT_FALSE(cs_status.isOK());
+        ASSERT_NOT_OK(cs_status);
     }
 }
 
@@ -357,4 +507,100 @@ TEST(MongoURI, CloneURIForServer) {
     ASSERT_EQ(clonedURIOptions.at("ssl"), "true");
 }
 
+/**
+ * These tests come from the Mongo Uri Specifications for the drivers found at:
+ * https://github.com/mongodb/specifications/tree/master/source/connection-string/tests
+ * They have been slighly altered as the Drivers specification is slighly different
+ * from the server specification.
+ */
+TEST(MongoURI, specTests) {
+    const std::string files[] = {
+        "mongo-uri-valid-auth.json",
+        "mongo-uri-options.json",
+        "mongo-uri-unix-sockets-absolute.json",
+        "mongo-uri-unix-sockets-relative.json",
+        "mongo-uri-warnings.json",
+        "mongo-uri-host-identifiers.json",
+        "mongo-uri-invalid.json",
+    };
+
+    for (const auto& file : files) {
+        const auto testBson = getBsonFromJsonFile(file);
+
+        for (const auto& testElement : testBson) {
+            ASSERT_EQ(testElement.type(), mongo::Object);
+            const auto test = testElement.Obj();
+
+            // First extract the valid field and the uri field
+            const auto validDoc = test.getField("valid");
+            ASSERT_FALSE(validDoc.eoo());
+            ASSERT_TRUE(validDoc.isBoolean());
+            const auto valid = validDoc.Bool();
+
+            const auto uriDoc = test.getField("uri");
+            ASSERT_FALSE(uriDoc.eoo());
+            ASSERT_EQ(uriDoc.type(), mongo::String);
+            const auto uri = uriDoc.String();
+
+            if (!valid) {
+                // This uri string is invalid --> parse the uri and ensure it fails
+                const InvalidURITestCase testCase = {uri};
+                mongo::unittest::log() << "Testing URI: " << testCase.URI << '\n';
+                auto cs_status = MongoURI::parse(testCase.URI);
+                ASSERT_NOT_OK(cs_status);
+            } else {
+                // This uri is valid -- > parse the remaining necessary fields
+
+                // parse the auth options
+                std::string database, username, password;
+
+                const auto auth = test.getField("auth");
+                ASSERT_FALSE(auth.eoo());
+                if (auth.type() != mongo::jstNULL) {
+                    ASSERT_EQ(auth.type(), mongo::Object);
+                    const auto authObj = auth.embeddedObject();
+                    database = returnStringFromElementOrNull(authObj.getField("db"));
+                    username = returnStringFromElementOrNull(authObj.getField("username"));
+                    password = returnStringFromElementOrNull(authObj.getField("password"));
+                }
+
+                // parse the hosts
+                const auto hosts = test.getField("hosts");
+                ASSERT_FALSE(hosts.eoo());
+                ASSERT_EQ(hosts.type(), mongo::Array);
+                const auto numHosts = static_cast<size_t>(hosts.Obj().nFields());
+
+                // parse the options
+                mongo::ConnectionString::ConnectionType connectionType = kMaster;
+                size_t numOptions = 0;
+                std::string setName;
+                const auto optionsElement = test.getField("options");
+                ASSERT_FALSE(optionsElement.eoo());
+                if (optionsElement.type() != mongo::jstNULL) {
+                    ASSERT_EQ(optionsElement.type(), mongo::Object);
+                    const auto optionsObj = optionsElement.Obj();
+                    numOptions = optionsObj.nFields();
+                    const auto replsetElement = optionsObj.getField("replicaSet");
+                    if (!replsetElement.eoo()) {
+                        ASSERT_EQ(replsetElement.type(), mongo::String);
+                        setName = replsetElement.String();
+                        connectionType = kSet;
+                    }
+                }
+
+                // Create the URITestCase abnd
+                const URITestCase testCase = {uri,
+                                              username,
+                                              password,
+                                              connectionType,
+                                              setName,
+                                              numHosts,
+                                              numOptions,
+                                              database};
+                testValidURIFormat(testCase);
+            }
+        }
+    }
+}
+
 }  // namespace
diff --git a/src/mongo/client/mongo_uri_tests/mongo-uri-host-identifiers.json b/src/mongo/client/mongo_uri_tests/mongo-uri-host-identifiers.json
new file mode 100644
index 00000000000..8a1fda580a8
--- /dev/null
+++ b/src/mongo/client/mongo_uri_tests/mongo-uri-host-identifiers.json
@@ -0,0 +1,203 @@
+{
+  "tests": [
+    {
+      "description": "Single IPv4 host without port",
+      "uri": "mongodb://127.0.0.1",
+      "valid": true,
+      "warning": false,
+      "hosts": [
+        {
+          "type": "ipv4",
+          "host": "127.0.0.1",
+          "port": null
+        }
+      ],
+      "auth": null,
+      "options": null
+    },
+    {
+      "description": "Single IPv4 host with port",
+      "uri": "mongodb://127.0.0.1:27018",
+      "valid": true,
+      "warning": false,
+      "hosts": [
+        {
+          "type": "ipv4",
+          "host": "127.0.0.1",
+          "port": 27018
+        }
+      ],
+      "auth": null,
+      "options": null
+    },
+    {
+      "description": "Single IP literal host without port",
+      "uri": "mongodb://[::1]",
+      "valid": true,
+      "warning": false,
+      "hosts": [
+        {
+          "type": "ip_literal",
+          "host": "::1",
+          "port": null
+        }
+      ],
+      "auth": null,
+      "options": null
+    },
+    {
+      "description": "Single IP literal host with port",
+      "uri": "mongodb://[::1]:27019",
+      "valid": true,
+      "warning": false,
+      "hosts": [
+        {
+          "type": "ip_literal",
+          "host": "::1",
+          "port": 27019
+        }
+      ],
+      "auth": null,
+      "options": null
+    },
+    {
+      "description": "Single hostname without port",
+      "uri": "mongodb://example.com",
+      "valid": true,
+      "warning": false,
+      "hosts": [
+        {
+          "type": "hostname",
+          "host": "example.com",
+          "port": null
+        }
+      ],
+      "auth": null,
+      "options": null
+    },
+    {
+      "description": "Single hostname with port",
+      "uri": "mongodb://example.com:27020",
+      "valid": true,
+      "warning": false,
+      "hosts": [
+        {
+          "type": "hostname",
+          "host": "example.com",
+          "port": 27020
+        }
+      ],
+      "auth": null,
+      "options": null
+    },
+    {
+      "description": "Single hostname (resembling IPv4) without port",
+      "uri": "mongodb://256.0.0.1",
+      "valid": true,
+      "warning": false,
+      "hosts": [
+        {
+          "type": "hostname",
+          "host": "256.0.0.1",
+          "port": null
+        }
+      ],
+      "auth": null,
+      "options": null
+    },
+    {
+      "description": "Multiple hosts (mixed formats)",
+      "uri": "mongodb://127.0.0.1,[::1]:27018,example.com:27019",
+      "valid": false,
+      "warning": false,
+      "hosts": [
+        {
+          "type": "ipv4",
+          "host": "127.0.0.1",
+          "port": null
+        },
+        {
+          "type": "ip_literal",
+          "host": "::1",
+          "port": 27018
+        },
+        {
+          "type": "hostname",
+          "host": "example.com",
+          "port": 27019
+        }
+      ],
+      "auth": null,
+      "options": null
+    },
+    {
+      "description": "Multiple hosts (mixed formats)",
+      "uri": "mongodb://127.0.0.1,[::1]:27018,example.com:27019/?replicaSet=replset",
+      "valid": true,
+      "warning": false,
+      "hosts": [
+        {
+          "type": "ipv4",
+          "host": "127.0.0.1",
+          "port": null
+        },
+        {
+          "type": "ip_literal",
+          "host": "::1",
+          "port": 27018
+        },
+        {
+          "type": "hostname",
+          "host": "example.com",
+          "port": 27019
+        }
+      ],
+      "auth": null,
+      "options": {
+        "replicaSet": "replset"
+      }
+    },
+    {
+      "description": "UTF-8 hosts",
+      "uri": "mongodb://bücher.example.com,umläut.example.com/",
+      "valid": false,
+      "warning": false,
+      "hosts": [
+        {
+          "type": "hostname",
+          "host": "bücher.example.com",
+          "port": null
+        },
+        {
+          "type": "hostname",
+          "host": "umläut.example.com",
+          "port": null
+        }
+      ],
+      "auth": null,
+      "options": null
+    }, 
+    {
+      "description": "UTF-8 hosts",
+      "uri": "mongodb://bücher.example.com,umläut.example.com/?replicaSet=replset",
+      "valid": true,
+      "warning": false,
+      "hosts": [
+        {
+          "type": "hostname",
+          "host": "bücher.example.com",
+          "port": null
+        },
+        {
+          "type": "hostname",
+          "host": "umläut.example.com",
+          "port": null
+        }
+      ],
+      "auth": null,
+      "options": {
+        "replicaSet": "replset"
+      }
+    }
+  ]
+}
\ No newline at end of file
diff --git a/src/mongo/client/mongo_uri_tests/mongo-uri-invalid.json b/src/mongo/client/mongo_uri_tests/mongo-uri-invalid.json
new file mode 100644
index 00000000000..3ee82bd535e
--- /dev/null
+++ b/src/mongo/client/mongo_uri_tests/mongo-uri-invalid.json
@@ -0,0 +1,247 @@
+{
+    "tests": [
+        {
+            "auth": null,
+            "description": "Empty string",
+            "hosts": null,
+            "options": null,
+            "uri": "",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Missing host",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Double colon in host identifier",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://localhost::27017",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Double colon in host identifier and trailing slash",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://localhost::27017/",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Double colon in host identifier with missing host and port",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://::",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Double colon in host identifier with missing port",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://localhost,localhost::",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Double colon in host identifier and second host",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://localhost::27017,abc",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Invalid port (negative number) with hostname",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://localhost:-1",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Invalid port (zero) with hostname",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://localhost:0/",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Invalid port (positive number) with hostname",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://localhost:65536",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Invalid port (positive number) with hostname and trailing slash",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://localhost:65536/",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Invalid port (non-numeric string) with hostname",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://localhost:foo",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Invalid port (negative number) with IP literal",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://[::1]:-1",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Invalid port (zero) with IP literal",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://[::1]:0/",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Invalid port (positive number) with IP literal",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://[::1]:65536",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Invalid port (positive number) with IP literal and trailing slash",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://[::1]:65536/",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Invalid port (non-numeric string) with IP literal",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://[::1]:foo",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Missing delimiting slash between hosts and options",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://example.com?w=1",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Incomplete key value pair for option",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://example.com/?w",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Username with password containing an unescaped colon",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://alice:foo:bar@127.0.0.1",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Username with password containing an unescaped colon",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://alice:foo:bar@127.0.0.1",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Username containing an unescaped at-sign",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://alice@@127.0.0.1",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Username with password containing an unescaped at-sign",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://alice@foo:bar@127.0.0.1",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Username containing an unescaped slash",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://alice/@localhost/db",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Username containing unescaped slash with password",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://alice/bob:foo@localhost/db",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Username with password containing an unescaped slash",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb://alice:foo/bar@localhost/db",
+            "valid": false,
+            "warning": null
+        },
+        {
+            "auth": null,
+            "description": "Host with unescaped slash",
+            "hosts": null,
+            "options": null,
+            "uri": "mongodb:///tmp/mongodb-27017.sock/",
+            "valid": false,
+            "warning": null
+        }
+    ]
+}
diff --git a/src/mongo/client/mongo_uri_tests/mongo-uri-options.json b/src/mongo/client/mongo_uri_tests/mongo-uri-options.json
new file mode 100644
index 00000000000..4c2bded9e72
--- /dev/null
+++ b/src/mongo/client/mongo_uri_tests/mongo-uri-options.json
@@ -0,0 +1,25 @@
+{
+  "tests": [
+    {
+      "description": "Option names are normalized to lowercase",
+      "uri": "mongodb://alice:secret@example.com/admin?AUTHMechanism=MONGODB-CR",
+      "valid": true,
+      "warning": false,
+      "hosts": [
+        {
+          "type": "hostname",
+          "host": "example.com",
+          "port": null
+        }
+      ],
+      "auth": {
+        "username": "alice",
+        "password": "secret",
+        "db": "admin"
+      },
+      "options": {
+        "authmechanism": "MONGODB-CR"
+      }
+    }
+  ]
+}
diff --git a/src/mongo/client/mongo_uri_tests/mongo-uri-unix-sockets-absolute.json b/src/mongo/client/mongo_uri_tests/mongo-uri-unix-sockets-absolute.json
new file mode 100644
index 00000000000..d167c315bbc
--- /dev/null
+++ b/src/mongo/client/mongo_uri_tests/mongo-uri-unix-sockets-absolute.json
@@ -0,0 +1,373 @@
+{
+    "tests": [
+        {
+            "auth": null,
+            "description": "Unix domain socket (absolute path with trailing slash)",
+            "hosts": [
+                {
+                    "host": "/tmp/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://%2Ftmp%2Fmongodb-27017.sock/",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": null,
+            "description": "Unix domain socket (absolute path without trailing slash)",
+            "hosts": [
+                {
+                    "host": "/tmp/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://%2Ftmp%2Fmongodb-27017.sock",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": null,
+            "description": "Unix domain socket (absolute path with spaces in path)",
+            "hosts": [
+                {
+                    "host": "/tmp/ /mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://%2Ftmp%2F %2Fmongodb-27017.sock",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": null,
+            "description": "Multiple Unix domain sockets (absolute paths)",
+            "hosts": [
+                {
+                    "host": "/tmp/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                },
+                {
+                    "host": "/tmp/mongodb-27018.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://%2Ftmp%2Fmongodb-27017.sock,%2Ftmp%2Fmongodb-27018.sock",
+            "valid": false,
+            "warning": false
+        },
+        {
+            "auth": null,
+            "description": "Multiple hosts (absolute path and ipv4)",
+            "hosts": [
+                {
+                    "host": "127.0.0.1",
+                    "port": 27017,
+                    "type": "ipv4"
+                },
+                {
+                    "host": "/tmp/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://127.0.0.1:27017,%2Ftmp%2Fmongodb-27017.sock",
+            "valid": false,
+            "warning": false
+        },
+        {
+            "auth": null,
+            "description": "Multiple hosts (absolute path and hostname resembling relative path)",
+            "hosts": [
+                {
+                    "host": "mongodb-27017.sock",
+                    "port": null,
+                    "type": "hostname"
+                },
+                {
+                    "host": "/tmp/mongodb-27018.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://mongodb-27017.sock,%2Ftmp%2Fmongodb-27018.sock",
+            "valid": false,
+            "warning": false
+        },
+        {
+            "auth": null,
+            "description": "Multiple Unix domain sockets (absolute paths)",
+            "hosts": [
+                {
+                    "host": "/tmp/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                },
+                {
+                    "host": "/tmp/mongodb-27018.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": {
+                "replicaSet": "replset"
+            },
+            "uri": "mongodb://%2Ftmp%2Fmongodb-27017.sock,%2Ftmp%2Fmongodb-27018.sock/?replicaSet=replset",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": null,
+            "description": "Multiple hosts (absolute path and ipv4)",
+            "hosts": [
+                {
+                    "host": "127.0.0.1",
+                    "port": 27017,
+                    "type": "ipv4"
+                },
+                {
+                    "host": "/tmp/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": {
+                "replicaSet": "replset"
+            },
+            "uri": "mongodb://127.0.0.1:27017,%2Ftmp%2Fmongodb-27017.sock/?replicaSet=replset",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": null,
+            "description": "Multiple hosts (absolute path and hostname resembling relative path)",
+            "hosts": [
+                {
+                    "host": "mongodb-27017.sock",
+                    "port": null,
+                    "type": "hostname"
+                },
+                {
+                    "host": "/tmp/mongodb-27018.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": {
+                "replicaSet": "replset"
+            },
+            "uri": "mongodb://mongodb-27017.sock,%2Ftmp%2Fmongodb-27018.sock/?replicaSet=replset",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": "admin",
+                "password": "foo",
+                "username": "alice"
+            },
+            "description": "Unix domain socket with auth database (absolute path)",
+            "hosts": [
+                {
+                    "host": "/tmp/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://alice:foo@%2Ftmp%2Fmongodb-27017.sock/admin",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": null,
+            "description": "Unix domain socket with path resembling socket file (absolute path with trailing slash)",
+            "hosts": [
+                {
+                    "host": "/tmp/path.to.sock/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://%2Ftmp%2Fpath.to.sock%2Fmongodb-27017.sock/",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": null,
+            "description": "Unix domain socket with path resembling socket file (absolute path without trailing slash)",
+            "hosts": [
+                {
+                    "host": "/tmp/path.to.sock/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://%2Ftmp%2Fpath.to.sock%2Fmongodb-27017.sock",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": "admin",
+                "password": "bar",
+                "username": "bob"
+            },
+            "description": "Unix domain socket with path resembling socket file and auth (absolute path)",
+            "hosts": [
+                {
+                    "host": "/tmp/path.to.sock/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://bob:bar@%2Ftmp%2Fpath.to.sock%2Fmongodb-27017.sock/admin",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": "admin",
+                "password": null,
+                "username": null
+            },
+            "description": "Multiple Unix domain sockets and auth DB (absolute path)",
+            "hosts": [
+                {
+                    "host": "/tmp/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                },
+                {
+                    "host": "/tmp/mongodb-27018.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://%2Ftmp%2Fmongodb-27017.sock,%2Ftmp%2Fmongodb-27018.sock/admin",
+            "valid": false,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": "admin",
+                "password": null,
+                "username": null
+            },
+            "description": "Multiple Unix domain sockets and auth DB (absolute path)",
+            "hosts": [
+                {
+                    "host": "/tmp/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                },
+                {
+                    "host": "/tmp/mongodb-27018.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": {
+                "replicaSet": "replset"
+            },
+            "uri": "mongodb://%2Ftmp%2Fmongodb-27017.sock,%2Ftmp%2Fmongodb-27018.sock/admin?replicaSet=replset",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": "admin",
+                "password": "bar",
+                "username": "bob"
+            },
+            "description": "Multiple Unix domain sockets with auth and query string (absolute path)",
+            "hosts": [
+                {
+                    "host": "/tmp/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                },
+                {
+                    "host": "/tmp/mongodb-27018.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": {
+                "w": 1
+            },
+            "uri": "mongodb://bob:bar@%2Ftmp%2Fmongodb-27017.sock,%2Ftmp%2Fmongodb-27018.sock/admin?w=1",
+            "valid": false,
+            "warning": false
+        }, 
+        {
+            "auth": {
+                "db": "admin",
+                "password": "bar",
+                "username": "bob"
+            },
+            "description": "Multiple Unix domain sockets with auth and query string (absolute path)",
+            "hosts": [
+                {
+                    "host": "/tmp/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                },
+                {
+                    "host": "/tmp/mongodb-27018.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": {
+                "w": 1, 
+                "replicaSet":"replset"
+            },
+            "uri": "mongodb://bob:bar@%2Ftmp%2Fmongodb-27017.sock,%2Ftmp%2Fmongodb-27018.sock/admin?w=1&replicaSet=replset",
+            "valid": true,
+            "warning": false
+        }, 
+        {
+            "auth": {
+                "db": "admin",
+                "password": "bar",
+                "username": "bob"
+            },
+            "description": "Multiple Unix domain sockets with auth and query string (absolute path)",
+            "hosts": [
+                {
+                    "host": "/tmp/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                },
+                {
+                    "host": "/tmp/mongodb-27018.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": {
+                "w": 1, 
+                "replicaSet":"replset"
+            },
+            "uri": "mongodb://bob:bar@%2Ftmp%2Fmongodb-27017.sock,%2Ftmp%2Fmongodb-27018.sock/admin?replicaSet=replset&w=1",
+            "valid": true,
+            "warning": false
+        }
+    ]
+}
diff --git a/src/mongo/client/mongo_uri_tests/mongo-uri-unix-sockets-relative.json b/src/mongo/client/mongo_uri_tests/mongo-uri-unix-sockets-relative.json
new file mode 100644
index 00000000000..8da45c4e1b9
--- /dev/null
+++ b/src/mongo/client/mongo_uri_tests/mongo-uri-unix-sockets-relative.json
@@ -0,0 +1,466 @@
+{
+    "tests": [
+        {
+            "auth": null,
+            "description": "Unix domain socket (relative path with trailing slash)",
+            "hosts": [
+                {
+                    "host": "rel/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://rel%2Fmongodb-27017.sock/",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": null,
+            "description": "Unix domain socket (relative path without trailing slash)",
+            "hosts": [
+                {
+                    "host": "rel/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://rel%2Fmongodb-27017.sock",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": null,
+            "description": "Unix domain socket (relative path with spaces)",
+            "hosts": [
+                {
+                    "host": "rel/ /mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://rel%2F %2Fmongodb-27017.sock",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": null,
+            "description": "Multiple Unix domain sockets (relative paths)",
+            "hosts": [
+                {
+                    "host": "rel/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                },
+                {
+                    "host": "rel/mongodb-27018.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://rel%2Fmongodb-27017.sock,rel%2Fmongodb-27018.sock",
+            "valid": false,
+            "warning": false
+        },
+        {
+            "auth": null,
+            "description": "Multiple Unix domain sockets (relative paths)",
+            "hosts": [
+                {
+                    "host": "rel/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                },
+                {
+                    "host": "rel/mongodb-27018.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": {
+                "replicaSet": "replset"
+            },
+            "uri": "mongodb://rel%2Fmongodb-27017.sock,rel%2Fmongodb-27018.sock/?replicaSet=replset",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": null,
+            "description": "Multiple Unix domain sockets (relative and absolute paths)",
+            "hosts": [
+                {
+                    "host": "rel/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                },
+                {
+                    "host": "/tmp/mongodb-27018.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://rel%2Fmongodb-27017.sock,%2Ftmp%2Fmongodb-27018.sock",
+            "valid": false,
+            "warning": false
+        },
+        {
+            "auth": null,
+            "description": "Multiple Unix domain sockets (relative and absolute paths)",
+            "hosts": [
+                {
+                    "host": "rel/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                },
+                {
+                    "host": "/tmp/mongodb-27018.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": {
+                "replicaSet": "replset"
+            },
+            "uri": "mongodb://rel%2Fmongodb-27017.sock,%2Ftmp%2Fmongodb-27018.sock/?replicaSet=replset",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": null,
+            "description": "Multiple hosts (relative path and ipv4)",
+            "hosts": [
+                {
+                    "host": "127.0.0.1",
+                    "port": 27017,
+                    "type": "ipv4"
+                },
+                {
+                    "host": "rel/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://127.0.0.1:27017,rel%2Fmongodb-27017.sock",
+            "valid": false,
+            "warning": false
+        },
+        {
+            "auth": null,
+            "description": "Multiple hosts (relative path and ipv4)",
+            "hosts": [
+                {
+                    "host": "127.0.0.1",
+                    "port": 27017,
+                    "type": "ipv4"
+                },
+                {
+                    "host": "rel/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": {
+                "replicaSet": "replset"
+            },
+            "uri": "mongodb://127.0.0.1:27017,rel%2Fmongodb-27017.sock/?replicaSet=replset",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": null,
+            "description": "Multiple hosts (relative path and hostname resembling relative path)",
+            "hosts": [
+                {
+                    "host": "mongodb-27017.sock",
+                    "port": null,
+                    "type": "hostname"
+                },
+                {
+                    "host": "rel/mongodb-27018.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://mongodb-27017.sock,rel%2Fmongodb-27018.sock",
+            "valid": false,
+            "warning": false
+        },
+        {
+            "auth": null,
+            "description": "Multiple hosts (relative path and hostname resembling relative path)",
+            "hosts": [
+                {
+                    "host": "mongodb-27017.sock",
+                    "port": null,
+                    "type": "hostname"
+                },
+                {
+                    "host": "rel/mongodb-27018.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": {
+                "replicaSet": "replset"
+            },
+            "uri": "mongodb://mongodb-27017.sock,rel%2Fmongodb-27018.sock/?replicaSet=replset",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": "admin",
+                "password": "foo",
+                "username": "alice"
+            },
+            "description": "Unix domain socket with auth database (relative path)",
+            "hosts": [
+                {
+                    "host": "rel/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://alice:foo@rel%2Fmongodb-27017.sock/admin",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": null,
+            "description": "Unix domain socket with path resembling socket file (relative path with trailing slash)",
+            "hosts": [
+                {
+                    "host": "rel/path.to.sock/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://rel%2Fpath.to.sock%2Fmongodb-27017.sock/",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": null,
+            "description": "Unix domain socket with path resembling socket file (relative path without trailing slash)",
+            "hosts": [
+                {
+                    "host": "rel/path.to.sock/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://rel%2Fpath.to.sock%2Fmongodb-27017.sock",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": "admin",
+                "password": "bar",
+                "username": "bob"
+            },
+            "description": "Unix domain socket with path resembling socket file and auth (relative path)",
+            "hosts": [
+                {
+                    "host": "rel/path.to.sock/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://bob:bar@rel%2Fpath.to.sock%2Fmongodb-27017.sock/admin",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": "admin",
+                "password": null,
+                "username": null
+            },
+            "description": "Multiple Unix domain sockets and auth DB resembling a socket (relative path)",
+            "hosts": [
+                {
+                    "host": "rel/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                },
+                {
+                    "host": "rel/mongodb-27018.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://rel%2Fmongodb-27017.sock,rel%2Fmongodb-27018.sock/admin",
+            "valid": false,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": "admin",
+                "password": null,
+                "username": null
+            },
+            "description": "Multiple Unix domain sockets and auth DB resembling a socket (relative path)",
+            "hosts": [
+                {
+                    "host": "rel/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                },
+                {
+                    "host": "rel/mongodb-27018.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": {
+                "replicaSet": "replset"
+            },
+            "uri": "mongodb://rel%2Fmongodb-27017.sock,rel%2Fmongodb-27018.sock/admin?replicaSet=replset",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": "admin",
+                "password": null,
+                "username": null
+            },
+            "description": "Multiple Unix domain sockets with auth DB resembling a path (relative path)",
+            "hosts": [
+                {
+                    "host": "rel/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                },
+                {
+                    "host": "rel/mongodb-27018.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://rel%2Fmongodb-27017.sock,rel%2Fmongodb-27018.sock/admin",
+            "valid": false,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": "admin",
+                "password": null,
+                "username": null
+            },
+            "description": "Multiple Unix domain sockets with auth DB resembling a path (relative path)",
+            "hosts": [
+                {
+                    "host": "rel/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                },
+                {
+                    "host": "rel/mongodb-27018.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": {
+                "replicaSet": "replset"
+            },
+            "uri": "mongodb://rel%2Fmongodb-27017.sock,rel%2Fmongodb-27018.sock/admin?replicaSet=replset",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": "admin",
+                "password": "bar",
+                "username": "bob"
+            },
+            "description": "Multiple Unix domain sockets with auth and query string (relative path)",
+            "hosts": [
+                {
+                    "host": "rel/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                },
+                {
+                    "host": "rel/mongodb-27018.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": {
+                "w": 1
+            },
+            "uri": "mongodb://bob:bar@rel%2Fmongodb-27017.sock,rel%2Fmongodb-27018.sock/admin?w=1",
+            "valid": false,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": "admin",
+                "password": "bar",
+                "username": "bob"
+            },
+            "description": "Multiple Unix domain sockets with auth and query string (relative path)",
+            "hosts": [
+                {
+                    "host": "rel/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                },
+                {
+                    "host": "rel/mongodb-27018.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": {
+                "w": 1,
+                "replicaSet": "replset"
+            },
+            "uri": "mongodb://bob:bar@rel%2Fmongodb-27017.sock,rel%2Fmongodb-27018.sock/admin?w=1&replicaSet=replset",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": "admin",
+                "password": "bar",
+                "username": "bob"
+            },
+            "description": "Multiple Unix domain sockets with auth and query string (relative path)",
+            "hosts": [
+                {
+                    "host": "rel/mongodb-27017.sock",
+                    "port": null,
+                    "type": "unix"
+                },
+                {
+                    "host": "rel/mongodb-27018.sock",
+                    "port": null,
+                    "type": "unix"
+                }
+            ],
+            "options": {
+                "w": 1,
+                "replicaSet": "replset", 
+                "b": 4
+            },
+            "uri": "mongodb://bob:bar@rel%2Fmongodb-27017.sock,rel%2Fmongodb-27018.sock/admin?w=1&replicaSet=replset&b=4",
+            "valid": true,
+            "warning": false
+        }
+    ]
+}
diff --git a/src/mongo/client/mongo_uri_tests/mongo-uri-valid-auth.json b/src/mongo/client/mongo_uri_tests/mongo-uri-valid-auth.json
new file mode 100644
index 00000000000..a44236e7b62
--- /dev/null
+++ b/src/mongo/client/mongo_uri_tests/mongo-uri-valid-auth.json
@@ -0,0 +1,311 @@
+{
+    "tests": [
+        {
+            "auth": {
+                "db": null,
+                "password": "foo",
+                "username": "alice"
+            },
+            "description": "User info for single IPv4 host without database",
+            "hosts": [
+                {
+                    "host": "127.0.0.1",
+                    "port": null,
+                    "type": "ipv4"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://alice:foo@127.0.0.1",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": "test",
+                "password": "foo",
+                "username": "alice"
+            },
+            "description": "User info for single IPv4 host with database",
+            "hosts": [
+                {
+                    "host": "127.0.0.1",
+                    "port": null,
+                    "type": "ipv4"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://alice:foo@127.0.0.1/test",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": null,
+                "password": "bar",
+                "username": "bob"
+            },
+            "description": "User info for single IP literal host without database",
+            "hosts": [
+                {
+                    "host": "::1",
+                    "port": 27018,
+                    "type": "ip_literal"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://bob:bar@[::1]:27018",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": "admin",
+                "password": "bar",
+                "username": "bob"
+            },
+            "description": "User info for single IP literal host with database",
+            "hosts": [
+                {
+                    "host": "::1",
+                    "port": 27018,
+                    "type": "ip_literal"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://bob:bar@[::1]:27018/admin",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": null,
+                "password": "baz",
+                "username": "eve"
+            },
+            "description": "User info for single hostname without database",
+            "hosts": [
+                {
+                    "host": "example.com",
+                    "port": null,
+                    "type": "hostname"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://eve:baz@example.com",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": "db2",
+                "password": "baz",
+                "username": "eve"
+            },
+            "description": "User info for single hostname with database",
+            "hosts": [
+                {
+                    "host": "example.com",
+                    "port": null,
+                    "type": "hostname"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://eve:baz@example.com/db2",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": null,
+                "password": "secret",
+                "username": "alice"
+            },
+            "description": "User info for multiple hosts without database",
+            "hosts": [
+                {
+                    "host": "127.0.0.1",
+                    "port": null,
+                    "type": "ipv4"
+                },
+                {
+                    "host": "example.com",
+                    "port": 27018,
+                    "type": "hostname"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://alice:secret@127.0.0.1,example.com:27018",
+            "valid": false,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": "admin",
+                "password": "secret",
+                "username": "alice"
+            },
+            "description": "User info for multiple hosts with database",
+            "hosts": [
+                {
+                    "host": "example.com",
+                    "port": null,
+                    "type": "hostname"
+                },
+                {
+                    "host": "::1",
+                    "port": 27019,
+                    "type": "ip_literal"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://alice:secret@example.com,[::1]:27019/admin",
+            "valid": false,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": null,
+                "password": null,
+                "username": "alice"
+            },
+            "description": "Username without password",
+            "hosts": [
+                {
+                    "host": "127.0.0.1",
+                    "port": null,
+                    "type": "ipv4"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://alice@127.0.0.1",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": null,
+                "password": "",
+                "username": "alice"
+            },
+            "description": "Username with empty password",
+            "hosts": [
+                {
+                    "host": "127.0.0.1",
+                    "port": null,
+                    "type": "ipv4"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://alice:@127.0.0.1",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": "my=db",
+                "password": null,
+                "username": "@l:ce/="
+            },
+            "description": "Escaped username and database without password",
+            "hosts": [
+                {
+                    "host": "example.com",
+                    "port": null,
+                    "type": "hostname"
+                }
+            ],
+            "options": null,
+            "uri": "mongodb://%40l%3Ace%2F%3D@example.com/my%3Ddb",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": "admin=",
+                "password": "fizzb@zz=",
+                "username": "$am"
+            },
+            "description": "Escaped user info and database (MONGODB-CR)",
+            "hosts": [
+                {
+                    "host": "127.0.0.1",
+                    "port": null,
+                    "type": "ipv4"
+                }
+            ],
+            "options": {
+                "authmechanism": "MONGODB-CR"
+            },
+            "uri": "mongodb://%24am:fizzb%40zz%3D@127.0.0.1/admin%3D?authMechanism=MONGODB-CR",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": null,
+                "password": null,
+                "username": "CN=myName,OU=myOrgUnit,O=myOrg,L=myLocality,ST=myState,C=myCountry"
+            },
+            "description": "Escaped username (MONGODB-X509)",
+            "hosts": [
+                {
+                    "host": "localhost",
+                    "port": null,
+                    "type": "hostname"
+                }
+            ],
+            "options": {
+                "authmechanism": "MONGODB-X509"
+            },
+            "uri": "mongodb://CN%3DmyName%2COU%3DmyOrgUnit%2CO%3DmyOrg%2CL%3DmyLocality%2CST%3DmyState%2CC%3DmyCountry@localhost/?authMechanism=MONGODB-X509",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": null,
+                "password": "secret",
+                "username": "user@EXAMPLE.COM"
+            },
+            "description": "Escaped username (GSSAPI)",
+            "hosts": [
+                {
+                    "host": "localhost",
+                    "port": null,
+                    "type": "hostname"
+                }
+            ],
+            "options": {
+                "authmechanism": "GSSAPI",
+                "authmechanismproperties": {
+                    "CANONICALIZE_HOST_NAME": true,
+                    "SERVICE_NAME": "other"
+                }
+            },
+            "uri": "mongodb://user%40EXAMPLE.COM:secret@localhost/?authMechanismProperties=SERVICE_NAME:other,CANONICALIZE_HOST_NAME:true&authMechanism=GSSAPI",
+            "valid": true,
+            "warning": false
+        },
+        {
+            "auth": {
+                "db": "admin",
+                "password": "secret",
+                "username": "alice"
+            },
+            "description": "At-signs in options aren't part of the userinfo",
+            "hosts": [
+                {
+                    "host": "example.com",
+                    "port": null,
+                    "type": "hostname"
+                }
+            ],
+            "options": {
+                "replicaSet": "my@replicaset"
+            },
+            "uri": "mongodb://alice:secret@example.com/admin?replicaSet=my@replicaset",
+            "valid": true,
+            "warning": false
+        }
+    ]
+}
diff --git a/src/mongo/client/mongo_uri_tests/mongo-uri-warnings.json b/src/mongo/client/mongo_uri_tests/mongo-uri-warnings.json
new file mode 100644
index 00000000000..a04246847ab
--- /dev/null
+++ b/src/mongo/client/mongo_uri_tests/mongo-uri-warnings.json
@@ -0,0 +1,56 @@
+{
+  "tests": [
+    {
+      "description": "Unrecognized option keys are ignored",
+      "uri": "mongodb://example.com/?foo=bar",
+      "valid": true,
+      "warning": true,
+      "hosts": [
+        {
+          "type": "hostname",
+          "host": "example.com",
+          "port": null
+        }
+      ],
+      "auth": null,
+      "options": {
+        "foo": "bar"
+      }
+    },
+    {
+      "description": "Unsupported option values are ignored",
+      "uri": "mongodb://example.com/?fsync=ifPossible",
+      "valid": true,
+      "warning": true,
+      "hosts": [
+        {
+          "type": "hostname",
+          "host": "example.com",
+          "port": null
+        }
+      ],
+      "auth": null,
+      "options": {
+        "fsync": "ifPossible"
+      }
+    },
+    {
+      "description": "Deprecated (or unknown) options are ignored if replacement exists",
+      "uri": "mongodb://example.com/?wtimeout=5&wtimeoutMS=10",
+      "valid": true,
+      "warning": true,
+      "hosts": [
+        {
+          "type": "hostname",
+          "host": "example.com",
+          "port": null
+        }
+      ],
+      "auth": null,
+      "options": {
+        "wtimeoutMS": 10, 
+        "wtimeout": 5
+      }
+    }
+  ]
+}
\ No newline at end of file
diff --git a/src/mongo/crypto/mechanism_scram.cpp b/src/mongo/crypto/mechanism_scram.cpp
index 8cd59142d67..6a2757d383d 100644
--- a/src/mongo/crypto/mechanism_scram.cpp
+++ b/src/mongo/crypto/mechanism_scram.cpp
@@ -226,6 +226,10 @@ bool verifyClientProof(StringData clientProof, StringData storedKey, StringData
     SHA1Block computedStoredKey =
         SHA1Block::computeHash(clientSignature.data(), clientSignature.size());
 
+    if (storedKey.size() != computedStoredKey.size()) {
+        return false;
+    }
+
     return consttimeMemEqual(reinterpret_cast<const unsigned char*>(storedKey.rawData()),
                              computedStoredKey.data(),
                              computedStoredKey.size());
diff --git a/src/mongo/crypto/mechanism_scram.h b/src/mongo/crypto/mechanism_scram.h
index b12cebbf2e1..6045f037078 100644
--- a/src/mongo/crypto/mechanism_scram.h
+++ b/src/mongo/crypto/mechanism_scram.h
@@ -53,15 +53,22 @@ const std::string serverKeyFieldName = "serverKey";
  */
 struct SCRAMPresecrets {
     SCRAMPresecrets(std::string hashedPassword,
-                    std::vector<std::uint8_t> salt,
-                    size_t iterationCount)
+                    std::vector<std::uint8_t> salt_,
+                    size_t iterationCount_)
         : hashedPassword(std::move(hashedPassword)),
-          salt(std::move(salt)),
-          iterationCount(iterationCount) {}
+          salt(std::move(salt_)),
+          iterationCount(iterationCount_) {
+        uassert(ErrorCodes::BadValue,
+                "Invalid salt for SCRAM mechanism",
+                salt.size() >= kSaltLengthMin);
+    }
 
     std::string hashedPassword;
     std::vector<std::uint8_t> salt;
     size_t iterationCount;
+
+private:
+    static const size_t kSaltLengthMin = 16;
 };
 
 inline bool operator==(const SCRAMPresecrets& lhs, const SCRAMPresecrets& rhs) {
diff --git a/src/mongo/db/SConscript b/src/mongo/db/SConscript
index e2be152f126..ded6480875c 100644
--- a/src/mongo/db/SConscript
+++ b/src/mongo/db/SConscript
@@ -445,6 +445,7 @@ env.Library(
     LIBDEPS=[
         '$BUILD_DIR/mongo/db/serveronly',
         '$BUILD_DIR/mongo/util/options_parser/options_parser_init',
+        'repl/replica_set_messages',
         'server_options',
     ],
 )
@@ -766,9 +767,12 @@ env.Library(
         'log_process_details.cpp',
     ],
     LIBDEPS=[
+        'repl/repl_coordinator_global',
+        'repl/repl_coordinator_interface',
+        'repl/replica_set_messages',
+        'server_options',
         '$BUILD_DIR/mongo/base',
         '$BUILD_DIR/mongo/util/net/network',
-        'server_options',
     ],
 )
 
@@ -818,3 +822,4 @@ asioEnv.CppIntegrationTest(
         '$BUILD_DIR/mongo/executor/network_interface_asio_fixture',
     ],
 )
+
diff --git a/src/mongo/db/auth/SConscript b/src/mongo/db/auth/SConscript
index 792314035b5..fc85149de7f 100644
--- a/src/mongo/db/auth/SConscript
+++ b/src/mongo/db/auth/SConscript
@@ -138,6 +138,7 @@ env.Library('authmongod',
             ],
             LIBDEPS=[
                 'authservercommon',
+                '$BUILD_DIR/mongo/db/catalog/index_key_validate',
                 '$BUILD_DIR/mongo/db/db_raii',
                 '$BUILD_DIR/mongo/db/dbdirectclient',
                 '$BUILD_DIR/mongo/db/dbhelpers',
diff --git a/src/mongo/db/auth/auth_index_d.cpp b/src/mongo/db/auth/auth_index_d.cpp
index 427900c8377..47987ac40c4 100644
--- a/src/mongo/db/auth/auth_index_d.cpp
+++ b/src/mongo/db/auth/auth_index_d.cpp
@@ -34,15 +34,20 @@
 
 #include "mongo/base/init.h"
 #include "mongo/base/status.h"
+#include "mongo/client/index_spec.h"
 #include "mongo/db/auth/authorization_manager.h"
 #include "mongo/db/auth/authorization_manager_global.h"
 #include "mongo/db/catalog/collection.h"
 #include "mongo/db/catalog/index_catalog.h"
+#include "mongo/db/catalog/index_create.h"
+#include "mongo/db/catalog/index_key_validate.h"
 #include "mongo/db/client.h"
-#include "mongo/db/client.h"
+#include "mongo/db/concurrency/write_conflict_exception.h"
+#include "mongo/db/curop.h"
 #include "mongo/db/db_raii.h"
 #include "mongo/db/index/index_descriptor.h"
 #include "mongo/db/jsobj.h"
+#include "mongo/db/storage/storage_options.h"
 #include "mongo/util/assert_util.h"
 #include "mongo/util/log.h"
 
@@ -58,6 +63,8 @@ BSONObj v3SystemUsersKeyPattern;
 BSONObj v3SystemRolesKeyPattern;
 std::string v3SystemUsersIndexName;
 std::string v3SystemRolesIndexName;
+IndexSpec v3SystemUsersIndexSpec;
+IndexSpec v3SystemRolesIndexSpec;
 
 MONGO_INITIALIZER(AuthIndexKeyPatterns)(InitializerContext*) {
     v1SystemUsersKeyPattern = BSON("user" << 1 << "userSource" << 1);
@@ -76,13 +83,66 @@ MONGO_INITIALIZER(AuthIndexKeyPatterns)(InitializerContext*) {
                                   << AuthorizationManager::ROLE_DB_FIELD_NAME
                                   << "_1");
 
+    v3SystemUsersIndexSpec.addKeys(v3SystemUsersKeyPattern);
+    v3SystemUsersIndexSpec.unique();
+    v3SystemUsersIndexSpec.name(v3SystemUsersIndexName);
+
+    v3SystemRolesIndexSpec.addKeys(v3SystemRolesKeyPattern);
+    v3SystemRolesIndexSpec.unique();
+    v3SystemRolesIndexSpec.name(v3SystemRolesIndexName);
+
     return Status::OK();
 }
 
+void generateSystemIndexForExistingCollection(OperationContext* opCtx,
+                                              Collection* collection,
+                                              const NamespaceString& ns,
+                                              const IndexSpec& spec) {
+    // Do not try and generate any system indexes in read only mode.
+    if (storageGlobalParams.readOnly) {
+        warning() << "Running in queryable backup mode. Unable to create authorization index on "
+                  << ns;
+        return;
+    }
+
+    try {
+        auto indexSpecStatus = index_key_validate::validateIndexSpec(
+            spec.toBSON(), ns, serverGlobalParams.featureCompatibility);
+        BSONObj indexSpec = fassertStatusOK(40452, indexSpecStatus);
+
+        log() << "No authorization index detected on " << ns
+              << " collection. Attempting to recover by creating an index with spec: " << indexSpec;
+
+        MultiIndexBlock indexer(opCtx, collection);
+
+        MONGO_WRITE_CONFLICT_RETRY_LOOP_BEGIN {
+            fassertStatusOK(40453, indexer.init(indexSpec));
+        }
+        MONGO_WRITE_CONFLICT_RETRY_LOOP_END(opCtx, "authorization index regeneration", ns.ns());
+
+        fassertStatusOK(40454, indexer.insertAllDocumentsInCollection());
+
+        MONGO_WRITE_CONFLICT_RETRY_LOOP_BEGIN {
+            WriteUnitOfWork wunit(opCtx);
+
+            indexer.commit();
+
+            wunit.commit();
+        }
+        MONGO_WRITE_CONFLICT_RETRY_LOOP_END(opCtx, "authorization index regeneration", ns.ns());
+
+        log() << "Authorization index construction on " << ns << " is complete";
+    } catch (const DBException& e) {
+        severe() << "Failed to regenerate index for " << ns << ". Exception: " << e.what();
+        throw;
+    }
+}
+
 }  // namespace
 
 Status verifySystemIndexes(OperationContext* txn) {
-    const NamespaceString systemUsers = AuthorizationManager::usersCollectionNamespace;
+    const NamespaceString& systemUsers = AuthorizationManager::usersCollectionNamespace;
+    const NamespaceString& systemRoles = AuthorizationManager::rolesCollectionNamespace;
 
     // Make sure the old unique index from v2.4 on system.users doesn't exist.
     ScopedTransaction scopedXact(txn, MODE_IX);
@@ -91,23 +151,54 @@ Status verifySystemIndexes(OperationContext* txn) {
         return Status::OK();
     }
 
-    Collection* collection = autoDb.getDb()->getCollection(NamespaceString(systemUsers));
-    if (!collection) {
-        return Status::OK();
+    Collection* collection = autoDb.getDb()->getCollection(systemUsers);
+    if (collection) {
+        IndexCatalog* indexCatalog = collection->getIndexCatalog();
+        invariant(indexCatalog);
+
+        // Make sure the old unique index from v2.4 on system.users doesn't exist.
+        std::vector<IndexDescriptor*> indexes;
+        indexCatalog->findIndexesByKeyPattern(txn, v1SystemUsersKeyPattern, false, &indexes);
+
+        if (!indexes.empty()) {
+            fassert(ErrorCodes::AmbiguousIndexKeyPattern, indexes.size() == 1);
+            return Status(ErrorCodes::AuthSchemaIncompatible,
+                          "Old 2.4 style user index identified. "
+                          "The authentication schema needs to be updated by "
+                          "running authSchemaUpgrade on a 2.6 server.");
+        }
+
+        // Ensure that system indexes exist for the user collection
+        indexCatalog->findIndexesByKeyPattern(txn, v3SystemUsersKeyPattern, false, &indexes);
+        if (indexes.empty()) {
+            try {
+                generateSystemIndexForExistingCollection(
+                    txn, collection, systemUsers, v3SystemUsersIndexSpec);
+            } catch (...) {
+                return exceptionToStatus();
+            }
+        }
     }
 
-    IndexCatalog* indexCatalog = collection->getIndexCatalog();
-    std::vector<IndexDescriptor*> indexes;
-    indexCatalog->findIndexesByKeyPattern(txn, v1SystemUsersKeyPattern, false, &indexes);
-
-    if (indexCatalog && !indexes.empty()) {
-        fassert(ErrorCodes::AmbiguousIndexKeyPattern, indexes.size() == 1);
-        return Status(ErrorCodes::AuthSchemaIncompatible,
-                      "Old 2.4 style user index identified. "
-                      "The authentication schema needs to be updated by "
-                      "running authSchemaUpgrade on a 2.6 server.");
+    // Ensure that system indexes exist for the roles collection, if it exists.
+    collection = autoDb.getDb()->getCollection(systemRoles);
+    if (collection) {
+        IndexCatalog* indexCatalog = collection->getIndexCatalog();
+        invariant(indexCatalog);
+
+        std::vector<IndexDescriptor*> indexes;
+        indexCatalog->findIndexesByKeyPattern(txn, v3SystemRolesKeyPattern, false, &indexes);
+        if (indexes.empty()) {
+            try {
+                generateSystemIndexForExistingCollection(
+                    txn, collection, systemRoles, v3SystemRolesIndexSpec);
+            } catch (...) {
+                return exceptionToStatus();
+            }
+        }
     }
 
+
     return Status::OK();
 }
 
@@ -115,19 +206,21 @@ void createSystemIndexes(OperationContext* txn, Collection* collection) {
     invariant(collection);
     const NamespaceString& ns = collection->ns();
     if (ns == AuthorizationManager::usersCollectionNamespace) {
-        collection->getIndexCatalog()->createIndexOnEmptyCollection(
-            txn,
-            BSON("name" << v3SystemUsersIndexName << "ns" << collection->ns().ns() << "key"
-                        << v3SystemUsersKeyPattern
-                        << "unique"
-                        << true));
+        auto indexSpec = fassertStatusOK(
+            40455,
+            index_key_validate::validateIndexSpec(
+                v3SystemUsersIndexSpec.toBSON(), ns, serverGlobalParams.featureCompatibility));
+
+        fassertStatusOK(
+            40456, collection->getIndexCatalog()->createIndexOnEmptyCollection(txn, indexSpec));
     } else if (ns == AuthorizationManager::rolesCollectionNamespace) {
-        collection->getIndexCatalog()->createIndexOnEmptyCollection(
-            txn,
-            BSON("name" << v3SystemRolesIndexName << "ns" << collection->ns().ns() << "key"
-                        << v3SystemRolesKeyPattern
-                        << "unique"
-                        << true));
+        auto indexSpec = fassertStatusOK(
+            40457,
+            index_key_validate::validateIndexSpec(
+                v3SystemRolesIndexSpec.toBSON(), ns, serverGlobalParams.featureCompatibility));
+
+        fassertStatusOK(
+            40458, collection->getIndexCatalog()->createIndexOnEmptyCollection(txn, indexSpec));
     }
 }
 
diff --git a/src/mongo/db/auth/sasl_scramsha1_server_conversation.cpp b/src/mongo/db/auth/sasl_scramsha1_server_conversation.cpp
index f48ab9b8953..900d0b818fc 100644
--- a/src/mongo/db/auth/sasl_scramsha1_server_conversation.cpp
+++ b/src/mongo/db/auth/sasl_scramsha1_server_conversation.cpp
@@ -198,6 +198,12 @@ StatusWith<bool> SaslSCRAMSHA1ServerConversation::_firstStep(std::vector<string>
         _creds.scram.serverKey = scramCreds[scram::serverKeyFieldName].String();
     }
 
+    if (!_creds.scram.isValid()) {
+        return Status(ErrorCodes::AuthenticationFailed,
+                      "Unable to perform SCRAM-SHA-1 authentication for a user with missing "
+                      "or invalid SCRAM credentials");
+    }
+
     // Generate server-first-message
     // Create text-based nonce as base64 encoding of a binary blob of length multiple of 3
     const int nonceLenQWords = 3;
@@ -280,6 +286,7 @@ StatusWith<bool> SaslSCRAMSHA1ServerConversation::_secondStep(const std::vector<
     // ClientSignature := HMAC(StoredKey, AuthMessage)
     // ClientKey := ClientSignature XOR ClientProof
     // ServerSignature := HMAC(ServerKey, AuthMessage)
+    invariant(_creds.scram.isValid());
 
     if (!scram::verifyClientProof(
             base64::decode(clientProof), base64::decode(_creds.scram.storedKey), _authMessage)) {
diff --git a/src/mongo/db/auth/user.h b/src/mongo/db/auth/user.h
index d4aea7e442b..cdef6e52ddd 100644
--- a/src/mongo/db/auth/user.h
+++ b/src/mongo/db/auth/user.h
@@ -66,6 +66,17 @@ public:
         std::string salt;
         std::string serverKey;
         std::string storedKey;
+
+        bool isValid() const {
+            // 160bit -> 20octets -> * 4/3 -> 26.667 -> padded to 28
+            const size_t kEncodedSHA1Length = 28;
+            // 128bit -> 16octets -> * 4/3 -> 21.333 -> padded to 24
+            const size_t kEncodedSaltLength = 24;
+
+            return (salt.size() == kEncodedSaltLength) &&
+                (serverKey.size() == kEncodedSHA1Length) &&
+                (storedKey.size() == kEncodedSHA1Length);
+        }
     };
     struct CredentialData {
         CredentialData() : password(""), scram(), isExternal(false) {}
diff --git a/src/mongo/db/auth/user_document_parser_test.cpp b/src/mongo/db/auth/user_document_parser_test.cpp
index 273eaff86f5..9ab688750d8 100644
--- a/src/mongo/db/auth/user_document_parser_test.cpp
+++ b/src/mongo/db/auth/user_document_parser_test.cpp
@@ -434,6 +434,7 @@ TEST_F(V2UserDocumentParsing, V2CredentialExtraction) {
                                                                       << BSON("MONGODB-CR"
                                                                               << "a"))));
     ASSERT(user->getCredentials().password == "a");
+    ASSERT(!user->getCredentials().scram.isValid());
     ASSERT(!user->getCredentials().isExternal);
 
     // Credentials are {external:true if users's db is $external
@@ -446,6 +447,7 @@ TEST_F(V2UserDocumentParsing, V2CredentialExtraction) {
                                                                 << "credentials"
                                                                 << BSON("external" << true))));
     ASSERT(user->getCredentials().password.empty());
+    ASSERT(!user->getCredentials().scram.isValid());
     ASSERT(user->getCredentials().isExternal);
 }
 
diff --git a/src/mongo/db/bson/dotted_path_support.cpp b/src/mongo/db/bson/dotted_path_support.cpp
index 1a9fda88501..31a92a5c54c 100644
--- a/src/mongo/db/bson/dotted_path_support.cpp
+++ b/src/mongo/db/bson/dotted_path_support.cpp
@@ -176,7 +176,7 @@ void extractAllElementsAlongPath(const BSONObj& obj,
 
 void extractAllElementsAlongPath(const BSONObj& obj,
                                  StringData path,
-                                 BSONElementMSet& elements,
+                                 BSONElementMultiSet& elements,
                                  bool expandArrayOnTrailingField,
                                  std::set<size_t>* arrayComponents) {
     const size_t initialDepth = 0;
diff --git a/src/mongo/db/bson/dotted_path_support.h b/src/mongo/db/bson/dotted_path_support.h
index c9c6676c75a..0a3d072c56a 100644
--- a/src/mongo/db/bson/dotted_path_support.h
+++ b/src/mongo/db/bson/dotted_path_support.h
@@ -31,6 +31,7 @@
 #include <cstddef>
 #include <set>
 
+#include "mongo/bson/bsonelement_comparator_interface.h"
 #include "mongo/bson/bsonobj.h"
 
 namespace mongo {
@@ -104,7 +105,7 @@ void extractAllElementsAlongPath(const BSONObj& obj,
 
 void extractAllElementsAlongPath(const BSONObj& obj,
                                  StringData path,
-                                 BSONElementMSet& elements,
+                                 BSONElementMultiSet& elements,
                                  bool expandArrayOnTrailingField = true,
                                  std::set<std::size_t>* arrayComponents = nullptr);
 
diff --git a/src/mongo/db/catalog/apply_ops.cpp b/src/mongo/db/catalog/apply_ops.cpp
index 672a37a10c7..ec5c7681f0e 100644
--- a/src/mongo/db/catalog/apply_ops.cpp
+++ b/src/mongo/db/catalog/apply_ops.cpp
@@ -32,6 +32,7 @@
 
 #include "mongo/db/catalog/apply_ops.h"
 
+#include "mongo/bson/util/bson_extract.h"
 #include "mongo/db/catalog/collection.h"
 #include "mongo/db/catalog/database.h"
 #include "mongo/db/catalog/database_holder.h"
@@ -42,21 +43,31 @@
 #include "mongo/db/curop.h"
 #include "mongo/db/db_raii.h"
 #include "mongo/db/dbdirectclient.h"
+#include "mongo/db/index/index_descriptor.h"
 #include "mongo/db/matcher/extensions_callback_disallow_extensions.h"
 #include "mongo/db/matcher/matcher.h"
 #include "mongo/db/op_observer.h"
 #include "mongo/db/operation_context.h"
+#include "mongo/db/query/collation/collation_spec.h"
 #include "mongo/db/repl/oplog.h"
 #include "mongo/db/repl/replication_coordinator_global.h"
 #include "mongo/db/service_context.h"
+#include "mongo/rpc/get_status_from_command_result.h"
+#include "mongo/util/fail_point_service.h"
 #include "mongo/util/log.h"
 
 namespace mongo {
 namespace {
+
+const auto kPreconditionFieldName = "preCondition"_sd;
+
+// If enabled, causes loop in _applyOps() to hang after applying current operation.
+MONGO_FP_DECLARE(applyOpsPauseBetweenOperations);
+
 /**
  * Return true iff the applyOpsCmd can be executed in a single WriteUnitOfWork.
  */
-bool canBeAtomic(const BSONObj& applyOpCmd) {
+bool _areOpsCrudOnly(const BSONObj& applyOpCmd) {
     for (const auto& elem : applyOpCmd.firstElement().Obj()) {
         const char* names[] = {"ns", "op"};
         BSONElement fields[2];
@@ -89,14 +100,11 @@ bool canBeAtomic(const BSONObj& applyOpCmd) {
     return true;
 }
 
-Status _applyOps(OperationContext* txn,
+Status _applyOps(OperationContext* opCtx,
                  const std::string& dbName,
                  const BSONObj& applyOpCmd,
                  BSONObjBuilder* result,
                  int* numApplied) {
-    dassert(txn->lockState()->isLockHeldForMode(
-        ResourceId(RESOURCE_GLOBAL, ResourceId::SINGLETON_GLOBAL), MODE_X));
-
     BSONObj ops = applyOpCmd.firstElement().Obj();
 
     // apply
@@ -107,148 +115,169 @@ Status _applyOps(OperationContext* txn,
     BSONArrayBuilder ab;
     const bool alwaysUpsert =
         applyOpCmd.hasField("alwaysUpsert") ? applyOpCmd["alwaysUpsert"].trueValue() : true;
-    const bool haveWrappingWUOW = txn->lockState()->inAWriteUnitOfWork();
-
-    {
-        repl::UnreplicatedWritesBlock uwb(txn);
+    const bool haveWrappingWUOW = opCtx->lockState()->inAWriteUnitOfWork();
 
-        while (i.more()) {
-            BSONElement e = i.next();
-            const BSONObj& opObj = e.Obj();
+    while (i.more()) {
+        BSONElement e = i.next();
+        const BSONObj& opObj = e.Obj();
 
-            // Ignore 'n' operations.
-            const char* opType = opObj["op"].valuestrsafe();
-            if (*opType == 'n')
-                continue;
+        // Ignore 'n' operations.
+        const char* opType = opObj["op"].valuestrsafe();
+        if (*opType == 'n')
+            continue;
 
-            const std::string ns = opObj["ns"].String();
+        const NamespaceString nss(opObj["ns"].String());
 
-            // Need to check this here, or OldClientContext may fail an invariant.
-            if (*opType != 'c' && !NamespaceString(ns).isValid())
-                return {ErrorCodes::InvalidNamespace, "invalid ns: " + ns};
+        // Need to check this here, or OldClientContext may fail an invariant.
+        if (*opType != 'c' && !nss.isValid())
+            return {ErrorCodes::InvalidNamespace, "invalid ns: " + nss.ns()};
 
-            Status status(ErrorCodes::InternalError, "");
+        Status status(ErrorCodes::InternalError, "");
 
-            if (haveWrappingWUOW) {
-                invariant(*opType != 'c');
+        if (haveWrappingWUOW) {
+            invariant(opCtx->lockState()->isW());
+            invariant(*opType != 'c');
+            auto db = dbHolder().get(opCtx, nss.ns());
+            if (!db) {
+                throw DBException(
+                    "cannot create a database in atomic applyOps mode; will retry without "
+                    "atomicity",
+                    ErrorCodes::NamespaceNotFound);
+            }
 
-                if (!dbHolder().get(txn, ns)) {
-                    throw DBException(
-                        "cannot create a database in atomic applyOps mode; will retry without "
-                        "atomicity",
-                        ErrorCodes::NamespaceNotFound);
-                }
+            // When processing an update on a non-existent collection, applyOperation_inlock()
+            // returns UpdateOperationFailed on updates and allows the collection to be
+            // implicitly created on upserts. We detect both cases here and fail early with
+            // NamespaceNotFound.
+            auto collection = db->getCollection(nss);
+            if (!collection && !nss.isSystemDotIndexes() && (*opType == 'i' || *opType == 'u')) {
+                throw DBException(str::stream() << "cannot apply insert or update operation on "
+                                                   "a non-existent namespace "
+                                                << nss.ns()
+                                                << ": "
+                                                << redact(opObj),
+                                  ErrorCodes::NamespaceNotFound);
+            }
 
-                OldClientContext ctx(txn, ns);
-                status = repl::applyOperation_inlock(txn, ctx.db(), opObj, alwaysUpsert);
-                if (!status.isOK())
-                    return status;
-                logOpForDbHash(txn, ns.c_str());
-            } else {
-                try {
-                    // Run operations under a nested lock as a hack to prevent yielding.
-                    //
-                    // The list of operations is supposed to be applied atomically; yielding
-                    // would break atomicity by allowing an interruption or a shutdown to occur
-                    // after only some operations are applied.  We are already locked globally
-                    // at this point, so taking a DBLock on the namespace creates a nested lock,
-                    // and yields are disallowed for operations that hold a nested lock.
-                    //
-                    // We do not have a wrapping WriteUnitOfWork so it is possible for a journal
-                    // commit to happen with a subset of ops applied.
-                    Lock::GlobalWrite globalWriteLockDisallowTempRelease(txn->lockState());
-
-                    // Ensures that yielding will not happen (see the comment above).
-                    DEV {
-                        Locker::LockSnapshot lockSnapshot;
-                        invariant(!txn->lockState()->saveLockStateAndUnlock(&lockSnapshot));
-                    };
-
-                    MONGO_WRITE_CONFLICT_RETRY_LOOP_BEGIN {
-                        if (*opType == 'c') {
-                            status = repl::applyCommand_inlock(txn, opObj, true);
+            OldClientContext ctx(opCtx, nss.ns());
+            status = repl::applyOperation_inlock(opCtx, ctx.db(), opObj, alwaysUpsert);
+            if (!status.isOK())
+                return status;
+            logOpForDbHash(opCtx, nss.ns().c_str());
+        } else {
+            try {
+                MONGO_WRITE_CONFLICT_RETRY_LOOP_BEGIN {
+                    if (*opType == 'c') {
+                        invariant(opCtx->lockState()->isW());
+                        uassertStatusOK(status = repl::applyCommand_inlock(opCtx, opObj, true));
+                    } else {
+                        const char* names[] = {"o", "ns"};
+                        BSONElement fields[2];
+                        opObj.getFields(2, names, fields);
+                        BSONElement& fieldO = fields[0];
+                        BSONElement& fieldNs = fields[1];
+                        const StringData ns = fieldNs.valueStringData();
+                        NamespaceString requestNss{ns};
+
+                        if (nss.isSystemDotIndexes()) {
+                            BSONObj indexSpec;
+                            NamespaceString indexNss;
+                            std::tie(indexSpec, indexNss) =
+                                repl::prepForApplyOpsIndexInsert(fieldO, opObj, requestNss);
+                            if (!indexSpec["collation"]) {
+                                // If the index spec does not include a collation, explicitly
+                                // specify the simple collation, so the index does not inherit the
+                                // collection default collation.
+                                auto indexVersion = indexSpec["v"];
+                                // The index version is populated by prepForApplyOpsIndexInsert().
+                                invariant(indexVersion);
+                                if (indexVersion.isNumber() &&
+                                    (indexVersion.numberInt() >=
+                                     static_cast<int>(IndexDescriptor::IndexVersion::kV2))) {
+                                    BSONObjBuilder bob;
+                                    bob.append("collation", CollationSpec::kSimpleSpec);
+                                    bob.appendElements(indexSpec);
+                                    indexSpec = bob.obj();
+                                }
+                            }
+                            BSONObjBuilder command;
+                            command.append("createIndexes", indexNss.coll());
+                            {
+                                BSONArrayBuilder indexes(command.subarrayStart("indexes"));
+                                indexes.append(indexSpec);
+                                indexes.doneFast();
+                            }
+                            const BSONObj commandObj = command.done();
+
+                            DBDirectClient client(opCtx);
+                            BSONObj infoObj;
+                            client.runCommand(nsToDatabase(ns), commandObj, infoObj);
+                            status = getStatusFromCommandResult(infoObj);
                         } else {
-                            OldClientContext ctx(txn, ns);
-
-                            status =
-                                repl::applyOperation_inlock(txn, ctx.db(), opObj, alwaysUpsert);
+                            AutoGetCollection autoColl(opCtx, nss, MODE_IX);
+                            if (!autoColl.getCollection() && !nss.isSystemDotIndexes()) {
+                                // For idempotency reasons, return success on delete operations.
+                                if (*opType == 'd') {
+                                    status = Status::OK();
+                                } else {
+                                    throw DBException(
+                                        str::stream()
+                                            << "cannot apply insert or update operation on"
+                                               " a non-existent namespace "
+                                            << nss.ns()
+                                            << ": "
+                                            << mongo::redact(opObj),
+                                        ErrorCodes::NamespaceNotFound);
+                                }
+                            } else {
+                                OldClientContext ctx(opCtx, nss.ns());
+                                status = repl::applyOperation_inlock(
+                                    opCtx, ctx.db(), opObj, alwaysUpsert);
+                            }
                         }
                     }
-                    MONGO_WRITE_CONFLICT_RETRY_LOOP_END(txn, "applyOps", ns);
-                } catch (const DBException& ex) {
-                    ab.append(false);
-                    result->append("applied", ++(*numApplied));
-                    result->append("code", ex.getCode());
-                    result->append("codeName",
-                                   ErrorCodes::errorString(ErrorCodes::fromInt(ex.getCode())));
-                    result->append("errmsg", ex.what());
-                    result->append("results", ab.arr());
-                    return Status(ErrorCodes::UnknownError, ex.what());
                 }
-                WriteUnitOfWork wuow(txn);
-                logOpForDbHash(txn, ns.c_str());
-                wuow.commit();
+                MONGO_WRITE_CONFLICT_RETRY_LOOP_END(opCtx, "applyOps", nss.ns());
+            } catch (const DBException& ex) {
+                ab.append(false);
+                result->append("applied", ++(*numApplied));
+                result->append("code", ex.getCode());
+                result->append("codeName",
+                               ErrorCodes::errorString(ErrorCodes::fromInt(ex.getCode())));
+                result->append("errmsg", ex.what());
+                result->append("results", ab.arr());
+                return ex.toStatus();
             }
-
-            ab.append(status.isOK());
-            if (!status.isOK()) {
-                log() << "applyOps error applying: " << status;
-                errors++;
-            }
-
-            (*numApplied)++;
+            WriteUnitOfWork wuow(opCtx);
+            logOpForDbHash(opCtx, nss.ns().c_str());
+            wuow.commit();
         }
 
-        result->append("applied", *numApplied);
-        result->append("results", ab.arr());
-    }  // set replicatedWrites back to original value
-
-    if (txn->writesAreReplicated()) {
-        // We want this applied atomically on slaves
-        // so we re-wrap without the pre-condition for speed
-
-        std::string tempNS = str::stream() << dbName << ".$cmd";
-
-        // TODO: possibly use mutable BSON to remove preCondition field
-        // once it is available
-        BSONObjBuilder cmdBuilder;
-
-        for (auto elem : applyOpCmd) {
-            auto name = elem.fieldNameStringData();
-            if (name == "preCondition")
-                continue;
-            if (name == "bypassDocumentValidation")
-                continue;
-            cmdBuilder.append(elem);
+        ab.append(status.isOK());
+        if (!status.isOK()) {
+            log() << "applyOps error applying: " << status;
+            errors++;
         }
 
-        const BSONObj cmdRewritten = cmdBuilder.done();
-
-        auto opObserver = getGlobalServiceContext()->getOpObserver();
-        invariant(opObserver);
-        if (haveWrappingWUOW) {
-            opObserver->onApplyOps(txn, tempNS, cmdRewritten);
-        } else {
-            // When executing applyOps outside of a wrapping WriteUnitOfWOrk, always logOp the
-            // command regardless of whether the individial ops succeeded and rely on any
-            // failures to also on secondaries. This isn't perfect, but it's what the command
-            // has always done and is part of its "correct" behavior.
-            while (true) {
-                try {
-                    WriteUnitOfWork wunit(txn);
-                    opObserver->onApplyOps(txn, tempNS, cmdRewritten);
-
-                    wunit.commit();
-                    break;
-                } catch (const WriteConflictException& wce) {
-                    LOG(2) << "WriteConflictException while logging applyOps command, retrying.";
-                    txn->recoveryUnit()->abandonSnapshot();
-                    continue;
-                }
+        (*numApplied)++;
+
+        if (MONGO_FAIL_POINT(applyOpsPauseBetweenOperations)) {
+            // While holding a database lock under MMAPv1, we would be implicitly holding the
+            // flush lock here. This would prevent other threads from acquiring the global
+            // lock or any database locks. We release all locks temporarily while the fail
+            // point is enabled to allow other threads to make progress.
+            boost::optional<Lock::TempRelease> release;
+            auto storageEngine = opCtx->getServiceContext()->getGlobalStorageEngine();
+            if (storageEngine->isMmapV1() && !opCtx->lockState()->isW()) {
+                release.emplace(opCtx->lockState());
             }
+            MONGO_FAIL_POINT_PAUSE_WHILE_SET(applyOpsPauseBetweenOperations);
         }
     }
 
+    result->append("applied", *numApplied);
+    result->append("results", ab.arr());
+
     if (errors != 0) {
         return Status(ErrorCodes::UnknownError, "applyOps had one or more errors applying ops");
     }
@@ -256,94 +285,144 @@ Status _applyOps(OperationContext* txn,
     return Status::OK();
 }
 
-Status preconditionOK(OperationContext* txn, const BSONObj& applyOpCmd, BSONObjBuilder* result) {
-    dassert(txn->lockState()->isLockHeldForMode(
-        ResourceId(RESOURCE_GLOBAL, ResourceId::SINGLETON_GLOBAL), MODE_X));
-
-    if (applyOpCmd["preCondition"].type() == Array) {
-        BSONObjIterator i(applyOpCmd["preCondition"].Obj());
-        while (i.more()) {
-            BSONObj preCondition = i.next().Obj();
-            if (preCondition["ns"].type() != BSONType::String) {
-                return {ErrorCodes::InvalidNamespace,
-                        str::stream() << "ns in preCondition must be a string, but found type: "
-                                      << typeName(preCondition["ns"].type())};
-            }
-            const NamespaceString nss(preCondition["ns"].valueStringData());
-            if (!nss.isValid()) {
-                return {ErrorCodes::InvalidNamespace, "invalid ns: " + nss.ns()};
-            }
+bool _hasPrecondition(const BSONObj& applyOpCmd) {
+    return applyOpCmd[kPreconditionFieldName].type() == Array;
+}
 
-            DBDirectClient db(txn);
-            BSONObj realres = db.findOne(nss.ns(), preCondition["q"].Obj());
+Status _checkPrecondition(OperationContext* opCtx,
+                          const BSONObj& applyOpCmd,
+                          BSONObjBuilder* result) {
+    invariant(opCtx->lockState()->isW());
+    invariant(_hasPrecondition(applyOpCmd));
+
+    for (auto elem : applyOpCmd[kPreconditionFieldName].Obj()) {
+        auto preCondition = elem.Obj();
+        if (preCondition["ns"].type() != BSONType::String) {
+            return {ErrorCodes::InvalidNamespace,
+                    str::stream() << "ns in preCondition must be a string, but found type: "
+                                  << typeName(preCondition["ns"].type())};
+        }
+        const NamespaceString nss(preCondition["ns"].valueStringData());
+        if (!nss.isValid()) {
+            return {ErrorCodes::InvalidNamespace, "invalid ns: " + nss.ns()};
+        }
 
-            // Get collection default collation.
-            Database* database = dbHolder().get(txn, nss.db());
-            if (!database) {
-                return {ErrorCodes::NamespaceNotFound,
-                        "database in ns does not exist: " + nss.ns()};
-            }
-            Collection* collection = database->getCollection(nss.ns());
-            if (!collection) {
-                return {ErrorCodes::NamespaceNotFound,
-                        "collection in ns does not exist: " + nss.ns()};
-            }
-            const CollatorInterface* collator = collection->getDefaultCollator();
-
-            // Apply-ops would never have a $where/$text matcher. Using the "DisallowExtensions"
-            // callback ensures that parsing will throw an error if $where or $text are found.
-            Matcher matcher(
-                preCondition["res"].Obj(), ExtensionsCallbackDisallowExtensions(), collator);
-            if (!matcher.matches(realres)) {
-                result->append("got", realres);
-                result->append("whatFailed", preCondition);
-                return {ErrorCodes::BadValue, "preCondition failed"};
-            }
+        DBDirectClient db(opCtx);
+        BSONObj realres = db.findOne(nss.ns(), preCondition["q"].Obj());
+
+        // Get collection default collation.
+        Database* database = dbHolder().get(opCtx, nss.db());
+        if (!database) {
+            return {ErrorCodes::NamespaceNotFound, "database in ns does not exist: " + nss.ns()};
+        }
+        Collection* collection = database->getCollection(nss.ns());
+        if (!collection) {
+            return {ErrorCodes::NamespaceNotFound, "collection in ns does not exist: " + nss.ns()};
+        }
+        const CollatorInterface* collator = collection->getDefaultCollator();
+
+        // Apply-ops would never have a $where/$text matcher. Using the "DisallowExtensions"
+        // callback ensures that parsing will throw an error if $where or $text are found.
+        Matcher matcher(
+            preCondition["res"].Obj(), ExtensionsCallbackDisallowExtensions(), collator);
+        if (!matcher.matches(realres)) {
+            result->append("got", realres);
+            result->append("whatFailed", preCondition);
+            return {ErrorCodes::BadValue, "preCondition failed"};
         }
     }
+
     return Status::OK();
 }
 }  // namespace
 
-Status applyOps(OperationContext* txn,
+Status applyOps(OperationContext* opCtx,
                 const std::string& dbName,
                 const BSONObj& applyOpCmd,
                 BSONObjBuilder* result) {
-    ScopedTransaction scopedXact(txn, MODE_X);
-    Lock::GlobalWrite globalWriteLock(txn->lockState());
+    bool allowAtomic = false;
+    uassertStatusOK(
+        bsonExtractBooleanFieldWithDefault(applyOpCmd, "allowAtomic", true, &allowAtomic));
+    auto areOpsCrudOnly = _areOpsCrudOnly(applyOpCmd);
+    auto isAtomic = allowAtomic && areOpsCrudOnly;
+    auto hasPrecondition = _hasPrecondition(applyOpCmd);
+
+    ScopedTransaction scopedXact(opCtx, MODE_X);
+    boost::optional<Lock::GlobalWrite> globalWriteLock;
+    boost::optional<Lock::DBLock> dbWriteLock;
+
+    // There's only one case where we are allowed to take the database lock instead of the global
+    // lock - no preconditions; only CRUD ops; and non-atomic mode.
+    if (!hasPrecondition && areOpsCrudOnly && !allowAtomic) {
+        dbWriteLock.emplace(opCtx->lockState(), dbName, MODE_IX);
+    } else {
+        globalWriteLock.emplace(opCtx->lockState());
+    }
 
-    bool userInitiatedWritesAndNotPrimary = txn->writesAreReplicated() &&
+    bool userInitiatedWritesAndNotPrimary = opCtx->writesAreReplicated() &&
         !repl::getGlobalReplicationCoordinator()->canAcceptWritesForDatabase(dbName);
 
     if (userInitiatedWritesAndNotPrimary)
         return Status(ErrorCodes::NotMaster,
                       str::stream() << "Not primary while applying ops to database " << dbName);
 
-    Status preconditionStatus = preconditionOK(txn, applyOpCmd, result);
-    if (!preconditionStatus.isOK()) {
-        return preconditionStatus;
+    if (hasPrecondition) {
+        auto status = _checkPrecondition(opCtx, applyOpCmd, result);
+        if (!status.isOK()) {
+            return status;
+        }
     }
 
     int numApplied = 0;
-    if (!canBeAtomic(applyOpCmd))
-        return _applyOps(txn, dbName, applyOpCmd, result, &numApplied);
+    if (!isAtomic)
+        return _applyOps(opCtx, dbName, applyOpCmd, result, &numApplied);
 
     // Perform write ops atomically
+    invariant(globalWriteLock);
     try {
         MONGO_WRITE_CONFLICT_RETRY_LOOP_BEGIN {
             BSONObjBuilder intermediateResult;
-            WriteUnitOfWork wunit(txn);
+            WriteUnitOfWork wunit(opCtx);
             numApplied = 0;
-            uassertStatusOK(_applyOps(txn, dbName, applyOpCmd, &intermediateResult, &numApplied));
+            {
+                // Suppress replication for atomic operations until end of applyOps.
+                repl::UnreplicatedWritesBlock uwb(opCtx);
+                uassertStatusOK(
+                    _applyOps(opCtx, dbName, applyOpCmd, &intermediateResult, &numApplied));
+            }
+            // Generate oplog entry for all atomic ops collectively.
+            if (opCtx->writesAreReplicated()) {
+                // We want this applied atomically on slaves so we rewrite the oplog entry without
+                // the pre-condition for speed.
+
+                std::string tempNS = str::stream() << dbName << ".$cmd";
+
+                BSONObjBuilder cmdBuilder;
+
+                for (auto elem : applyOpCmd) {
+                    auto name = elem.fieldNameStringData();
+                    if (name == kPreconditionFieldName)
+                        continue;
+                    if (name == "bypassDocumentValidation")
+                        continue;
+                    cmdBuilder.append(elem);
+                }
+
+                const BSONObj cmdRewritten = cmdBuilder.done();
+
+                auto opObserver = getGlobalServiceContext()->getOpObserver();
+                invariant(opObserver);
+                opObserver->onApplyOps(opCtx, tempNS, cmdRewritten);
+            }
             wunit.commit();
             result->appendElements(intermediateResult.obj());
         }
-        MONGO_WRITE_CONFLICT_RETRY_LOOP_END(txn, "applyOps", dbName);
+        MONGO_WRITE_CONFLICT_RETRY_LOOP_END(opCtx, "applyOps", dbName);
     } catch (const DBException& ex) {
         if (ex.getCode() == ErrorCodes::NamespaceNotFound) {
             // Retry in non-atomic mode, since MMAP cannot implicitly create a new database
             // within an active WriteUnitOfWork.
-            return _applyOps(txn, dbName, applyOpCmd, result, &numApplied);
+            return _applyOps(opCtx, dbName, applyOpCmd, result, &numApplied);
         }
         BSONArrayBuilder ab;
         ++numApplied;
diff --git a/src/mongo/db/catalog/apply_ops.h b/src/mongo/db/catalog/apply_ops.h
index 588d3bb370b..3a742891573 100644
--- a/src/mongo/db/catalog/apply_ops.h
+++ b/src/mongo/db/catalog/apply_ops.h
@@ -37,7 +37,7 @@ class OperationContext;
  * Applies ops contained in "applyOpCmd" and populates fields in "result" to be returned to the
  * user.
  */
-Status applyOps(OperationContext* txn,
+Status applyOps(OperationContext* opCtx,
                 const std::string& dbName,
                 const BSONObj& applyOpCmd,
                 BSONObjBuilder* result);
diff --git a/src/mongo/db/catalog/database.cpp b/src/mongo/db/catalog/database.cpp
index 9e24afd5ffb..23fdfb84e95 100644
--- a/src/mongo/db/catalog/database.cpp
+++ b/src/mongo/db/catalog/database.cpp
@@ -475,8 +475,8 @@ Status Database::renameCollection(OperationContext* txn,
         Top::get(txn->getClient()->getServiceContext()).collectionDropped(fromNS.toString());
     }
 
-    txn->recoveryUnit()->registerChange(new AddCollectionChange(txn, this, toNS));
     Status s = _dbEntry->renameCollection(txn, fromNS, toNS, stayTemp);
+    txn->recoveryUnit()->registerChange(new AddCollectionChange(txn, this, toNS));
     _collections[toNS] = _getOrCreateCollectionInstance(txn, toNS);
     return s;
 }
diff --git a/src/mongo/db/catalog/rename_collection.cpp b/src/mongo/db/catalog/rename_collection.cpp
index d9d25a951dc..f5f62ce9636 100644
--- a/src/mongo/db/catalog/rename_collection.cpp
+++ b/src/mongo/db/catalog/rename_collection.cpp
@@ -71,7 +71,15 @@ Status renameCollection(OperationContext* txn,
     DisableDocumentValidation validationDisabler(txn);
 
     ScopedTransaction transaction(txn, MODE_X);
-    Lock::GlobalWrite globalWriteLock(txn->lockState());
+    boost::optional<Lock::GlobalWrite> globalWriteLock;
+    boost::optional<Lock::DBLock> dbWriteLock;
+
+    // If the rename is known not to be a cross-database rename, just a database lock suffices.
+    if (source.db() == target.db())
+        dbWriteLock.emplace(txn->lockState(), source.db(), MODE_X);
+    else
+        globalWriteLock.emplace(txn->lockState());
+
     // We stay in source context the whole time. This is mostly to set the CurOp namespace.
     OldClientContext ctx(txn, source.ns());
 
diff --git a/src/mongo/db/commands/apply_ops_cmd.cpp b/src/mongo/db/commands/apply_ops_cmd.cpp
index c252b8fd073..262e8ef93d3 100644
--- a/src/mongo/db/commands/apply_ops_cmd.cpp
+++ b/src/mongo/db/commands/apply_ops_cmd.cpp
@@ -116,22 +116,17 @@ public:
             }
         }
 
-        auto client = txn->getClient();
-        auto lastOpAtOperationStart = repl::ReplClientInfo::forClient(client).getLastOp();
-        ScopeGuard lastOpSetterGuard =
-            MakeObjGuard(repl::ReplClientInfo::forClient(client),
-                         &repl::ReplClientInfo::setLastOpToSystemLastOpTime,
-                         txn);
-
+        // TODO (SERVER-30217): When a write concern is provided to the applyOps command, we
+        // normally wait on the OpTime of whichever operation successfully completed last. This is
+        // erroneous, however, if the last operation in the array happens to be a write no-op and
+        // thus isn’t assigned an OpTime. Let the second to last operation in the applyOps be write
+        // A, the last operation in applyOps be write B. Let B do a no-op write and let the
+        // operation that caused B to be a no-op be C. If C has an OpTime after A but before B,
+        // then we won’t wait for C to be replicated and it could be rolled back, even though B
+        // was acknowledged. To fix this, we should wait for replication of the node’s last applied
+        // OpTime if the last write operation was a no-op write.
         auto applyOpsStatus = appendCommandStatus(result, applyOps(txn, dbname, cmdObj, &result));
 
-        if (repl::ReplClientInfo::forClient(client).getLastOp() != lastOpAtOperationStart) {
-            // If this operation has already generated a new lastOp, don't bother setting it
-            // here. No-op applyOps will not generate a new lastOp, so we still need the guard to
-            // fire in that case.
-            lastOpSetterGuard.Dismiss();
-        }
-
         return applyOpsStatus;
     }
 
diff --git a/src/mongo/db/commands/create_indexes.cpp b/src/mongo/db/commands/create_indexes.cpp
index 489e080f7a6..ee0133da46b 100644
--- a/src/mongo/db/commands/create_indexes.cpp
+++ b/src/mongo/db/commands/create_indexes.cpp
@@ -290,12 +290,6 @@ public:
         const int numIndexesBefore = collection->getIndexCatalog()->numIndexesTotal(txn);
         result.append("numIndexesBefore", numIndexesBefore);
 
-        auto client = txn->getClient();
-        ScopeGuard lastOpSetterGuard =
-            MakeObjGuard(repl::ReplClientInfo::forClient(client),
-                         &repl::ReplClientInfo::setLastOpToSystemLastOpTime,
-                         txn);
-
         MultiIndexBlock indexer(txn, collection);
         indexer.allowBackgroundBuilding();
         indexer.allowInterruption();
@@ -403,8 +397,6 @@ public:
 
         result.append("numIndexesAfter", collection->getIndexCatalog()->numIndexesTotal(txn));
 
-        lastOpSetterGuard.Dismiss();
-
         return true;
     }
 
diff --git a/src/mongo/db/commands/dbcommands.cpp b/src/mongo/db/commands/dbcommands.cpp
index 37bcf783d47..38eae01ce9d 100644
--- a/src/mongo/db/commands/dbcommands.cpp
+++ b/src/mongo/db/commands/dbcommands.cpp
@@ -61,6 +61,7 @@
 #include "mongo/db/commands.h"
 #include "mongo/db/commands/server_status.h"
 #include "mongo/db/commands/shutdown.h"
+#include "mongo/db/concurrency/global_lock_acquisition_tracker.h"
 #include "mongo/db/concurrency/write_conflict_exception.h"
 #include "mongo/db/db_raii.h"
 #include "mongo/db/dbdirectclient.h"
@@ -1256,13 +1257,21 @@ namespace {
 
 void _waitForWriteConcernAndAddToCommandResponse(OperationContext* opCtx,
                                                  const std::string& commandName,
+                                                 const repl::OpTime& lastOpBeforeRun,
                                                  BSONObjBuilder* commandResponseBuilder) {
+    auto lastOpAfterRun = repl::ReplClientInfo::forClient(opCtx->getClient()).getLastOp();
+
+    // Ensures that if we tried to do a write, we wait for write concern, even if that write was
+    // a noop.
+    if ((lastOpAfterRun == lastOpBeforeRun) &&
+        GlobalLockAcquisitionTracker::get(opCtx).getGlobalExclusiveLockTaken()) {
+        repl::ReplClientInfo::forClient(opCtx->getClient()).setLastOpToSystemLastOpTime(opCtx);
+        lastOpAfterRun = repl::ReplClientInfo::forClient(opCtx->getClient()).getLastOp();
+    }
+
     WriteConcernResult res;
     auto waitForWCStatus =
-        waitForWriteConcern(opCtx,
-                            repl::ReplClientInfo::forClient(opCtx->getClient()).getLastOp(),
-                            opCtx->getWriteConcern(),
-                            &res);
+        waitForWriteConcern(opCtx, lastOpAfterRun, opCtx->getWriteConcern(), &res);
     Command::appendCommandWCStatus(*commandResponseBuilder, waitForWCStatus, res);
 
     // SERVER-22421: This code is to ensure error response backwards compatibility with the
@@ -1558,13 +1567,17 @@ bool Command::run(OperationContext* txn,
             return result;
         }
 
+        auto lastOpBeforeRun = repl::ReplClientInfo::forClient(txn->getClient()).getLastOp();
+
         // Change the write concern while running the command.
         const auto oldWC = txn->getWriteConcern();
         ON_BLOCK_EXIT([&] { txn->setWriteConcern(oldWC); });
         txn->setWriteConcern(wcResult.getValue());
 
-        ON_BLOCK_EXIT(
-            [&] { _waitForWriteConcernAndAddToCommandResponse(txn, getName(), &inPlaceReplyBob); });
+        ON_BLOCK_EXIT([&] {
+            _waitForWriteConcernAndAddToCommandResponse(
+                txn, getName(), lastOpBeforeRun, &inPlaceReplyBob);
+        });
 
         result = run(txn, db, cmd, 0, errmsg, inPlaceReplyBob);
 
diff --git a/src/mongo/db/commands/feature_compatibility_version.cpp b/src/mongo/db/commands/feature_compatibility_version.cpp
index f354efa88d9..00c6de52a5c 100644
--- a/src/mongo/db/commands/feature_compatibility_version.cpp
+++ b/src/mongo/db/commands/feature_compatibility_version.cpp
@@ -248,12 +248,8 @@ void FeatureCompatibilityVersion::set(OperationContext* txn, StringData version)
         client.runCommand(nss.db().toString(),
                           makeUpdateCommand(version, WriteConcernOptions::Majority),
                           updateResult);
-        uassertStatusOK(getStatusFromCommandResult(updateResult));
-        uassertStatusOK(getWriteConcernStatusFromCommandResult(updateResult));
+        uassertStatusOK(getStatusFromWriteCommandReply(updateResult));
 
-        // We then update the value of the featureCompatibilityVersion server parameter.
-        serverGlobalParams.featureCompatibility.version.store(
-            ServerGlobalParams::FeatureCompatibility::Version::k34);
     } else if (version == FeatureCompatibilityVersionCommandParser::kVersion32) {
         // We update the featureCompatibilityVersion document stored in the "admin.system.version"
         // collection. We do this before dropping the v=2 index in order to maintain the invariant
@@ -262,7 +258,7 @@ void FeatureCompatibilityVersion::set(OperationContext* txn, StringData version)
         // concern to this update because we're going to do so anyway for the "dropIndexes" command.
         BSONObj updateResult;
         client.runCommand(nss.db().toString(), makeUpdateCommand(version, BSONObj()), updateResult);
-        uassertStatusOK(getStatusFromCommandResult(updateResult));
+        uassertStatusOK(getStatusFromWriteCommandReply(updateResult));
 
         // We then drop the v=2 index on the "admin.system.version" collection to enable 3.2
         // secondaries to sync from this mongod.
@@ -278,10 +274,6 @@ void FeatureCompatibilityVersion::set(OperationContext* txn, StringData version)
             uassertStatusOK(status);
         }
         uassertStatusOK(getWriteConcernStatusFromCommandResult(dropIndexesResult));
-
-        // We then update the value of the featureCompatibilityVersion server parameter.
-        serverGlobalParams.featureCompatibility.version.store(
-            ServerGlobalParams::FeatureCompatibility::Version::k32);
     }
 }
 
@@ -323,7 +315,8 @@ void FeatureCompatibilityVersion::setIfCleanStartup(OperationContext* txn,
         }
 
         // We then insert the featureCompatibilityVersion document into the "admin.system.version"
-        // collection. We do this after creating the v=2 index in order to maintain the invariant
+        // collection. The server parameter will be updated on commit by the op observer.
+        // We do this after creating the v=2 index in order to maintain the invariant
         // that if the featureCompatibilityVersion is 3.4, then 'k32IncompatibleIndexSpec' index
         // exists on the "admin.system.version" collection. If we happened to fail to insert the
         // document when starting up, then on a subsequent start-up we'd no longer consider the data
@@ -334,14 +327,10 @@ void FeatureCompatibilityVersion::setIfCleanStartup(OperationContext* txn,
             BSON("_id" << FeatureCompatibilityVersion::kParameterName
                        << FeatureCompatibilityVersion::kVersionField
                        << FeatureCompatibilityVersionCommandParser::kVersion34)));
-
-        // We then update the value of the featureCompatibilityVersion server parameter.
-        serverGlobalParams.featureCompatibility.version.store(
-            ServerGlobalParams::FeatureCompatibility::Version::k34);
     }
 }
 
-void FeatureCompatibilityVersion::onInsertOrUpdate(const BSONObj& doc) {
+void FeatureCompatibilityVersion::onInsertOrUpdate(OperationContext* opCtx, const BSONObj& doc) {
     auto idElement = doc["_id"];
     if (idElement.type() != BSONType::String ||
         idElement.String() != FeatureCompatibilityVersion::kParameterName) {
@@ -350,10 +339,11 @@ void FeatureCompatibilityVersion::onInsertOrUpdate(const BSONObj& doc) {
     auto newVersion = uassertStatusOK(FeatureCompatibilityVersion::parse(doc));
     log() << "setting featureCompatibilityVersion to "
           << getFeatureCompatibilityVersionString(newVersion);
-    serverGlobalParams.featureCompatibility.version.store(newVersion);
+    opCtx->recoveryUnit()->onCommit(
+        [newVersion]() { serverGlobalParams.featureCompatibility.version.store(newVersion); });
 }
 
-void FeatureCompatibilityVersion::onDelete(const BSONObj& doc) {
+void FeatureCompatibilityVersion::onDelete(OperationContext* opCtx, const BSONObj& doc) {
     auto idElement = doc["_id"];
     if (idElement.type() != BSONType::String ||
         idElement.String() != FeatureCompatibilityVersion::kParameterName) {
@@ -361,15 +351,19 @@ void FeatureCompatibilityVersion::onDelete(const BSONObj& doc) {
     }
     log() << "setting featureCompatibilityVersion to "
           << FeatureCompatibilityVersionCommandParser::kVersion32;
-    serverGlobalParams.featureCompatibility.version.store(
-        ServerGlobalParams::FeatureCompatibility::Version::k32);
+    opCtx->recoveryUnit()->onCommit([]() {
+        serverGlobalParams.featureCompatibility.version.store(
+            ServerGlobalParams::FeatureCompatibility::Version::k32);
+    });
 }
 
-void FeatureCompatibilityVersion::onDropCollection() {
+void FeatureCompatibilityVersion::onDropCollection(OperationContext* opCtx) {
     log() << "setting featureCompatibilityVersion to "
           << FeatureCompatibilityVersionCommandParser::kVersion32;
-    serverGlobalParams.featureCompatibility.version.store(
-        ServerGlobalParams::FeatureCompatibility::Version::k32);
+    opCtx->recoveryUnit()->onCommit([]() {
+        serverGlobalParams.featureCompatibility.version.store(
+            ServerGlobalParams::FeatureCompatibility::Version::k32);
+    });
 }
 
 /**
diff --git a/src/mongo/db/commands/feature_compatibility_version.h b/src/mongo/db/commands/feature_compatibility_version.h
index 4bcb4b56e55..8e7969c2e70 100644
--- a/src/mongo/db/commands/feature_compatibility_version.h
+++ b/src/mongo/db/commands/feature_compatibility_version.h
@@ -76,21 +76,22 @@ public:
 
     /**
      * Examines a document inserted or updated in admin.system.version. If it is the
-     * featureCompatibilityVersion document, validates the document and updates the server
-     * parameter.
+     * featureCompatibilityVersion document, validates the document and on commit, updates
+     * the server parameter.
      */
-    static void onInsertOrUpdate(const BSONObj& doc);
+    static void onInsertOrUpdate(OperationContext* opCtx, const BSONObj& doc);
 
     /**
      * Examines the _id of a document removed from admin.system.version. If it is the
      * featureCompatibilityVersion document, resets the server parameter to its default value (3.2).
+     * on commit.
      */
-    static void onDelete(const BSONObj& doc);
+    static void onDelete(OperationContext* opCtx, const BSONObj& doc);
 
     /**
-     * Resets the server parameter to its default value (3.2).
+     * Resets the server parameter to its default value (3.2) on commit.
      */
-    static void onDropCollection();
+    static void onDropCollection(OperationContext* opCtx);
 };
 
 }  // namespace mongo
diff --git a/src/mongo/db/commands/find_and_modify.cpp b/src/mongo/db/commands/find_and_modify.cpp
index 79ee34203ce..61dbc4f6509 100644
--- a/src/mongo/db/commands/find_and_modify.cpp
+++ b/src/mongo/db/commands/find_and_modify.cpp
@@ -364,18 +364,6 @@ public:
         if (shouldBypassDocumentValidationForCommand(cmdObj))
             maybeDisableValidation.emplace(txn);
 
-        auto client = txn->getClient();
-        auto lastOpAtOperationStart = repl::ReplClientInfo::forClient(client).getLastOp();
-        ScopeGuard lastOpSetterGuard =
-            MakeObjGuard(repl::ReplClientInfo::forClient(client),
-                         &repl::ReplClientInfo::setLastOpToSystemLastOpTime,
-                         txn);
-
-        // If this is the local database, don't set last op.
-        if (dbName == "local") {
-            lastOpSetterGuard.Dismiss();
-        }
-
         auto curOp = CurOp::get(txn);
         OpDebug* opDebug = &curOp->debug();
 
@@ -567,13 +555,6 @@ public:
         }
         MONGO_WRITE_CONFLICT_RETRY_LOOP_END(txn, "findAndModify", nsString.ns());
 
-        if (repl::ReplClientInfo::forClient(client).getLastOp() != lastOpAtOperationStart) {
-            // If this operation has already generated a new lastOp, don't bother setting it here.
-            // No-op updates will not generate a new lastOp, so we still need the guard to fire in
-            // that case.
-            lastOpSetterGuard.Dismiss();
-        }
-
         return true;
     }
 
diff --git a/src/mongo/db/commands/mr.cpp b/src/mongo/db/commands/mr.cpp
index 4aeed6d5978..136656b8dd2 100644
--- a/src/mongo/db/commands/mr.cpp
+++ b/src/mongo/db/commands/mr.cpp
@@ -314,7 +314,7 @@ Config::Config(const string& _dbname, const BSONObj& cmdObj) {
         // scope and code
 
         if (cmdObj["scope"].type() == Object)
-            scopeSetup = cmdObj["scope"].embeddedObjectUserCheck();
+            scopeSetup = cmdObj["scope"].embeddedObjectUserCheck().getOwned();
 
         mapper.reset(new JSMapper(cmdObj["map"]));
         reducer.reset(new JSReducer(cmdObj["reduce"]));
@@ -322,7 +322,7 @@ Config::Config(const string& _dbname, const BSONObj& cmdObj) {
             finalizer.reset(new JSFinalizer(cmdObj["finalize"]));
 
         if (cmdObj["mapparams"].type() == Array) {
-            mapParams = cmdObj["mapparams"].embeddedObjectUserCheck();
+            mapParams = cmdObj["mapparams"].embeddedObjectUserCheck().getOwned();
         }
     }
 
@@ -836,6 +836,7 @@ void State::init() {
         AuthorizationSession::get(Client::getCurrent())->getAuthenticatedUserNamesToken();
     _scope.reset(getGlobalScriptEngine()->newScopeForCurrentThread());
     _scope->registerOperation(_txn);
+    _scope->requireOwnedObjects();
     _scope->setLocalDB(_config.dbname);
     _scope->loadStored(_txn, true);
 
@@ -1522,6 +1523,7 @@ public:
                 BSONObj o;
                 PlanExecutor::ExecState execState;
                 while (PlanExecutor::ADVANCED == (execState = exec->getNext(&o, NULL))) {
+                    o = o.getOwned();  // we will be accessing outside of the lock
                     // check to see if this is a new object we don't own yet
                     // because of a chunk migration
                     if (collMetadata) {
diff --git a/src/mongo/db/commands/pipeline_command.cpp b/src/mongo/db/commands/pipeline_command.cpp
index 6fe6de34381..03f8ec87e30 100644
--- a/src/mongo/db/commands/pipeline_command.cpp
+++ b/src/mongo/db/commands/pipeline_command.cpp
@@ -407,10 +407,6 @@ public:
                     return appendCommandStatus(result, resolvedView.getStatus());
                 }
 
-                auto collationSpec = ctx.getView()->defaultCollator()
-                    ? ctx.getView()->defaultCollator()->getSpec().toBSON().getOwned()
-                    : CollationSpec::kSimpleSpec;
-
                 // With the view & collation resolved, we can relinquish locks.
                 ctx.releaseLocksForView();
 
@@ -424,7 +420,6 @@ public:
                 if (!newRequest.isOK()) {
                     return appendCommandStatus(result, newRequest.getStatus());
                 }
-                newRequest.getValue().setCollation(collationSpec);
 
                 bool status = runParsed(
                     txn, origNss, newRequest.getValue(), newCmd.getValue(), errmsg, result);
diff --git a/src/mongo/db/concurrency/SConscript b/src/mongo/db/concurrency/SConscript
index 942b7a4f7f1..023d3b6b09e 100644
--- a/src/mongo/db/concurrency/SConscript
+++ b/src/mongo/db/concurrency/SConscript
@@ -13,6 +13,17 @@ env.Library(
 )
 
 env.Library(
+    target='global_lock_acquisition_tracker',
+    source=[
+        'global_lock_acquisition_tracker.cpp'
+        ],
+    LIBDEPS=[
+        '$BUILD_DIR/mongo/util/decorable',
+    ]
+)
+
+
+env.Library(
     target='lock_manager',
     source=[
         'd_concurrency.cpp',
@@ -21,6 +32,7 @@ env.Library(
         'lock_stats.cpp',
     ],
     LIBDEPS=[
+        'global_lock_acquisition_tracker',
         '$BUILD_DIR/mongo/util/background_job',
         '$BUILD_DIR/mongo/base',
         # Temporary crutch since the ssl cleanup is hard coded in background.cpp
@@ -44,6 +56,7 @@ env.CppUnitTest(
     LIBDEPS=[
         '$BUILD_DIR/mongo/db/service_context_noop_init',
         '$BUILD_DIR/mongo/util/progress_meter',
+        'global_lock_acquisition_tracker',
         'lock_manager',
     ]
 )
diff --git a/src/mongo/db/concurrency/d_concurrency.cpp b/src/mongo/db/concurrency/d_concurrency.cpp
index f7dabcc2248..6e7bd1ee7d6 100644
--- a/src/mongo/db/concurrency/d_concurrency.cpp
+++ b/src/mongo/db/concurrency/d_concurrency.cpp
@@ -34,6 +34,7 @@
 
 #include <string>
 
+#include "mongo/db/concurrency/global_lock_acquisition_tracker.h"
 #include "mongo/db/namespace_string.h"
 #include "mongo/db/server_parameters.h"
 #include "mongo/db/service_context.h"
@@ -91,6 +92,13 @@ void Lock::GlobalLock::waitForLock(unsigned timeoutMs) {
     if (_result != LOCK_OK && _locker->shouldConflictWithSecondaryBatchApplication()) {
         _pbwm.unlock();
     }
+
+    if (_locker->isWriteLocked() && haveClient()) {
+        auto opCtx = cc().getOperationContext();
+        if (opCtx) {
+            GlobalLockAcquisitionTracker::get(opCtx).setGlobalExclusiveLockTaken();
+        }
+    }
 }
 
 void Lock::GlobalLock::_unlock() {
diff --git a/src/mongo/db/concurrency/d_concurrency.h b/src/mongo/db/concurrency/d_concurrency.h
index 6274814ca1c..b8df869089e 100644
--- a/src/mongo/db/concurrency/d_concurrency.h
+++ b/src/mongo/db/concurrency/d_concurrency.h
@@ -163,6 +163,9 @@ public:
         /**
          * Enqueues lock but does not block on lock acquisition.
          * Call waitForLock() to complete locking process.
+         *
+         * Does not set that the global lock was taken on the GlobalLockAcquisitionTracker. Call
+         * waitForLock to do so.
          */
         GlobalLock(Locker* locker, LockMode lockMode, unsigned timeoutMs, EnqueueOnly enqueueOnly);
 
@@ -171,7 +174,8 @@ public:
         }
 
         /**
-         * Waits for lock to be granted.
+         * Waits for lock to be granted. Sets that the global lock was taken on the
+         * GlobalLockAcquisitionTracker.
          */
         void waitForLock(unsigned timeoutMs);
 
diff --git a/src/mongo/db/concurrency/d_concurrency_test.cpp b/src/mongo/db/concurrency/d_concurrency_test.cpp
index 22aae759650..799aae8a7e3 100644
--- a/src/mongo/db/concurrency/d_concurrency_test.cpp
+++ b/src/mongo/db/concurrency/d_concurrency_test.cpp
@@ -34,6 +34,7 @@
 #include <vector>
 
 #include "mongo/db/concurrency/d_concurrency.h"
+#include "mongo/db/concurrency/global_lock_acquisition_tracker.h"
 #include "mongo/db/concurrency/lock_manager_test_help.h"
 #include "mongo/db/operation_context.h"
 #include "mongo/stdx/functional.h"
@@ -80,12 +81,13 @@ private:
 class UseGlobalThrottling {
 public:
     explicit UseGlobalThrottling(OperationContext* opCtx, int numTickets)
-        : _opCtx(opCtx), _holder(1) {
+        : _opCtx(opCtx), _holder(numTickets) {
         _opCtx->lockState()->setGlobalThrottling(&_holder, &_holder);
     }
-    ~UseGlobalThrottling() {
+    ~UseGlobalThrottling() noexcept(false) {
         // Reset the global setting as we're about to destroy the ticket holder.
         _opCtx->lockState()->setGlobalThrottling(nullptr, nullptr);
+        ASSERT_EQ(_holder.used(), 0);
     }
 
 private:
@@ -115,6 +117,16 @@ makeKClientsWithLockers(int k) {
 }
 
 /**
+ * Returns an operation context that has an MMAPV1 locker attached to it.
+ */
+ServiceContext::UniqueOperationContext makeMMAPOperationContext() {
+    auto opCtx = cc().makeOperationContext();
+    opCtx->releaseLockState();
+    opCtx->setLockState(stdx::make_unique<MMAPV1LockerImpl>());
+    return opCtx;
+}
+
+/**
  * Calls fn the given number of iterations, spread out over up to maxThreads threads.
  * The threadNr passed is an integer between 0 and maxThreads exclusive. Logs timing
  * statistics for for all power-of-two thread counts from 1 up to maxThreds.
@@ -292,6 +304,86 @@ TEST(DConcurrency, GlobalLockX_Timeout) {
     }
 }
 
+TEST(DConcurrency, GlobalLockXSetsGlobalLockTakenOnOperationContext) {
+    Client::initThreadIfNotAlready();
+    auto opCtx = makeMMAPOperationContext();
+    ASSERT_FALSE(GlobalLockAcquisitionTracker::get(opCtx.get()).getGlobalExclusiveLockTaken());
+
+    {
+        Lock::GlobalLock globalWrite(opCtx->lockState(), MODE_X, 0);
+        ASSERT(globalWrite.isLocked());
+    }
+    ASSERT_TRUE(GlobalLockAcquisitionTracker::get(opCtx.get()).getGlobalExclusiveLockTaken());
+}
+
+TEST(DConcurrency, GlobalLockIXSetsGlobalLockTakenOnOperationContext) {
+    Client::initThreadIfNotAlready();
+    auto opCtx = makeMMAPOperationContext();
+    ASSERT_FALSE(GlobalLockAcquisitionTracker::get(opCtx.get()).getGlobalExclusiveLockTaken());
+    {
+        Lock::GlobalLock globalWrite(opCtx->lockState(), MODE_IX, 0);
+        ASSERT(globalWrite.isLocked());
+    }
+    ASSERT_TRUE(GlobalLockAcquisitionTracker::get(opCtx.get()).getGlobalExclusiveLockTaken());
+}
+
+TEST(DConcurrency, GlobalLockSDoesNotSetGlobalLockTakenOnOperationContext) {
+    Client::initThreadIfNotAlready();
+    auto opCtx = makeMMAPOperationContext();
+    ASSERT_FALSE(GlobalLockAcquisitionTracker::get(opCtx.get()).getGlobalExclusiveLockTaken());
+    {
+        Lock::GlobalLock globalRead(opCtx->lockState(), MODE_S, 0);
+        ASSERT(globalRead.isLocked());
+    }
+    ASSERT_FALSE(GlobalLockAcquisitionTracker::get(opCtx.get()).getGlobalExclusiveLockTaken());
+}
+
+TEST(DConcurrency, GlobalLockISDoesNotSetGlobalLockTakenOnOperationContext) {
+    Client::initThreadIfNotAlready();
+    auto opCtx = makeMMAPOperationContext();
+    ASSERT_FALSE(GlobalLockAcquisitionTracker::get(opCtx.get()).getGlobalExclusiveLockTaken());
+    {
+        Lock::GlobalLock globalRead(opCtx->lockState(), MODE_IS, 0);
+        ASSERT(globalRead.isLocked());
+    }
+    ASSERT_FALSE(GlobalLockAcquisitionTracker::get(opCtx.get()).getGlobalExclusiveLockTaken());
+}
+
+TEST(DConcurrency, DBLockXSetsGlobalLockTakenOnOperationContext) {
+    Client::initThreadIfNotAlready();
+    auto opCtx = makeMMAPOperationContext();
+    ASSERT_FALSE(GlobalLockAcquisitionTracker::get(opCtx.get()).getGlobalExclusiveLockTaken());
+
+    { Lock::DBLock dbWrite(opCtx->lockState(), "db", MODE_X); }
+    ASSERT_TRUE(GlobalLockAcquisitionTracker::get(opCtx.get()).getGlobalExclusiveLockTaken());
+}
+
+TEST(DConcurrency, DBLockSDoesNotSetGlobalLockTakenOnOperationContext) {
+    Client::initThreadIfNotAlready();
+    auto opCtx = makeMMAPOperationContext();
+    ASSERT_FALSE(GlobalLockAcquisitionTracker::get(opCtx.get()).getGlobalExclusiveLockTaken());
+
+    { Lock::DBLock dbRead(opCtx->lockState(), "db", MODE_S); }
+    ASSERT_FALSE(GlobalLockAcquisitionTracker::get(opCtx.get()).getGlobalExclusiveLockTaken());
+}
+
+TEST(DConcurrency, GlobalLockXDoesNotSetGlobalLockTakenWhenLockAcquisitionTimesOut) {
+    Client::initThreadIfNotAlready();
+    auto clients = makeKClientsWithLockers<MMAPV1LockerImpl>(1);
+
+    // Take a global lock so that the next one times out.
+    Lock::GlobalLock globalWrite0(clients[0].second.get()->lockState(), MODE_X, 0);
+    ASSERT(globalWrite0.isLocked());
+
+    auto opCtx = makeMMAPOperationContext();
+    ASSERT_FALSE(GlobalLockAcquisitionTracker::get(opCtx.get()).getGlobalExclusiveLockTaken());
+    {
+        Lock::GlobalLock globalWrite1(opCtx->lockState(), MODE_X, 1);
+        ASSERT_FALSE(globalWrite1.isLocked());
+    }
+    ASSERT_FALSE(GlobalLockAcquisitionTracker::get(opCtx.get()).getGlobalExclusiveLockTaken());
+}
+
 TEST(DConcurrency, GlobalLockS_NoTimeoutDueToGlobalLockS) {
     MMAPV1LockerImpl ls;
     Lock::GlobalRead globalRead(&ls);
@@ -720,6 +812,231 @@ TEST(DConcurrency, Throttling) {
     ASSERT(!overlongWait);
 }
 
+TEST(DConcurrency, NoThrottlingWhenNotAcquiringTickets) {
+    auto clientOpctxPairs = makeKClientsWithLockers<DefaultLockerImpl>(2);
+    auto opctx1 = clientOpctxPairs[0].second.get();
+    auto opctx2 = clientOpctxPairs[1].second.get();
+    // Limit the locker to 1 ticket at a time.
+    UseGlobalThrottling throttle(opctx1, 1);
+
+    // Prevent the enforcement of ticket throttling.
+    opctx1->lockState()->setShouldAcquireTicket(false);
+
+    // Both locks should be acquired immediately because there is no throttling.
+    Lock::GlobalRead R1(opctx1->lockState(), 0);
+    ASSERT(R1.isLocked());
+
+    Lock::GlobalRead R2(opctx2->lockState(), 0);
+    ASSERT(R2.isLocked());
+}
+
+TEST(DConcurrency, CompatibleFirstWithSXIS) {
+    auto clientOpctxPairs = makeKClientsWithLockers<DefaultLockerImpl>(3);
+    auto opctx1 = clientOpctxPairs[0].second.get();
+    auto opctx2 = clientOpctxPairs[1].second.get();
+    auto opctx3 = clientOpctxPairs[2].second.get();
+
+    // Build a queue of MODE_S <- MODE_X <- MODE_IS, with MODE_S granted.
+    Lock::GlobalRead lockS(opctx1->lockState());
+    ASSERT(lockS.isLocked());
+    Lock::GlobalLock lockX(opctx2->lockState(), MODE_X, UINT_MAX, Lock::GlobalLock::EnqueueOnly());
+    ASSERT(!lockX.isLocked());
+
+    // A MODE_IS should be granted due to compatibleFirst policy.
+    Lock::GlobalLock lockIS(opctx3->lockState(), MODE_IS, 0);
+    ASSERT(lockIS.isLocked());
+
+    lockX.waitForLock(0);
+    ASSERT(!lockX.isLocked());
+}
+
+
+TEST(DConcurrency, CompatibleFirstWithXSIXIS) {
+    auto clientOpctxPairs = makeKClientsWithLockers<DefaultLockerImpl>(4);
+    auto opctx1 = clientOpctxPairs[0].second.get();
+    auto opctx2 = clientOpctxPairs[1].second.get();
+    auto opctx3 = clientOpctxPairs[2].second.get();
+    auto opctx4 = clientOpctxPairs[3].second.get();
+
+    // Build a queue of MODE_X <- MODE_S <- MODE_IX <- MODE_IS, with MODE_X granted.
+    boost::optional<Lock::GlobalWrite> lockX;
+    lockX.emplace(opctx1->lockState());
+    ASSERT(lockX->isLocked());
+    boost::optional<Lock::GlobalLock> lockS;
+    lockS.emplace(opctx2->lockState(), MODE_S, UINT_MAX, Lock::GlobalLock::EnqueueOnly());
+    ASSERT(!lockS->isLocked());
+    Lock::GlobalLock lockIX(
+        opctx3->lockState(), MODE_IX, UINT_MAX, Lock::GlobalLock::EnqueueOnly());
+    ASSERT(!lockIX.isLocked());
+    Lock::GlobalLock lockIS(
+        opctx4->lockState(), MODE_IS, UINT_MAX, Lock::GlobalLock::EnqueueOnly());
+    ASSERT(!lockIS.isLocked());
+
+
+    // Now release the MODE_X and ensure that MODE_S will switch policy to compatibleFirst
+    lockX.reset();
+    lockS->waitForLock(0);
+    ASSERT(lockS->isLocked());
+    ASSERT(!lockIX.isLocked());
+    lockIS.waitForLock(0);
+    ASSERT(lockIS.isLocked());
+
+    // Now release the MODE_S and ensure that MODE_IX gets locked.
+    lockS.reset();
+    lockIX.waitForLock(0);
+    ASSERT(lockIX.isLocked());
+}
+
+TEST(DConcurrency, CompatibleFirstWithXSXIXIS) {
+    auto clientOpctxPairs = makeKClientsWithLockers<DefaultLockerImpl>(5);
+    auto opctx1 = clientOpctxPairs[0].second.get();
+    auto opctx2 = clientOpctxPairs[1].second.get();
+    auto opctx3 = clientOpctxPairs[2].second.get();
+    auto opctx4 = clientOpctxPairs[3].second.get();
+    auto opctx5 = clientOpctxPairs[4].second.get();
+
+    // Build a queue of MODE_X <- MODE_S <- MODE_X <- MODE_IX <- MODE_IS, with the first MODE_X
+    // granted and check that releasing it will result in the MODE_IS being granted.
+    boost::optional<Lock::GlobalWrite> lockXgranted;
+    lockXgranted.emplace(opctx1->lockState());
+    ASSERT(lockXgranted->isLocked());
+
+    boost::optional<Lock::GlobalLock> lockX;
+    lockX.emplace(opctx3->lockState(), MODE_X, UINT_MAX, Lock::GlobalLock::EnqueueOnly());
+    ASSERT(!lockX->isLocked());
+
+    // Now request MODE_S: it will be first in the pending list due to EnqueueAtFront policy.
+    boost::optional<Lock::GlobalLock> lockS;
+    lockS.emplace(opctx2->lockState(), MODE_S, UINT_MAX, Lock::GlobalLock::EnqueueOnly());
+    ASSERT(!lockS->isLocked());
+
+    Lock::GlobalLock lockIX(
+        opctx4->lockState(), MODE_IX, UINT_MAX, Lock::GlobalLock::EnqueueOnly());
+    ASSERT(!lockIX.isLocked());
+    Lock::GlobalLock lockIS(
+        opctx5->lockState(), MODE_IS, UINT_MAX, Lock::GlobalLock::EnqueueOnly());
+    ASSERT(!lockIS.isLocked());
+
+
+    // Now release the granted MODE_X and ensure that MODE_S will switch policy to compatibleFirst,
+    // not locking the MODE_X or MODE_IX, but instead granting the final MODE_IS.
+    lockXgranted.reset();
+    lockS->waitForLock(0);
+    ASSERT(lockS->isLocked());
+
+    lockX->waitForLock(0);
+    ASSERT(!lockX->isLocked());
+    lockIX.waitForLock(0);
+    ASSERT(!lockIX.isLocked());
+
+    lockIS.waitForLock(0);
+    ASSERT(lockIS.isLocked());
+}
+
+TEST(DConcurrency, CompatibleFirstStress) {
+    int numThreads = 8;
+    int testMicros = 500000;
+    AtomicUInt64 readOnlyInterval{0};
+    AtomicBool done{false};
+    std::vector<uint64_t> acquisitionCount(numThreads);
+    std::vector<uint64_t> timeoutCount(numThreads);
+    std::vector<uint64_t> busyWaitCount(numThreads);
+    auto clientOpctxPairs = makeKClientsWithLockers<DefaultLockerImpl>(numThreads);
+
+    // Do some busy waiting to trigger different timings. The atomic load prevents compilers
+    // from optimizing the loop away.
+    auto busyWait = [&done, &busyWaitCount](int threadId, long long iters) {
+        while (iters-- > 0) {
+            for (int i = 0; i < 100 && !done.load(); i++) {
+                busyWaitCount[threadId]++;
+            }
+        }
+    };
+
+    std::vector<stdx::thread> threads;
+
+    // Thread putting state in/out of read-only CompatibleFirst mode.
+    threads.emplace_back([&]() {
+        Timer t;
+        auto endTime = t.micros() + testMicros;
+        uint64_t readOnlyIntervalCount = 0;
+        OperationContext* opCtx = clientOpctxPairs[0].second.get();
+        for (int iters = 0; (t.micros() < endTime); iters++) {
+            busyWait(0, iters % 20);
+            Lock::GlobalRead readLock(opCtx->lockState(), iters % 2);
+            if (!readLock.isLocked()) {
+                timeoutCount[0]++;
+                continue;
+            }
+            acquisitionCount[0]++;
+            readOnlyInterval.store(++readOnlyIntervalCount);
+            busyWait(0, iters % 200);
+            readOnlyInterval.store(0);
+        };
+        done.store(true);
+    });
+
+    for (int threadId = 1; threadId < numThreads; threadId++) {
+        threads.emplace_back([&, threadId]() {
+            Timer t;
+            for (int iters = 0; !done.load(); iters++) {
+                OperationContext* opCtx = clientOpctxPairs[threadId].second.get();
+                boost::optional<Lock::GlobalLock> lock;
+                switch (threadId) {
+                    case 1:
+                    case 2:
+                    case 3:
+                    case 4: {
+                        // Here, actually try to acquire a lock without waiting, and check whether
+                        // we should have gotten the lock or not. Use MODE_IS in 95% of the cases,
+                        // and MODE_S in only 5, as that stressing the partitioning scheme and
+                        // policy changes more as thread 0 acquires/releases its MODE_S lock.
+                        busyWait(threadId, iters % 100);
+                        auto interval = readOnlyInterval.load();
+                        lock.emplace(opCtx->lockState(),
+                                     iters % 20 ? MODE_IS : MODE_S,
+                                     0,
+                                     Lock::GlobalLock::EnqueueOnly());
+                        // If thread 0 is holding the MODE_S lock while we tried to acquire a
+                        // MODE_IS or MODE_S lock, the CompatibleFirst policy guarantees success.
+                        auto newInterval = readOnlyInterval.load();
+                        invariant(!interval || interval != newInterval || lock->isLocked());
+                        lock->waitForLock(0);
+                        break;
+                    }
+                    case 5:
+                        busyWait(threadId, iters % 150);
+                        lock.emplace(opCtx->lockState(), MODE_X, iters % 2);
+                        busyWait(threadId, iters % 10);
+                        break;
+                    case 6:
+                        lock.emplace(opCtx->lockState(), iters % 25 ? MODE_IX : MODE_S, iters % 2);
+                        busyWait(threadId, iters % 100);
+                        break;
+                    case 7:
+                        busyWait(threadId, iters % 100);
+                        lock.emplace(opCtx->lockState(), iters % 20 ? MODE_IS : MODE_X, 0);
+                        break;
+                    default:
+                        MONGO_UNREACHABLE;
+                }
+                if (lock->isLocked())
+                    acquisitionCount[threadId]++;
+                else
+                    timeoutCount[threadId]++;
+            };
+        });
+    }
+
+    for (auto& thread : threads)
+        thread.join();
+    for (int threadId = 0; threadId < numThreads; threadId++) {
+        log() << "thread " << threadId << " stats: " << acquisitionCount[threadId]
+              << " acquisitions, " << timeoutCount[threadId] << " timeouts, "
+              << busyWaitCount[threadId] / 1000000 << "M busy waits";
+    }
+}
+
 // These tests exercise single- and multi-threaded performance of uncontended lock acquisition. It
 // is neither practical nor useful to run them on debug builds.
 
diff --git a/src/mongo/db/concurrency/global_lock_acquisition_tracker.cpp b/src/mongo/db/concurrency/global_lock_acquisition_tracker.cpp
new file mode 100644
index 00000000000..21ac9272c93
--- /dev/null
+++ b/src/mongo/db/concurrency/global_lock_acquisition_tracker.cpp
@@ -0,0 +1,45 @@
+/**
+ *    Copyright (C) 2017 MongoDB Inc.
+ *
+ *    This program is free software: you can redistribute it and/or  modify
+ *    it under the terms of the GNU Affero General Public License, version 3,
+ *    as published by the Free Software Foundation.
+ *
+ *    This program is distributed in the hope that it will be useful,
+ *    but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *    GNU Affero General Public License for more details.
+ *
+ *    You should have received a copy of the GNU Affero General Public License
+ *    along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ *    As a special exception, the copyright holders give permission to link the
+ *    code of portions of this program with the OpenSSL library under certain
+ *    conditions as described in each individual source file and distribute
+ *    linked combinations including the program with the OpenSSL library. You
+ *    must comply with the GNU Affero General Public License in all respects for
+ *    all of the code used other than as permitted herein. If you modify file(s)
+ *    with this exception, you may extend this exception to your version of the
+ *    file(s), but you are not obligated to do so. If you do not wish to do so,
+ *    delete this exception statement from your version. If you delete this
+ *    exception statement from all source files in the program, then also delete
+ *    it in the license file.
+ */
+
+#include "mongo/platform/basic.h"
+
+#include "mongo/db/concurrency/global_lock_acquisition_tracker.h"
+
+namespace mongo {
+
+const OperationContext::Decoration<GlobalLockAcquisitionTracker> GlobalLockAcquisitionTracker::get =
+    OperationContext::declareDecoration<GlobalLockAcquisitionTracker>();
+
+bool GlobalLockAcquisitionTracker::getGlobalExclusiveLockTaken() const {
+    return _globalExclusiveLockTaken;
+}
+
+void GlobalLockAcquisitionTracker::setGlobalExclusiveLockTaken() {
+    _globalExclusiveLockTaken = true;
+}
+}  // namespace mongo
diff --git a/src/mongo/db/s/collection_range_deleter.h b/src/mongo/db/concurrency/global_lock_acquisition_tracker.h
index 9d2e1ae8cb4..5bd68453f57 100644
--- a/src/mongo/db/s/collection_range_deleter.h
+++ b/src/mongo/db/concurrency/global_lock_acquisition_tracker.h
@@ -1,5 +1,5 @@
 /**
- *    Copyright (C) 2016 MongoDB Inc.
+ *    Copyright (C) 2017 MongoDB Inc.
  *
  *    This program is free software: you can redistribute it and/or  modify
  *    it under the terms of the GNU Affero General Public License, version 3,
@@ -25,53 +25,38 @@
  *    exception statement from all source files in the program, then also delete
  *    it in the license file.
  */
+
 #pragma once
 
-#include "mongo/base/disallow_copying.h"
-#include "mongo/db/namespace_string.h"
-#include "mongo/s/catalog/type_chunk.h"
+#include "mongo/db/operation_context.h"
 
 namespace mongo {
 
-class BSONObj;
-class Collection;
-class OperationContext;
-
-class CollectionRangeDeleter {
-    MONGO_DISALLOW_COPYING(CollectionRangeDeleter);
-
+/**
+ * The GlobalLockAcquisitionTracker keeps track of if the global lock has ever been taken in X or
+ * IX mode. This class is used to track if we ever did a transaction with the intent to do a
+ * write, so that we can enforce write concern on noop writes.
+ */
+class GlobalLockAcquisitionTracker {
 public:
-    CollectionRangeDeleter(NamespaceString nss);
+    static const OperationContext::Decoration<GlobalLockAcquisitionTracker> get;
 
-    /**
-     * Starts deleting ranges and cleans up this object when it is finished.
-     */
-    void run();
+    // Decoration requires a default constructor.
+    GlobalLockAcquisitionTracker() = default;
 
     /**
-     * Acquires the collection IX lock and checks whether there are new entries for the collection's
-     * rangesToClean structure.  If there are, deletes some small amount of entries and yields using
-     * the standard query yielding logic.
-     *
-     * Returns true if there are more entries in rangesToClean, false if there is no more progress
-     * to be made.
+     * Returns whether we have ever taken a global lock in X or IX mode in this operation.
      */
-    bool cleanupNextRange(OperationContext* txn);
+    bool getGlobalExclusiveLockTaken() const;
 
-private:
     /**
-     * Performs the deletion of a small amount of entries within the range in progress.
-     * This function will invariant if called while _rangeInProgress is not set.
-     *
-     * Returns the number of documents deleted (0 if deletion is finished), or -1 for error.
+     * Sets that we have ever taken a global lock in X or IX mode in this operation.
      */
-    int _doDeletion(OperationContext* txn, Collection* collection, const BSONObj& keyPattern);
+    void setGlobalExclusiveLockTaken();
 
-    NamespaceString _nss;
-
-    // Holds a range for which deletion has begun. If empty, then a new range
-    // must be requested from rangesToClean
-    boost::optional<ChunkRange> _rangeInProgress;
+private:
+    // Set to true when the global lock is first taken in X or IX mode. Never set back to false.
+    bool _globalExclusiveLockTaken = false;
 };
 
 }  // namespace mongo
diff --git a/src/mongo/db/concurrency/lock_manager.cpp b/src/mongo/db/concurrency/lock_manager.cpp
index f90141e79be..8ae7919012e 100644
--- a/src/mongo/db/concurrency/lock_manager.cpp
+++ b/src/mongo/db/concurrency/lock_manager.cpp
@@ -592,7 +592,9 @@ bool LockManager::unlock(LockRequest* request) {
         lock->decGrantedModeCount(request->mode);
 
         if (request->compatibleFirst) {
+            invariant(lock->compatibleFirstCount > 0);
             lock->compatibleFirstCount--;
+            invariant(lock->compatibleFirstCount == 0 || !lock->grantedList.empty());
         }
 
         _onLockModeChanged(lock, lock->grantedCounts[request->mode] == 0);
diff --git a/src/mongo/db/concurrency/lock_state.cpp b/src/mongo/db/concurrency/lock_state.cpp
index 45ea279c137..3625cef24f7 100644
--- a/src/mongo/db/concurrency/lock_state.cpp
+++ b/src/mongo/db/concurrency/lock_state.cpp
@@ -311,7 +311,7 @@ LockResult LockerImpl<IsForMMAPV1>::_lockGlobalBegin(LockMode mode, Milliseconds
     dassert(isLocked() == (_modeForTicket != MODE_NONE));
     if (_modeForTicket == MODE_NONE) {
         const bool reader = isSharedLockMode(mode);
-        auto holder = ticketHolders[mode];
+        auto holder = shouldAcquireTicket() ? ticketHolders[mode] : nullptr;
         if (holder) {
             _clientState.store(reader ? kQueuedReader : kQueuedWriter);
             if (timeout == Milliseconds::max()) {
@@ -787,7 +787,10 @@ LockResult LockerImpl<IsForMMAPV1>::lockComplete(ResourceId resId,
         }
     }
 
-    // Cleanup the state, since this is an unused lock now
+    // Cleanup the state, since this is an unused lock now.
+    // Note: in case of the _notify object returning LOCK_TIMEOUT, it is possible to find that the
+    // lock was still granted after all, but we don't try to take advantage of that and will return
+    // a timeout.
     if (result != LOCK_OK) {
         LockRequestsMap::Iterator it = _requests.find(resId);
         _unlockImpl(&it);
@@ -807,7 +810,7 @@ bool LockerImpl<IsForMMAPV1>::_unlockImpl(LockRequestsMap::Iterator* it) {
     if (globalLockManager.unlock(it->objAddr())) {
         if (it->key() == resourceIdGlobal) {
             invariant(_modeForTicket != MODE_NONE);
-            auto holder = ticketHolders[_modeForTicket];
+            auto holder = shouldAcquireTicket() ? ticketHolders[_modeForTicket] : nullptr;
             _modeForTicket = MODE_NONE;
             if (holder) {
                 holder->release();
diff --git a/src/mongo/db/concurrency/locker.h b/src/mongo/db/concurrency/locker.h
index 37af569124f..c1443865326 100644
--- a/src/mongo/db/concurrency/locker.h
+++ b/src/mongo/db/concurrency/locker.h
@@ -334,11 +334,25 @@ public:
         return _shouldConflictWithSecondaryBatchApplication;
     }
 
+    /**
+     * If set to false, this opts out of the ticket mechanism. This should be used sparingly
+     * for special purpose threads, such as FTDC.
+     */
+    void setShouldAcquireTicket(bool newValue) {
+        invariant(!isLocked());
+        _shouldAcquireTicket = newValue;
+    }
+    bool shouldAcquireTicket() const {
+        return _shouldAcquireTicket;
+    }
+
+
 protected:
     Locker() {}
 
 private:
     bool _shouldConflictWithSecondaryBatchApplication = true;
+    bool _shouldAcquireTicket = true;
 };
 
 }  // namespace mongo
diff --git a/src/mongo/db/db.cpp b/src/mongo/db/db.cpp
index 2dbb2b1065e..c6272425bf5 100644
--- a/src/mongo/db/db.cpp
+++ b/src/mongo/db/db.cpp
@@ -103,10 +103,10 @@
 #include "mongo/db/startup_warnings_mongod.h"
 #include "mongo/db/stats/counters.h"
 #include "mongo/db/stats/snapshots.h"
+#include "mongo/db/storage/encryption_hooks.h"
 #include "mongo/db/storage/mmap_v1/mmap_v1_options.h"
 #include "mongo/db/storage/storage_engine.h"
 #include "mongo/db/storage/storage_options.h"
-#include "mongo/db/storage/wiredtiger/wiredtiger_customization_hooks.h"
 #include "mongo/db/ttl.h"
 #include "mongo/db/wire_version.h"
 #include "mongo/executor/network_interface_factory.h"
@@ -596,7 +596,7 @@ ExitCode _initAndListen(int listenPort) {
     getGlobalServiceContext()->initializeGlobalStorageEngine();
 
 #ifdef MONGO_CONFIG_WIREDTIGER_ENABLED
-    if (WiredTigerCustomizationHooks::get(getGlobalServiceContext())->restartRequired()) {
+    if (EncryptionHooks::get(getGlobalServiceContext())->restartRequired()) {
         exitCleanly(EXIT_CLEAN);
     }
 #endif
@@ -704,7 +704,11 @@ ExitCode _initAndListen(int listenPort) {
         Status status = authindex::verifySystemIndexes(startupOpCtx.get());
         if (!status.isOK()) {
             log() << redact(status);
-            exitCleanly(EXIT_NEED_UPGRADE);
+            if (status.code() == ErrorCodes::AuthSchemaIncompatible) {
+                exitCleanly(EXIT_NEED_UPGRADE);
+            } else {
+                quickExit(EXIT_FAILURE);
+            }
         }
 
         // SERVER-14090: Verify that auth schema version is schemaVersion26Final.
@@ -752,7 +756,7 @@ ExitCode _initAndListen(int listenPort) {
     if (!storageGlobalParams.readOnly) {
         logStartup(startupOpCtx.get());
 
-        startFTDC();
+        startMongoDFTDC();
 
         getDeleter()->startWorkers();
 
@@ -1108,7 +1112,7 @@ static void shutdownTask() {
 #endif
 
     // Shutdown Full-Time Data Capture
-    stopFTDC();
+    stopMongoDFTDC();
 
     if (txn) {
         ShardingState::get(txn)->shutDown(txn);
diff --git a/src/mongo/db/dbhelpers.cpp b/src/mongo/db/dbhelpers.cpp
index ebed6807c7e..96eb011018d 100644
--- a/src/mongo/db/dbhelpers.cpp
+++ b/src/mongo/db/dbhelpers.cpp
@@ -62,8 +62,8 @@
 #include "mongo/db/s/sharding_state.h"
 #include "mongo/db/service_context.h"
 #include "mongo/db/storage/data_protector.h"
+#include "mongo/db/storage/encryption_hooks.h"
 #include "mongo/db/storage/storage_options.h"
-#include "mongo/db/storage/wiredtiger/wiredtiger_customization_hooks.h"
 #include "mongo/db/write_concern.h"
 #include "mongo/db/write_concern_options.h"
 #include "mongo/s/shard_key_pattern.h"
@@ -270,6 +270,7 @@ long long Helpers::removeRange(OperationContext* txn,
                                const KeyRange& range,
                                BoundInclusion boundInclusion,
                                const WriteConcernOptions& writeConcern,
+                               Milliseconds& replWaitDuration,
                                RemoveSaver* callback,
                                bool fromMigrate,
                                bool onlyRemoveOrphanedDocs) {
@@ -283,6 +284,7 @@ long long Helpers::removeRange(OperationContext* txn,
     BSONObj max;
 
     {
+        ScopedTransaction scopedXact(txn, MODE_IS);
         AutoGetCollectionForRead ctx(txn, ns);
         Collection* collection = ctx.getCollection();
         if (!collection) {
@@ -326,11 +328,12 @@ long long Helpers::removeRange(OperationContext* txn,
 
     long long numDeleted = 0;
 
-    Milliseconds millisWaitingForReplication{0};
+    replWaitDuration = Milliseconds::zero();
 
     while (1) {
         // Scoping for write lock.
         {
+            ScopedTransaction scopedXact(txn, MODE_IX);
             AutoGetCollection ctx(txn, NamespaceString(ns), MODE_IX, MODE_IX);
             Collection* collection = ctx.getCollection();
             if (!collection)
@@ -438,20 +441,21 @@ long long Helpers::removeRange(OperationContext* txn,
                     txn,
                     repl::ReplClientInfo::forClient(txn->getClient()).getLastOp(),
                     writeConcern);
-            if (replStatus.status.code() == ErrorCodes::ExceededTimeLimit) {
+            if (replStatus.status.code() == ErrorCodes::ExceededTimeLimit ||
+                replStatus.status.code() == ErrorCodes::WriteConcernFailed) {
                 warning(LogComponent::kSharding) << "replication to secondaries for removeRange at "
                                                     "least 60 seconds behind";
             } else {
                 uassertStatusOK(replStatus.status);
             }
-            millisWaitingForReplication += replStatus.duration;
+            replWaitDuration += replStatus.duration;
         }
     }
 
     if (writeConcern.shouldWaitForOtherNodes())
         log(LogComponent::kSharding)
             << "Helpers::removeRangeUnlocked time spent waiting for replication: "
-            << durationCount<Milliseconds>(millisWaitingForReplication) << "ms" << endl;
+            << durationCount<Milliseconds>(replWaitDuration) << "ms" << endl;
 
     MONGO_LOG_COMPONENT(1, LogComponent::kSharding) << "end removal of " << min << " to " << max
                                                     << " in " << ns << " (took "
@@ -485,19 +489,19 @@ Helpers::RemoveSaver::RemoveSaver(const string& a, const string& b, const string
     ss << why << "." << terseCurrentTime(false) << "." << NUM++ << ".bson";
     _file /= ss.str();
 
-    auto hooks = WiredTigerCustomizationHooks::get(getGlobalServiceContext());
-    if (hooks->enabled()) {
-        _protector = hooks->getDataProtector();
-        _file += hooks->getProtectedPathSuffix();
+    auto encryptionHooks = EncryptionHooks::get(getGlobalServiceContext());
+    if (encryptionHooks->enabled()) {
+        _protector = encryptionHooks->getDataProtector();
+        _file += encryptionHooks->getProtectedPathSuffix();
     }
 }
 
 Helpers::RemoveSaver::~RemoveSaver() {
     if (_protector && _out) {
-        auto hooks = WiredTigerCustomizationHooks::get(getGlobalServiceContext());
-        invariant(hooks->enabled());
+        auto encryptionHooks = EncryptionHooks::get(getGlobalServiceContext());
+        invariant(encryptionHooks->enabled());
 
-        size_t protectedSizeMax = hooks->additionalBytesForProtectedBuffer();
+        size_t protectedSizeMax = encryptionHooks->additionalBytesForProtectedBuffer();
         std::unique_ptr<uint8_t[]> protectedBuffer(new uint8_t[protectedSizeMax]);
 
         size_t resultLen;
@@ -560,10 +564,10 @@ Status Helpers::RemoveSaver::goingToDelete(const BSONObj& o) {
 
     std::unique_ptr<uint8_t[]> protectedBuffer;
     if (_protector) {
-        auto hooks = WiredTigerCustomizationHooks::get(getGlobalServiceContext());
-        invariant(hooks->enabled());
+        auto encryptionHooks = EncryptionHooks::get(getGlobalServiceContext());
+        invariant(encryptionHooks->enabled());
 
-        size_t protectedSizeMax = dataSize + hooks->additionalBytesForProtectedBuffer();
+        size_t protectedSizeMax = dataSize + encryptionHooks->additionalBytesForProtectedBuffer();
         protectedBuffer.reset(new uint8_t[protectedSizeMax]);
 
         size_t resultLen;
diff --git a/src/mongo/db/dbhelpers.h b/src/mongo/db/dbhelpers.h
index 650f265265d..59a4a317db7 100644
--- a/src/mongo/db/dbhelpers.h
+++ b/src/mongo/db/dbhelpers.h
@@ -141,11 +141,13 @@ struct Helpers {
     /**
      * Takes a namespace range, specified by a min and max and qualified by an index pattern,
      * and removes all the documents in that range found by iterating
-     * over the given index. Caller is responsible for insuring that min/max are
+     * over the given index. Caller is responsible for ensuring that min/max are
      * compatible with the given keyPattern (e.g min={a:100} is compatible with
      * keyPattern={a:1,b:1} since it can be extended to {a:100,b:minKey}, but
      * min={b:100} is not compatible).
      *
+     * Returns time spent waiting for majority replication in replWaitDuration.
+     *
      * Caller must hold a write lock on 'ns'
      *
      * Returns -1 when no usable index exists
@@ -157,6 +159,7 @@ struct Helpers {
                                  const KeyRange& range,
                                  BoundInclusion boundInclusion,
                                  const WriteConcernOptions& secondaryThrottle,
+                                 Milliseconds& replWaitDuration,
                                  RemoveSaver* callback = NULL,
                                  bool fromMigrate = false,
                                  bool onlyRemoveOrphanedDocs = false);
diff --git a/src/mongo/db/dbwebserver.cpp b/src/mongo/db/dbwebserver.cpp
index b202e0500b9..50fa68a8176 100644
--- a/src/mongo/db/dbwebserver.cpp
+++ b/src/mongo/db/dbwebserver.cpp
@@ -65,6 +65,7 @@
 namespace mongo {
 
 using std::map;
+using std::string;
 using std::stringstream;
 using std::vector;
 
diff --git a/src/mongo/db/ftdc/SConscript b/src/mongo/db/ftdc/SConscript
index 0d16aec93f2..81df9ad0f4d 100644
--- a/src/mongo/db/ftdc/SConscript
+++ b/src/mongo/db/ftdc/SConscript
@@ -21,7 +21,6 @@ ftdcEnv.Library(
     LIBDEPS=[
         '$BUILD_DIR/mongo/base',
         '$BUILD_DIR/mongo/bson/util/bson_extract',
-        '$BUILD_DIR/mongo/db/server_options_core',
         '$BUILD_DIR/mongo/db/service_context',
         '$BUILD_DIR/third_party/s2/s2', # For VarInt
         '$BUILD_DIR/third_party/shim_zlib',
@@ -40,24 +39,44 @@ elif env.TargetOSIs('windows'):
     ]
 
 env.Library(
-    target='ftdc_mongod',
+    target='ftdc_server',
     source=[
-        'ftdc_commands.cpp',
-        'ftdc_mongod.cpp',
+        'ftdc_server.cpp',
         'ftdc_system_stats.cpp',
         'ftdc_system_stats_${TARGET_OS}.cpp',
     ],
     LIBDEPS=[
         '$BUILD_DIR/mongo/base',
         '$BUILD_DIR/mongo/db/commands',
-        '$BUILD_DIR/mongo/db/repl/repl_coordinator_global',
         '$BUILD_DIR/mongo/db/server_parameters',
-        '$BUILD_DIR/mongo/db/storage/storage_options',
         '$BUILD_DIR/mongo/util/processinfo',
         'ftdc'
     ] + platform_libs,
 )
 
+env.Library(
+    target='ftdc_mongod',
+    source=[
+        'ftdc_commands.cpp',
+        'ftdc_mongod.cpp',
+    ],
+    LIBDEPS=[
+        '$BUILD_DIR/mongo/db/repl/repl_coordinator_global',
+        '$BUILD_DIR/mongo/db/storage/storage_options',
+        'ftdc_server'
+    ],
+)
+
+env.Library(
+    target='ftdc_mongos',
+    source=[
+        'ftdc_mongos.cpp',
+    ],
+    LIBDEPS=[
+        'ftdc_server'
+    ],
+)
+
 env.CppUnitTest(
     target='ftdc_test',
     source=[
diff --git a/src/mongo/db/ftdc/collector.cpp b/src/mongo/db/ftdc/collector.cpp
index 611f12dff5a..7c35fb12bb6 100644
--- a/src/mongo/db/ftdc/collector.cpp
+++ b/src/mongo/db/ftdc/collector.cpp
@@ -67,6 +67,7 @@ std::tuple<BSONObj, Date_t> FTDCCollectorCollection::collect(Client* client) {
     // batches that are taking a long time.
     auto txn = client->makeOperationContext();
     txn->lockState()->setShouldConflictWithSecondaryBatchApplication(false);
+    txn->lockState()->setShouldAcquireTicket(false);
 
     for (auto& collector : _collectors) {
         BSONObjBuilder subObjBuilder(builder.subobjStart(collector->name()));
diff --git a/src/mongo/db/ftdc/compressor_test.cpp b/src/mongo/db/ftdc/compressor_test.cpp
index 688197a392a..0c01bd58040 100644
--- a/src/mongo/db/ftdc/compressor_test.cpp
+++ b/src/mongo/db/ftdc/compressor_test.cpp
@@ -368,7 +368,7 @@ TEST(FTDCCompressor, TestNumbersCompat) {
 }
 
 // Test various date time types
-TEST(AFTDCCompressor, TestDateTimeTypes) {
+TEST(FTDCCompressor, TestDateTimeTypes) {
     TestTie c;
     for (int i = 0; i < 10; i++) {
         BSONObjBuilder builder1;
diff --git a/src/mongo/db/ftdc/constants.h b/src/mongo/db/ftdc/constants.h
index a989fd35b64..3fe5191ff44 100644
--- a/src/mongo/db/ftdc/constants.h
+++ b/src/mongo/db/ftdc/constants.h
@@ -26,6 +26,10 @@
  * then also delete it in the license file.
  */
 
+#pragma once
+
+#include "mongo/base/string_data.h"
+
 namespace mongo {
 
 extern const char kFTDCInterimFile[];
@@ -42,4 +46,6 @@ extern const char kFTDCDocsField[];
 extern const char kFTDCCollectStartField[];
 extern const char kFTDCCollectEndField[];
 
+constexpr StringData kFTDCDefaultDirectory = "diagnostic.data"_sd;
+
 }  // namespace mongo
diff --git a/src/mongo/db/ftdc/controller.cpp b/src/mongo/db/ftdc/controller.cpp
index 9e2987db0f9..bc3d4df444c 100644
--- a/src/mongo/db/ftdc/controller.cpp
+++ b/src/mongo/db/ftdc/controller.cpp
@@ -47,9 +47,19 @@
 
 namespace mongo {
 
-void FTDCController::setEnabled(bool enabled) {
+Status FTDCController::setEnabled(bool enabled) {
     stdx::lock_guard<stdx::mutex> lock(_mutex);
+
+    if (_path.empty()) {
+        return Status(ErrorCodes::FTDCPathNotSet,
+                      str::stream() << "FTDC cannot be enabled without setting the set parameter "
+                                       "'diagnosticDataCollectionDirectoryPath' first.");
+    }
+
     _configTemp.enabled = enabled;
+    _condvar.notify_one();
+
+    return Status::OK();
 }
 
 void FTDCController::setPeriod(Milliseconds millis) {
@@ -82,6 +92,23 @@ void FTDCController::setMaxSamplesPerInterimMetricChunk(size_t size) {
     _condvar.notify_one();
 }
 
+Status FTDCController::setDirectory(const boost::filesystem::path& path) {
+    stdx::lock_guard<stdx::mutex> lock(_mutex);
+
+    if (!_path.empty()) {
+        return Status(ErrorCodes::FTDCPathAlreadySet,
+                      str::stream() << "FTDC path has already been set to '" << _path.string()
+                                    << "'. It cannot be changed.");
+    }
+
+    _path = path;
+
+    // Do not notify for the change since it has to be enabled via setEnabled.
+
+    return Status::OK();
+}
+
+
 void FTDCController::addPeriodicCollector(std::unique_ptr<FTDCCollectorInterface> collector) {
     {
         stdx::lock_guard<stdx::mutex> lock(_mutex);
@@ -203,7 +230,7 @@ void FTDCController::doLoop() {
             }
 
             // TODO: consider only running this thread if we are enabled
-            // for now, we just keep an idle thread as it is simplier
+            // for now, we just keep an idle thread as it is simpler
             if (_config.enabled) {
                 // Delay initialization of FTDCFileManager until we are sure the user has enabled
                 // FTDC
diff --git a/src/mongo/db/ftdc/controller.h b/src/mongo/db/ftdc/controller.h
index 44b5e838506..e73d570598d 100644
--- a/src/mongo/db/ftdc/controller.h
+++ b/src/mongo/db/ftdc/controller.h
@@ -62,8 +62,12 @@ public:
 
     /*
      * Set whether the controller is enabled, and collects data.
+     *
+     * Returns ErrorCodes::FTDCPathNotSet if no log path has been specified for FTDC. This occurs
+     * in MongoS in some situations since MongoS is not required to have a storage directory like
+     * MongoD does.
      */
-    void setEnabled(bool enabled);
+    Status setEnabled(bool enabled);
 
     /**
      * Set the period for data collection.
@@ -94,6 +98,13 @@ public:
      */
     void setMaxSamplesPerInterimMetricChunk(size_t size);
 
+    /*
+     * Set the path to store FTDC files if not already set.
+     *
+     * Returns ErrorCodes::FTDCPathAlreadySet if the path has already been set.
+     */
+    Status setDirectory(const boost::filesystem::path& path);
+
     /**
      * Add a metric collector to collect periodically. i.e., serverStatus
      */
@@ -172,7 +183,7 @@ private:
     State _state{State::kNotStarted};
 
     // Directory to store files
-    const boost::filesystem::path _path;
+    boost::filesystem::path _path;
 
     // Mutex to protect the condvar, configuration changes, and most recent periodic document.
     stdx::mutex _mutex;
diff --git a/src/mongo/db/ftdc/controller_test.cpp b/src/mongo/db/ftdc/controller_test.cpp
index 365f06580cd..c75a3fe44d6 100644
--- a/src/mongo/db/ftdc/controller_test.cpp
+++ b/src/mongo/db/ftdc/controller_test.cpp
@@ -254,7 +254,7 @@ TEST(FTDCControllerTest, TestStartAsDisabled) {
 
     ASSERT_EQUALS(files0.size(), 0UL);
 
-    c.setEnabled(true);
+    ASSERT_OK(c.setEnabled(true));
 
     c1Ptr->setSignalOnCount(50);
 
diff --git a/src/mongo/db/ftdc/ftdc_mongod.cpp b/src/mongo/db/ftdc/ftdc_mongod.cpp
index 094a2b05d06..1f53ec345ea 100644
--- a/src/mongo/db/ftdc/ftdc_mongod.cpp
+++ b/src/mongo/db/ftdc/ftdc_mongod.cpp
@@ -31,295 +31,18 @@
 #include "mongo/db/ftdc/ftdc_mongod.h"
 
 #include <boost/filesystem.hpp>
-#include <fstream>
-#include <memory>
 
-#include "mongo/base/init.h"
-#include "mongo/base/status.h"
-#include "mongo/bson/bsonobjbuilder.h"
-#include "mongo/db/commands.h"
-#include "mongo/db/ftdc/collector.h"
-#include "mongo/db/ftdc/config.h"
+#include "mongo/db/ftdc/constants.h"
 #include "mongo/db/ftdc/controller.h"
-#include "mongo/db/ftdc/ftdc_system_stats.h"
-#include "mongo/db/jsobj.h"
+#include "mongo/db/ftdc/ftdc_server.h"
 #include "mongo/db/repl/replication_coordinator.h"
 #include "mongo/db/repl/replication_coordinator_global.h"
-#include "mongo/db/server_parameters.h"
-#include "mongo/db/service_context.h"
 #include "mongo/db/storage/storage_options.h"
 
 namespace mongo {
 
 namespace {
-
-const auto getFTDCController = ServiceContext::declareDecoration<std::unique_ptr<FTDCController>>();
-
-FTDCController* getGlobalFTDCController() {
-    if (!hasGlobalServiceContext()) {
-        return nullptr;
-    }
-
-    return getFTDCController(getGlobalServiceContext()).get();
-}
-
-std::atomic<bool> localEnabledFlag(FTDCConfig::kEnabledDefault);  // NOLINT
-
-class ExportedFTDCEnabledParameter
-    : public ExportedServerParameter<bool, ServerParameterType::kStartupAndRuntime> {
-public:
-    ExportedFTDCEnabledParameter()
-        : ExportedServerParameter<bool, ServerParameterType::kStartupAndRuntime>(
-              ServerParameterSet::getGlobal(),
-              "diagnosticDataCollectionEnabled",
-              &localEnabledFlag) {}
-
-    virtual Status validate(const bool& potentialNewValue) {
-        auto controller = getGlobalFTDCController();
-        if (controller) {
-            controller->setEnabled(potentialNewValue);
-        }
-
-        return Status::OK();
-    }
-
-} exportedFTDCEnabledParameter;
-
-std::atomic<std::int32_t> localPeriodMillis(FTDCConfig::kPeriodMillisDefault);  // NOLINT
-
-class ExportedFTDCPeriodParameter
-    : public ExportedServerParameter<std::int32_t, ServerParameterType::kStartupAndRuntime> {
-public:
-    ExportedFTDCPeriodParameter()
-        : ExportedServerParameter<std::int32_t, ServerParameterType::kStartupAndRuntime>(
-              ServerParameterSet::getGlobal(),
-              "diagnosticDataCollectionPeriodMillis",
-              &localPeriodMillis) {}
-
-    virtual Status validate(const std::int32_t& potentialNewValue) {
-        if (potentialNewValue < 100) {
-            return Status(
-                ErrorCodes::BadValue,
-                "diagnosticDataCollectionPeriodMillis must be greater than or equal to 100ms");
-        }
-
-        auto controller = getGlobalFTDCController();
-        if (controller) {
-            controller->setPeriod(Milliseconds(potentialNewValue));
-        }
-
-        return Status::OK();
-    }
-
-} exportedFTDCPeriodParameter;
-
-// Scale the values down since are defaults are in bytes, but the user interface is MB
-std::atomic<std::int32_t> localMaxDirectorySizeMB(  // NOLINT
-    FTDCConfig::kMaxDirectorySizeBytesDefault / (1024 * 1024));
-
-std::atomic<std::int32_t> localMaxFileSizeMB(FTDCConfig::kMaxFileSizeBytesDefault /  // NOLINT
-                                             (1024 * 1024));
-
-class ExportedFTDCDirectorySizeParameter
-    : public ExportedServerParameter<std::int32_t, ServerParameterType::kStartupAndRuntime> {
-public:
-    ExportedFTDCDirectorySizeParameter()
-        : ExportedServerParameter<std::int32_t, ServerParameterType::kStartupAndRuntime>(
-              ServerParameterSet::getGlobal(),
-              "diagnosticDataCollectionDirectorySizeMB",
-              &localMaxDirectorySizeMB) {}
-
-    virtual Status validate(const std::int32_t& potentialNewValue) {
-        if (potentialNewValue < 10) {
-            return Status(
-                ErrorCodes::BadValue,
-                "diagnosticDataCollectionDirectorySizeMB must be greater than or equal to 10");
-        }
-
-        if (potentialNewValue < localMaxFileSizeMB) {
-            return Status(
-                ErrorCodes::BadValue,
-                str::stream()
-                    << "diagnosticDataCollectionDirectorySizeMB must be greater than or equal to '"
-                    << localMaxFileSizeMB
-                    << "' which is the current value of diagnosticDataCollectionFileSizeMB.");
-        }
-
-        auto controller = getGlobalFTDCController();
-        if (controller) {
-            controller->setMaxDirectorySizeBytes(potentialNewValue * 1024 * 1024);
-        }
-
-        return Status::OK();
-    }
-
-} exportedFTDCDirectorySizeParameter;
-
-class ExportedFTDCFileSizeParameter
-    : public ExportedServerParameter<std::int32_t, ServerParameterType::kStartupAndRuntime> {
-public:
-    ExportedFTDCFileSizeParameter()
-        : ExportedServerParameter<std::int32_t, ServerParameterType::kStartupAndRuntime>(
-              ServerParameterSet::getGlobal(),
-              "diagnosticDataCollectionFileSizeMB",
-              &localMaxFileSizeMB) {}
-
-    virtual Status validate(const std::int32_t& potentialNewValue) {
-        if (potentialNewValue < 1) {
-            return Status(ErrorCodes::BadValue,
-                          "diagnosticDataCollectionFileSizeMB must be greater than or equal to 1");
-        }
-
-        if (potentialNewValue > localMaxDirectorySizeMB) {
-            return Status(
-                ErrorCodes::BadValue,
-                str::stream()
-                    << "diagnosticDataCollectionFileSizeMB must be less than or equal to '"
-                    << localMaxDirectorySizeMB
-                    << "' which is the current value of diagnosticDataCollectionDirectorySizeMB.");
-        }
-
-        auto controller = getGlobalFTDCController();
-        if (controller) {
-            controller->setMaxFileSizeBytes(potentialNewValue * 1024 * 1024);
-        }
-
-        return Status::OK();
-    }
-
-} exportedFTDCFileSizeParameter;
-
-std::atomic<std::int32_t> localMaxSamplesPerArchiveMetricChunk(  // NOLINT
-    FTDCConfig::kMaxSamplesPerArchiveMetricChunkDefault);
-
-class ExportedFTDCArchiveChunkSizeParameter
-    : public ExportedServerParameter<std::int32_t, ServerParameterType::kStartupAndRuntime> {
-public:
-    ExportedFTDCArchiveChunkSizeParameter()
-        : ExportedServerParameter<std::int32_t, ServerParameterType::kStartupAndRuntime>(
-              ServerParameterSet::getGlobal(),
-              "diagnosticDataCollectionSamplesPerChunk",
-              &localMaxSamplesPerArchiveMetricChunk) {}
-
-    virtual Status validate(const std::int32_t& potentialNewValue) {
-        if (potentialNewValue < 2) {
-            return Status(
-                ErrorCodes::BadValue,
-                "diagnosticDataCollectionSamplesPerChunk must be greater than or equal to 2");
-        }
-
-        auto controller = getGlobalFTDCController();
-        if (controller) {
-            controller->setMaxSamplesPerArchiveMetricChunk(potentialNewValue);
-        }
-
-        return Status::OK();
-    }
-
-} exportedFTDCArchiveChunkSizeParameter;
-
-std::atomic<std::int32_t> localMaxSamplesPerInterimMetricChunk(  // NOLINT
-    FTDCConfig::kMaxSamplesPerInterimMetricChunkDefault);
-
-class ExportedFTDCInterimChunkSizeParameter
-    : public ExportedServerParameter<std::int32_t, ServerParameterType::kStartupAndRuntime> {
-public:
-    ExportedFTDCInterimChunkSizeParameter()
-        : ExportedServerParameter<std::int32_t, ServerParameterType::kStartupAndRuntime>(
-              ServerParameterSet::getGlobal(),
-              "diagnosticDataCollectionSamplesPerInterimUpdate",
-              &localMaxSamplesPerInterimMetricChunk) {}
-
-    virtual Status validate(const std::int32_t& potentialNewValue) {
-        if (potentialNewValue < 2) {
-            return Status(ErrorCodes::BadValue,
-                          "diagnosticDataCollectionSamplesPerInterimUpdate must be greater than or "
-                          "equal to 2");
-        }
-
-        auto controller = getGlobalFTDCController();
-        if (controller) {
-            controller->setMaxSamplesPerInterimMetricChunk(potentialNewValue);
-        }
-
-        return Status::OK();
-    }
-
-} exportedFTDCInterimChunkSizeParameter;
-
-class FTDCSimpleInternalCommandCollector final : public FTDCCollectorInterface {
-public:
-    FTDCSimpleInternalCommandCollector(StringData command,
-                                       StringData name,
-                                       StringData ns,
-                                       BSONObj cmdObj)
-        : _name(name.toString()), _ns(ns.toString()), _cmdObj(std::move(cmdObj)) {
-        _command = Command::findCommand(command);
-        invariant(_command);
-    }
-
-    void collect(OperationContext* txn, BSONObjBuilder& builder) override {
-        std::string errmsg;
-
-        bool ret = _command->run(txn, _ns, _cmdObj, 0, errmsg, builder);
-
-        // Some commands return errmsgs when they return false (collstats)
-        // Some commands return bson objs when they return false (replGetStatus)
-        // We append the status as needed to ensure readers of the collected data can check the
-        // status of any individual command.
-        _command->appendCommandStatus(builder, ret, errmsg);
-    }
-
-    std::string name() const override {
-        return _name;
-    }
-
-private:
-    std::string _name;
-    std::string _ns;
-    BSONObj _cmdObj;
-
-    // Not owned
-    Command* _command;
-};
-
-}  // namespace
-
-// Register the FTDC system
-// Note: This must be run before the server parameters are parsed during startup
-// so that the FTDCController is initialized.
-//
-void startFTDC() {
-    boost::filesystem::path dir(storageGlobalParams.dbpath);
-    dir /= "diagnostic.data";
-
-
-    FTDCConfig config;
-    config.period = Milliseconds(localPeriodMillis.load());
-    config.enabled = localEnabledFlag;
-    config.maxFileSizeBytes = localMaxFileSizeMB * 1024 * 1024;
-    config.maxDirectorySizeBytes = localMaxDirectorySizeMB * 1024 * 1024;
-    config.maxSamplesPerArchiveMetricChunk = localMaxSamplesPerArchiveMetricChunk;
-    config.maxSamplesPerInterimMetricChunk = localMaxSamplesPerInterimMetricChunk;
-
-    auto controller = stdx::make_unique<FTDCController>(dir, config);
-
-    // Install periodic collectors
-    // These are collected on the period interval in FTDCConfig.
-    // NOTE: For each command here, there must be an equivalent privilege check in
-    // GetDiagnosticDataCommand
-
-    // CmdServerStatus
-    // The "sharding" section is filtered out because at this time it only consists of strings in
-    // migration status. This section triggers too many schema changes in the serverStatus which
-    // hurt ftdc compression efficiency, because its output varies depending on the list of active
-    // migrations.
-    controller->addPeriodicCollector(stdx::make_unique<FTDCSimpleInternalCommandCollector>(
-        "serverStatus",
-        "serverStatus",
-        "",
-        BSON("serverStatus" << 1 << "tcMalloc" << true << "sharding" << false)));
-
+void registerMongoDCollectors(FTDCController* controller) {
     // These metrics are only collected if replication is enabled
     if (repl::getGlobalReplicationCoordinator()->getReplicationMode() !=
         repl::ReplicationCoordinator::modeNone) {
@@ -335,43 +58,19 @@ void startFTDC() {
                                                                   BSON("collStats"
                                                                        << "oplog.rs")));
     }
-
-    // Install System Metric Collector as a periodic collector
-    installSystemMetricsCollector(controller.get());
-
-    // Install file rotation collectors
-    // These are collected on each file rotation.
-
-    // CmdBuildInfo
-    controller->addOnRotateCollector(stdx::make_unique<FTDCSimpleInternalCommandCollector>(
-        "buildInfo", "buildInfo", "", BSON("buildInfo" << 1)));
-
-    // CmdGetCmdLineOpts
-    controller->addOnRotateCollector(stdx::make_unique<FTDCSimpleInternalCommandCollector>(
-        "getCmdLineOpts", "getCmdLineOpts", "", BSON("getCmdLineOpts" << 1)));
-
-    // HostInfoCmd
-    controller->addOnRotateCollector(stdx::make_unique<FTDCSimpleInternalCommandCollector>(
-        "hostInfo", "hostInfo", "", BSON("hostInfo" << 1)));
-
-    // Install the new controller
-    auto& staticFTDC = getFTDCController(getGlobalServiceContext());
-
-    staticFTDC = std::move(controller);
-
-    staticFTDC->start();
 }
 
-void stopFTDC() {
-    auto controller = getGlobalFTDCController();
+}  // namespace
 
-    if (controller) {
-        controller->stop();
-    }
+void startMongoDFTDC() {
+    boost::filesystem::path dir(storageGlobalParams.dbpath);
+    dir /= kFTDCDefaultDirectory.toString();
+
+    startFTDC(dir, FTDCStartMode::kStart, registerMongoDCollectors);
 }
 
-FTDCController* FTDCController::get(ServiceContext* serviceContext) {
-    return getFTDCController(serviceContext).get();
+void stopMongoDFTDC() {
+    stopFTDC();
 }
 
 }  // namespace mongo
diff --git a/src/mongo/db/ftdc/ftdc_mongod.h b/src/mongo/db/ftdc/ftdc_mongod.h
index 1e4f20b8b17..b4409dde902 100644
--- a/src/mongo/db/ftdc/ftdc_mongod.h
+++ b/src/mongo/db/ftdc/ftdc_mongod.h
@@ -1,30 +1,30 @@
 /**
-* Copyright (C) 2015 MongoDB Inc.
-*
-* This program is free software: you can redistribute it and/or  modify
-* it under the terms of the GNU Affero General Public License, version 3,
-* as published by the Free Software Foundation.
-*
-* This program is distributed in the hope that it will be useful,
-* but WITHOUT ANY WARRANTY; without even the implied warranty of
-* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-* GNU Affero General Public License for more details.
-*
-* You should have received a copy of the GNU Affero General Public License
-* along with this program.  If not, see <http://www.gnu.org/licenses/>.
-*
-* As a special exception, the copyright holders give permission to link the
-* code of portions of this program with the OpenSSL library under certain
-* conditions as described in each individual source file and distribute
-* linked combinations including the program with the OpenSSL library. You
-* must comply with the GNU Affero General Public License in all respects
-* for all of the code used other than as permitted herein. If you modify
-* file(s) with this exception, you may extend this exception to your
-* version of the file(s), but you are not obligated to do so. If you do not
-* wish to do so, delete this exception statement from your version. If you
-* delete this exception statement from all source files in the program,
-* then also delete it in the license file.
-*/
+ * Copyright (C) 2015 MongoDB Inc.
+ *
+ * This program is free software: you can redistribute it and/or  modify
+ * it under the terms of the GNU Affero General Public License, version 3,
+ * as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU Affero General Public License for more details.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * As a special exception, the copyright holders give permission to link the
+ * code of portions of this program with the OpenSSL library under certain
+ * conditions as described in each individual source file and distribute
+ * linked combinations including the program with the OpenSSL library. You
+ * must comply with the GNU Affero General Public License in all respects
+ * for all of the code used other than as permitted herein. If you modify
+ * file(s) with this exception, you may extend this exception to your
+ * version of the file(s), but you are not obligated to do so. If you do not
+ * wish to do so, delete this exception statement from your version. If you
+ * delete this exception statement from all source files in the program,
+ * then also delete it in the license file.
+ */
 
 #pragma once
 
@@ -34,11 +34,11 @@ namespace mongo {
  * Start Full Time Data Capture
  * Starts 1 thread.
  */
-void startFTDC();
+void startMongoDFTDC();
 
 /**
  * Stop Full Time Data Capture
  */
-void stopFTDC();
+void stopMongoDFTDC();
 
 }  // namespace mongo
diff --git a/src/mongo/db/ftdc/ftdc_mongos.cpp b/src/mongo/db/ftdc/ftdc_mongos.cpp
new file mode 100644
index 00000000000..bdacaa7b6e1
--- /dev/null
+++ b/src/mongo/db/ftdc/ftdc_mongos.cpp
@@ -0,0 +1,153 @@
+/**
+ * Copyright (C) 2017 MongoDB Inc.
+ *
+ * This program is free software: you can redistribute it and/or  modify
+ * it under the terms of the GNU Affero General Public License, version 3,
+ * as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU Affero General Public License for more details.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * As a special exception, the copyright holders give permission to link the
+ * code of portions of this program with the OpenSSL library under certain
+ * conditions as described in each individual source file and distribute
+ * linked combinations including the program with the OpenSSL library. You
+ * must comply with the GNU Affero General Public License in all respects
+ * for all of the code used other than as permitted herein. If you modify
+ * file(s) with this exception, you may extend this exception to your
+ * version of the file(s), but you are not obligated to do so. If you do not
+ * wish to do so, delete this exception statement from your version. If you
+ * delete this exception statement from all source files in the program,
+ * then also delete it in the license file.
+ */
+#define MONGO_LOG_DEFAULT_COMPONENT ::mongo::logger::LogComponent::kFTDC
+
+#include "mongo/platform/basic.h"
+
+#include "mongo/db/ftdc/ftdc_mongos.h"
+
+#include <boost/filesystem.hpp>
+
+#include "mongo/db/ftdc/controller.h"
+#include "mongo/db/ftdc/ftdc_server.h"
+#include "mongo/db/server_parameters.h"
+#include "mongo/stdx/thread.h"
+#include "mongo/util/log.h"
+
+namespace mongo {
+
+namespace {
+
+/**
+ * Expose diagnosticDataCollectionDirectoryPath set parameter to specify the MongoS FTDC path.
+ */
+class ExportedFTDCDirectoryPathParameter : public ServerParameter {
+public:
+    ExportedFTDCDirectoryPathParameter()
+        : ServerParameter(ServerParameterSet::getGlobal(),
+                          "diagnosticDataCollectionDirectoryPath",
+                          true,
+                          true) {}
+
+
+    void append(OperationContext* opCtx, BSONObjBuilder& b, const std::string& name) final {
+        stdx::lock_guard<stdx::mutex> guard(_lock);
+        b.append(name, _path.generic_string());
+    }
+
+    Status set(const BSONElement& newValueElement) {
+        if (newValueElement.type() != String) {
+            return Status(ErrorCodes::BadValue,
+                          "diagnosticDataCollectionDirectoryPath only supports type string");
+        }
+
+        std::string str = newValueElement.str();
+        return setFromString(str);
+    }
+
+    Status setFromString(const std::string& str) final {
+        stdx::lock_guard<stdx::mutex> guard(_lock);
+
+        FTDCController* controller = nullptr;
+
+        if (hasGlobalServiceContext()) {
+            controller = FTDCController::get(getGlobalServiceContext());
+        }
+
+        if (controller) {
+            Status s = controller->setDirectory(str);
+            if (!s.isOK()) {
+                return s;
+            }
+        }
+
+        _path = str;
+
+        return Status::OK();
+    }
+
+    boost::filesystem::path getDirectory() {
+        stdx::lock_guard<stdx::mutex> guard(_lock);
+        return _path;
+    }
+
+    void setDirectory(boost::filesystem::path& path) {
+        stdx::lock_guard<stdx::mutex> guard(_lock);
+        _path = path;
+    }
+
+private:
+    // Lock to guard _path
+    stdx::mutex _lock;
+
+    // Directory location of ftdc files, guarded by _lock
+    boost::filesystem::path _path;
+} exportedFTDCDirectoryPathParameter;
+
+void registerMongoSCollectors(FTDCController* controller) {
+    // PoolStats
+    controller->addPeriodicCollector(stdx::make_unique<FTDCSimpleInternalCommandCollector>(
+        "connPoolStats", "connPoolStats", "", BSON("connPoolStats" << 1)));
+}
+
+}  // namespace
+
+void startMongoSFTDC() {
+    // Get the path to use for FTDC:
+    // 1. Check if the user set one.
+    // 2. If not, check if the user has a logpath and derive one.
+    // 3. Otherwise, tell the user FTDC cannot run.
+
+    // Only attempt to enable FTDC if we have a path to log files to.
+    FTDCStartMode startMode = FTDCStartMode::kStart;
+    auto directory = exportedFTDCDirectoryPathParameter.getDirectory();
+
+    if (directory.empty()) {
+        if (serverGlobalParams.logpath.empty()) {
+            warning() << "FTDC is disabled because neither '--logpath' nor set parameter "
+                         "'diagnosticDataCollectionDirectoryPath' are specified.";
+            startMode = FTDCStartMode::kSkipStart;
+        } else {
+            directory = FTDCUtil::getMongoSPath(serverGlobalParams.logpath);
+
+            // Update the server parameter with the computed path.
+            // Note: If the computed FTDC directory conflicts with an existing file, then FTDC will
+            // warn about the conflict, and not startup. It will not terminate MongoS in this
+            // situation.
+            exportedFTDCDirectoryPathParameter.setDirectory(directory);
+        }
+    }
+
+    startFTDC(directory, startMode, registerMongoSCollectors);
+}
+
+void stopMongoSFTDC() {
+    stopFTDC();
+}
+
+}  // namespace mongo
diff --git a/src/mongo/db/ftdc/ftdc_mongos.h b/src/mongo/db/ftdc/ftdc_mongos.h
new file mode 100644
index 00000000000..4d8ff8fc08b
--- /dev/null
+++ b/src/mongo/db/ftdc/ftdc_mongos.h
@@ -0,0 +1,43 @@
+/**
+ * Copyright (C) 2017 MongoDB Inc.
+ *
+ * This program is free software: you can redistribute it and/or  modify
+ * it under the terms of the GNU Affero General Public License, version 3,
+ * as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU Affero General Public License for more details.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * As a special exception, the copyright holders give permission to link the
+ * code of portions of this program with the OpenSSL library under certain
+ * conditions as described in each individual source file and distribute
+ * linked combinations including the program with the OpenSSL library. You
+ * must comply with the GNU Affero General Public License in all respects
+ * for all of the code used other than as permitted herein. If you modify
+ * file(s) with this exception, you may extend this exception to your
+ * version of the file(s), but you are not obligated to do so. If you do not
+ * wish to do so, delete this exception statement from your version. If you
+ * delete this exception statement from all source files in the program,
+ * then also delete it in the license file.
+ */
+
+#pragma once
+
+namespace mongo {
+
+/**
+ * Start Full Time Data Capture
+ */
+void startMongoSFTDC();
+
+/**
+ * Stop Full Time Data Capture
+ */
+void stopMongoSFTDC();
+
+}  // namespace mongo
diff --git a/src/mongo/db/ftdc/ftdc_server.cpp b/src/mongo/db/ftdc/ftdc_server.cpp
new file mode 100644
index 00000000000..167587a8722
--- /dev/null
+++ b/src/mongo/db/ftdc/ftdc_server.cpp
@@ -0,0 +1,350 @@
+/**
+ * Copyright (C) 2017 MongoDB Inc.
+ *
+ * This program is free software: you can redistribute it and/or  modify
+ * it under the terms of the GNU Affero General Public License, version 3,
+ * as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU Affero General Public License for more details.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * As a special exception, the copyright holders give permission to link the
+ * code of portions of this program with the OpenSSL library under certain
+ * conditions as described in each individual source file and distribute
+ * linked combinations including the program with the OpenSSL library. You
+ * must comply with the GNU Affero General Public License in all respects
+ * for all of the code used other than as permitted herein. If you modify
+ * file(s) with this exception, you may extend this exception to your
+ * version of the file(s), but you are not obligated to do so. If you do not
+ * wish to do so, delete this exception statement from your version. If you
+ * delete this exception statement from all source files in the program,
+ * then also delete it in the license file.
+ */
+
+#include "mongo/platform/basic.h"
+
+#include "mongo/db/ftdc/ftdc_server.h"
+
+#include <boost/filesystem.hpp>
+#include <fstream>
+#include <memory>
+
+#include "mongo/base/status.h"
+#include "mongo/bson/bsonobjbuilder.h"
+#include "mongo/db/commands.h"
+#include "mongo/db/ftdc/collector.h"
+#include "mongo/db/ftdc/config.h"
+#include "mongo/db/ftdc/controller.h"
+#include "mongo/db/ftdc/ftdc_system_stats.h"
+#include "mongo/db/jsobj.h"
+#include "mongo/db/server_parameters.h"
+#include "mongo/db/service_context.h"
+#include "mongo/stdx/memory.h"
+
+namespace mongo {
+
+namespace {
+
+const auto getFTDCController = ServiceContext::declareDecoration<std::unique_ptr<FTDCController>>();
+
+FTDCController* getGlobalFTDCController() {
+    if (!hasGlobalServiceContext()) {
+        return nullptr;
+    }
+
+    return getFTDCController(getGlobalServiceContext()).get();
+}
+
+std::atomic<bool> localEnabledFlag(FTDCConfig::kEnabledDefault);  // NOLINT
+
+class ExportedFTDCEnabledParameter
+    : public ExportedServerParameter<bool, ServerParameterType::kStartupAndRuntime> {
+public:
+    ExportedFTDCEnabledParameter()
+        : ExportedServerParameter<bool, ServerParameterType::kStartupAndRuntime>(
+              ServerParameterSet::getGlobal(),
+              "diagnosticDataCollectionEnabled",
+              &localEnabledFlag) {}
+
+    virtual Status validate(const bool& potentialNewValue) {
+        auto controller = getGlobalFTDCController();
+        if (controller) {
+            return controller->setEnabled(potentialNewValue);
+        }
+
+        return Status::OK();
+    }
+
+} exportedFTDCEnabledParameter;
+
+std::atomic<std::int32_t> localPeriodMillis(FTDCConfig::kPeriodMillisDefault);  // NOLINT
+
+class ExportedFTDCPeriodParameter
+    : public ExportedServerParameter<std::int32_t, ServerParameterType::kStartupAndRuntime> {
+public:
+    ExportedFTDCPeriodParameter()
+        : ExportedServerParameter<std::int32_t, ServerParameterType::kStartupAndRuntime>(
+              ServerParameterSet::getGlobal(),
+              "diagnosticDataCollectionPeriodMillis",
+              &localPeriodMillis) {}
+
+    virtual Status validate(const std::int32_t& potentialNewValue) {
+        if (potentialNewValue < 100) {
+            return Status(
+                ErrorCodes::BadValue,
+                "diagnosticDataCollectionPeriodMillis must be greater than or equal to 100ms");
+        }
+
+        auto controller = getGlobalFTDCController();
+        if (controller) {
+            controller->setPeriod(Milliseconds(potentialNewValue));
+        }
+
+        return Status::OK();
+    }
+
+} exportedFTDCPeriodParameter;
+
+// Scale the values down since are defaults are in bytes, but the user interface is MB
+std::atomic<std::int32_t> localMaxDirectorySizeMB(FTDCConfig::kMaxDirectorySizeBytesDefault /
+                                                  (1024 * 1024));  // NOLINT
+
+std::atomic<std::int32_t> localMaxFileSizeMB(FTDCConfig::kMaxFileSizeBytesDefault /
+                                             (1024 * 1024));  // NOLINT
+
+class ExportedFTDCDirectorySizeParameter
+    : public ExportedServerParameter<std::int32_t, ServerParameterType::kStartupAndRuntime> {
+public:
+    ExportedFTDCDirectorySizeParameter()
+        : ExportedServerParameter<std::int32_t, ServerParameterType::kStartupAndRuntime>(
+              ServerParameterSet::getGlobal(),
+              "diagnosticDataCollectionDirectorySizeMB",
+              &localMaxDirectorySizeMB) {}
+
+    virtual Status validate(const std::int32_t& potentialNewValue) {
+        if (potentialNewValue < 10) {
+            return Status(
+                ErrorCodes::BadValue,
+                "diagnosticDataCollectionDirectorySizeMB must be greater than or equal to 10");
+        }
+
+        if (potentialNewValue < localMaxFileSizeMB.load()) {
+            return Status(
+                ErrorCodes::BadValue,
+                str::stream()
+                    << "diagnosticDataCollectionDirectorySizeMB must be greater than or equal to '"
+                    << localMaxFileSizeMB.load()
+                    << "' which is the current value of diagnosticDataCollectionFileSizeMB.");
+        }
+
+        auto controller = getGlobalFTDCController();
+        if (controller) {
+            controller->setMaxDirectorySizeBytes(potentialNewValue * 1024 * 1024);
+        }
+
+        return Status::OK();
+    }
+
+} exportedFTDCDirectorySizeParameter;
+
+class ExportedFTDCFileSizeParameter
+    : public ExportedServerParameter<std::int32_t, ServerParameterType::kStartupAndRuntime> {
+public:
+    ExportedFTDCFileSizeParameter()
+        : ExportedServerParameter<std::int32_t, ServerParameterType::kStartupAndRuntime>(
+              ServerParameterSet::getGlobal(),
+              "diagnosticDataCollectionFileSizeMB",
+              &localMaxFileSizeMB) {}
+
+    virtual Status validate(const std::int32_t& potentialNewValue) {
+        if (potentialNewValue < 1) {
+            return Status(ErrorCodes::BadValue,
+                          "diagnosticDataCollectionFileSizeMB must be greater than or equal to 1");
+        }
+
+        if (potentialNewValue > localMaxDirectorySizeMB.load()) {
+            return Status(
+                ErrorCodes::BadValue,
+                str::stream()
+                    << "diagnosticDataCollectionFileSizeMB must be less than or equal to '"
+                    << localMaxDirectorySizeMB.load()
+                    << "' which is the current value of diagnosticDataCollectionDirectorySizeMB.");
+        }
+
+        auto controller = getGlobalFTDCController();
+        if (controller) {
+            controller->setMaxFileSizeBytes(potentialNewValue * 1024 * 1024);
+        }
+
+        return Status::OK();
+    }
+
+} exportedFTDCFileSizeParameter;
+
+std::atomic<std::int32_t> localMaxSamplesPerArchiveMetricChunk(  // NOLINT
+    FTDCConfig::kMaxSamplesPerArchiveMetricChunkDefault);
+
+class ExportedFTDCArchiveChunkSizeParameter
+    : public ExportedServerParameter<std::int32_t, ServerParameterType::kStartupAndRuntime> {
+public:
+    ExportedFTDCArchiveChunkSizeParameter()
+        : ExportedServerParameter<std::int32_t, ServerParameterType::kStartupAndRuntime>(
+              ServerParameterSet::getGlobal(),
+              "diagnosticDataCollectionSamplesPerChunk",
+              &localMaxSamplesPerArchiveMetricChunk) {}
+
+    virtual Status validate(const std::int32_t& potentialNewValue) {
+        if (potentialNewValue < 2) {
+            return Status(
+                ErrorCodes::BadValue,
+                "diagnosticDataCollectionSamplesPerChunk must be greater than or equal to 2");
+        }
+
+        auto controller = getGlobalFTDCController();
+        if (controller) {
+            controller->setMaxSamplesPerArchiveMetricChunk(potentialNewValue);
+        }
+
+        return Status::OK();
+    }
+
+} exportedFTDCArchiveChunkSizeParameter;
+
+std::atomic<std::int32_t> localMaxSamplesPerInterimMetricChunk(  // NOLINT
+    FTDCConfig::kMaxSamplesPerInterimMetricChunkDefault);
+
+class ExportedFTDCInterimChunkSizeParameter
+    : public ExportedServerParameter<std::int32_t, ServerParameterType::kStartupAndRuntime> {
+public:
+    ExportedFTDCInterimChunkSizeParameter()
+        : ExportedServerParameter<std::int32_t, ServerParameterType::kStartupAndRuntime>(
+              ServerParameterSet::getGlobal(),
+              "diagnosticDataCollectionSamplesPerInterimUpdate",
+              &localMaxSamplesPerInterimMetricChunk) {}
+
+    virtual Status validate(const std::int32_t& potentialNewValue) {
+        if (potentialNewValue < 2) {
+            return Status(ErrorCodes::BadValue,
+                          "diagnosticDataCollectionSamplesPerInterimUpdate must be greater than or "
+                          "equal to 2");
+        }
+
+        auto controller = getGlobalFTDCController();
+        if (controller) {
+            controller->setMaxSamplesPerInterimMetricChunk(potentialNewValue);
+        }
+
+        return Status::OK();
+    }
+
+} exportedFTDCInterimChunkSizeParameter;
+}  // namespace
+
+FTDCSimpleInternalCommandCollector::FTDCSimpleInternalCommandCollector(StringData command,
+                                                                       StringData name,
+                                                                       StringData ns,
+                                                                       BSONObj cmdObj)
+    : _name(name.toString()), _ns(ns.toString()), _cmdObj(std::move(cmdObj)) {
+    _command = Command::findCommand(command);
+    invariant(_command);
+}
+
+void FTDCSimpleInternalCommandCollector::collect(OperationContext* opCtx, BSONObjBuilder& builder) {
+    std::string errmsg;
+
+    bool ret = _command->run(opCtx, _ns, _cmdObj, 0, errmsg, builder);
+
+    // Some commands return errmsgs when they return false (collstats)
+    // Some commands return bson objs when they return false (replGetStatus)
+    // We append the status as needed to ensure readers of the collected data can check the
+    // status of any individual command.
+    _command->appendCommandStatus(builder, ret, errmsg);
+}
+
+std::string FTDCSimpleInternalCommandCollector::name() const {
+    return _name;
+}
+
+// Register the FTDC system
+// Note: This must be run before the server parameters are parsed during startup
+// so that the FTDCController is initialized.
+//
+void startFTDC(boost::filesystem::path& path,
+               FTDCStartMode startupMode,
+               RegisterCollectorsFunction registerCollectors) {
+    FTDCConfig config;
+    config.period = Milliseconds(localPeriodMillis.load());
+    // Only enable FTDC if our caller says to enable FTDC, MongoS may not have a valid path to write
+    // files to so update the diagnosticDataCollectionEnabled set parameter to reflect that.
+    localEnabledFlag.store(startupMode == FTDCStartMode::kStart && localEnabledFlag.load());
+    config.enabled = localEnabledFlag.load();
+    config.maxFileSizeBytes = localMaxFileSizeMB.load() * 1024 * 1024;
+    config.maxDirectorySizeBytes = localMaxDirectorySizeMB.load() * 1024 * 1024;
+    config.maxSamplesPerArchiveMetricChunk = localMaxSamplesPerArchiveMetricChunk.load();
+    config.maxSamplesPerInterimMetricChunk = localMaxSamplesPerInterimMetricChunk.load();
+
+    auto controller = stdx::make_unique<FTDCController>(path, config);
+
+    // Install periodic collectors
+    // These are collected on the period interval in FTDCConfig.
+    // NOTE: For each command here, there must be an equivalent privilege check in
+    // GetDiagnosticDataCommand
+
+    // CmdServerStatus
+    // The "sharding" section is filtered out because at this time it only consists of strings in
+    // migration status. This section triggers too many schema changes in the serverStatus which
+    // hurt ftdc compression efficiency, because its output varies depending on the list of active
+    // migrations.
+    // TODO: do we need to enable "sharding" on MongoS?
+    controller->addPeriodicCollector(stdx::make_unique<FTDCSimpleInternalCommandCollector>(
+        "serverStatus",
+        "serverStatus",
+        "",
+        BSON("serverStatus" << 1 << "tcMalloc" << true << "sharding" << false)));
+
+    registerCollectors(controller.get());
+
+    // Install System Metric Collector as a periodic collector
+    installSystemMetricsCollector(controller.get());
+
+    // Install file rotation collectors
+    // These are collected on each file rotation.
+
+    // CmdBuildInfo
+    controller->addOnRotateCollector(stdx::make_unique<FTDCSimpleInternalCommandCollector>(
+        "buildInfo", "buildInfo", "", BSON("buildInfo" << 1)));
+
+    // CmdGetCmdLineOpts
+    controller->addOnRotateCollector(stdx::make_unique<FTDCSimpleInternalCommandCollector>(
+        "getCmdLineOpts", "getCmdLineOpts", "", BSON("getCmdLineOpts" << 1)));
+
+    // HostInfoCmd
+    controller->addOnRotateCollector(stdx::make_unique<FTDCSimpleInternalCommandCollector>(
+        "hostInfo", "hostInfo", "", BSON("hostInfo" << 1)));
+
+    // Install the new controller
+    auto& staticFTDC = getFTDCController(getGlobalServiceContext());
+
+    staticFTDC = std::move(controller);
+
+    staticFTDC->start();
+}
+
+void stopFTDC() {
+    auto controller = getGlobalFTDCController();
+
+    if (controller) {
+        controller->stop();
+    }
+}
+
+FTDCController* FTDCController::get(ServiceContext* serviceContext) {
+    return getFTDCController(serviceContext).get();
+}
+
+}  // namespace mongo
diff --git a/src/mongo/db/ftdc/ftdc_server.h b/src/mongo/db/ftdc/ftdc_server.h
new file mode 100644
index 00000000000..4b4583ae153
--- /dev/null
+++ b/src/mongo/db/ftdc/ftdc_server.h
@@ -0,0 +1,103 @@
+/**
+ * Copyright (C) 2017 MongoDB Inc.
+ *
+ * This program is free software: you can redistribute it and/or  modify
+ * it under the terms of the GNU Affero General Public License, version 3,
+ * as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU Affero General Public License for more details.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * As a special exception, the copyright holders give permission to link the
+ * code of portions of this program with the OpenSSL library under certain
+ * conditions as described in each individual source file and distribute
+ * linked combinations including the program with the OpenSSL library. You
+ * must comply with the GNU Affero General Public License in all respects
+ * for all of the code used other than as permitted herein. If you modify
+ * file(s) with this exception, you may extend this exception to your
+ * version of the file(s), but you are not obligated to do so. If you do not
+ * wish to do so, delete this exception statement from your version. If you
+ * delete this exception statement from all source files in the program,
+ * then also delete it in the license file.
+ */
+
+#pragma once
+
+#include <string>
+
+#include "mongo/bson/bsonobjbuilder.h"
+#include "mongo/db/commands.h"
+#include "mongo/db/ftdc/collector.h"
+#include "mongo/db/ftdc/controller.h"
+#include "mongo/db/jsobj.h"
+#include "mongo/db/operation_context.h"
+#include "mongo/stdx/functional.h"
+
+namespace mongo {
+
+/**
+ * Function that allows FTDC server components to register their own collectors as needed.
+ */
+using RegisterCollectorsFunction = stdx::function<void(FTDCController*)>;
+
+/**
+ * An enum that decides whether FTDC will startup as part of startup or if its deferred to later.
+ */
+enum class FTDCStartMode {
+
+    /**
+     * Skip starting FTDC since it missing a file storage location.
+     */
+    kSkipStart,
+
+    /**
+     * Start FTDC because it has a path to store files.
+     */
+    kStart,
+};
+
+/**
+ * Start Full Time Data Capture
+ * Starts 1 thread.
+ *
+ * See MongoD and MongoS specific functions.
+ */
+void startFTDC(boost::filesystem::path& path,
+               FTDCStartMode startupMode,
+               RegisterCollectorsFunction registerCollectors);
+
+/**
+ * Stop Full Time Data Capture
+ *
+ * See MongoD and MongoS specific functions.
+ */
+void stopFTDC();
+
+/**
+ * A simple FTDC Collector that runs Commands.
+ */
+class FTDCSimpleInternalCommandCollector final : public FTDCCollectorInterface {
+public:
+    FTDCSimpleInternalCommandCollector(StringData command,
+                                       StringData name,
+                                       StringData ns,
+                                       BSONObj cmdObj);
+
+    void collect(OperationContext* opCtx, BSONObjBuilder& builder) override;
+    std::string name() const override;
+
+private:
+    std::string _name;
+    std::string _ns;
+    BSONObj _cmdObj;
+
+    // Not owned
+    Command* _command;
+};
+
+}  // namespace mongo
diff --git a/src/mongo/db/ftdc/ftdc_system_stats.h b/src/mongo/db/ftdc/ftdc_system_stats.h
index 20d21ef4f39..97bc0a67cf5 100644
--- a/src/mongo/db/ftdc/ftdc_system_stats.h
+++ b/src/mongo/db/ftdc/ftdc_system_stats.h
@@ -25,6 +25,7 @@
  * delete this exception statement from all source files in the program,
  * then also delete it in the license file.
  */
+#pragma once
 
 #include <string>
 
diff --git a/src/mongo/db/ftdc/util.cpp b/src/mongo/db/ftdc/util.cpp
index d56eb8ca380..243ba90b666 100644
--- a/src/mongo/db/ftdc/util.cpp
+++ b/src/mongo/db/ftdc/util.cpp
@@ -103,6 +103,19 @@ Date_t roundTime(Date_t now, Milliseconds period) {
     return Date_t::fromMillisSinceEpoch(next_time);
 }
 
+boost::filesystem::path getMongoSPath(const boost::filesystem::path& logFile) {
+    auto base = logFile;
+
+    // Keep stripping file extensions until we are only left with the file name
+    while (base.has_extension()) {
+        auto full_path = base.generic_string();
+        base = full_path.substr(0, full_path.size() - base.extension().size());
+    }
+
+    base += "." + kFTDCDefaultDirectory.toString();
+    return base;
+}
+
 }  // namespace FTDCUtil
 
 
diff --git a/src/mongo/db/ftdc/util.h b/src/mongo/db/ftdc/util.h
index 0cb4d19ce7f..4816c534f96 100644
--- a/src/mongo/db/ftdc/util.h
+++ b/src/mongo/db/ftdc/util.h
@@ -190,6 +190,11 @@ boost::filesystem::path getInterimTempFile(const boost::filesystem::path& file);
  */
 Date_t roundTime(Date_t now, Milliseconds period);
 
+/**
+ * Get the storage path for MongoS from the log file path.
+ */
+boost::filesystem::path getMongoSPath(const boost::filesystem::path& logFile);
+
 }  // namespace FTDCUtil
 
 }  // namespace mongo
diff --git a/src/mongo/db/ftdc/util_test.cpp b/src/mongo/db/ftdc/util_test.cpp
index aa99006c73a..0adf0f73c15 100644
--- a/src/mongo/db/ftdc/util_test.cpp
+++ b/src/mongo/db/ftdc/util_test.cpp
@@ -46,4 +46,25 @@ TEST(FTDCUtilTest, TestRoundTime) {
     checkTime(14, 13, 7);
 }
 
+// Validate the MongoS FTDC path is computed correctly from a log file path.
+TEST(FTDCUtilTest, TestMongoSPath) {
+
+    std::vector<std::pair<std::string, std::string>> testCases = {
+        {"/var/log/mongos.log", "/var/log/mongos.diagnostic.data"},
+        {"/var/log/mongos.foo.log", "/var/log/mongos.diagnostic.data"},
+        {"/var/log/log_file", "/var/log/log_file.diagnostic.data"},
+        {"./mongos.log", "./mongos.diagnostic.data"},
+        {"../mongos.log", "../mongos.diagnostic.data"},
+        {"c:\\var\\log\\mongos.log", "c:\\var\\log\\mongos.diagnostic.data"},
+        {"c:\\var\\log\\mongos.foo.log", "c:\\var\\log\\mongos.diagnostic.data"},
+        {"c:\\var\\log\\log_file", "c:\\var\\log\\log_file.diagnostic.data"},
+        {"/var/some.log/mongos.log", "/var/some.log/mongos.diagnostic.data"},
+        {"/var/some.log/log_file", "/var/some.log/log_file.diagnostic.data"},
+    };
+
+    for (const auto& p : testCases) {
+        ASSERT_EQUALS(FTDCUtil::getMongoSPath(p.first), p.second);
+    }
+}
+
 }  // namespace mongo
diff --git a/src/mongo/db/fts/unicode/string.cpp b/src/mongo/db/fts/unicode/string.cpp
index 10737acc3ed..8c1890f3669 100644
--- a/src/mongo/db/fts/unicode/string.cpp
+++ b/src/mongo/db/fts/unicode/string.cpp
@@ -30,6 +30,7 @@
 
 #include <algorithm>
 #include <boost/algorithm/searching/boyer_moore.hpp>
+#include <boost/version.hpp>
 
 #include "mongo/db/fts/unicode/byte_vector.h"
 #include "mongo/platform/bits.h"
@@ -272,9 +273,15 @@ bool String::substrMatch(const std::string& str,
     auto haystack = caseFoldAndStripDiacritics(&haystackBuf, str, options, cfMode);
     auto needle = caseFoldAndStripDiacritics(&needleBuf, find, options, cfMode);
 
-    // Case sensitive and diacritic sensitive.
+// Case sensitive and diacritic sensitive.
+#if BOOST_VERSION < 106200
     return boost::algorithm::boyer_moore_search(
                haystack.begin(), haystack.end(), needle.begin(), needle.end()) != haystack.end();
+#else
+    return boost::algorithm::boyer_moore_search(
+               haystack.begin(), haystack.end(), needle.begin(), needle.end()) !=
+        std::make_pair(haystack.end(), haystack.end());
+#endif
 }
 
 }  // namespace unicode
diff --git a/src/mongo/db/index/expression_keys_private.cpp b/src/mongo/db/index/expression_keys_private.cpp
index d58c5ce1dc1..9fe3365f393 100644
--- a/src/mongo/db/index/expression_keys_private.cpp
+++ b/src/mongo/db/index/expression_keys_private.cpp
@@ -32,6 +32,7 @@
 
 #include <utility>
 
+#include "mongo/bson/bsonelement_comparator_interface.h"
 #include "mongo/bson/simple_bsonobj_comparator.h"
 #include "mongo/db/bson/dotted_path_support.h"
 #include "mongo/db/field_ref.h"
@@ -247,7 +248,7 @@ void ExpressionKeysPrivate::get2DKeys(const BSONObj& obj,
                                       const TwoDIndexingParams& params,
                                       BSONObjSet* keys,
                                       std::vector<BSONObj>* locs) {
-    BSONElementMSet bSet;
+    BSONElementMultiSet bSet;
 
     // Get all the nested location fields, but don't return individual elements from
     // the last array, if it exists.
@@ -256,7 +257,7 @@ void ExpressionKeysPrivate::get2DKeys(const BSONObj& obj,
     if (bSet.empty())
         return;
 
-    for (BSONElementMSet::iterator setI = bSet.begin(); setI != bSet.end(); ++setI) {
+    for (BSONElementMultiSet::iterator setI = bSet.begin(); setI != bSet.end(); ++setI) {
         BSONElement geo = *setI;
 
         if (geo.eoo() || !geo.isABSONObj())
diff --git a/src/mongo/db/instance.cpp b/src/mongo/db/instance.cpp
index c10d0a3174b..5bbf3339a37 100644
--- a/src/mongo/db/instance.cpp
+++ b/src/mongo/db/instance.cpp
@@ -206,14 +206,15 @@ void receivedCommand(OperationContext* txn,
 
     const int32_t responseToMsgId = message.header().getId();
 
-    DbMessage dbMessage(message);
-    QueryMessage queryMessage(dbMessage);
-
     CurOp* op = CurOp::get(txn);
-
     rpc::LegacyReplyBuilder builder{};
 
     try {
+        DbMessage dbMessage(message);
+
+        // Can throw, so make sure it's under the try statement.
+        QueryMessage queryMessage(dbMessage);
+
         // This will throw if the request is on an invalid namespace.
         rpc::LegacyRequest request{&message};
         // Auth checking for Commands happens later.
diff --git a/src/mongo/db/log_process_details.cpp b/src/mongo/db/log_process_details.cpp
index 125106fd331..ea6e9d945a8 100644
--- a/src/mongo/db/log_process_details.cpp
+++ b/src/mongo/db/log_process_details.cpp
@@ -34,6 +34,9 @@
 
 #include "mongo/db/log_process_details.h"
 
+#include "mongo/db/repl/repl_set_config.h"
+#include "mongo/db/repl/replication_coordinator.h"
+#include "mongo/db/repl/replication_coordinator_global.h"
 #include "mongo/db/server_options.h"
 #include "mongo/db/server_options_helpers.h"
 #include "mongo/util/log.h"
@@ -51,6 +54,7 @@ void logProcessDetails() {
     auto&& vii = VersionInfoInterface::instance();
     log() << mongodVersion(vii);
     vii.logBuildInfo();
+
     printCommandLineOpts();
 }
 
@@ -59,6 +63,19 @@ void logProcessDetailsForLogRotate() {
           << (is32bit() ? " 32" : " 64") << "-bit "
           << "host=" << getHostNameCached();
 
+    auto replCoord = repl::getGlobalReplicationCoordinator();
+    if (replCoord != nullptr &&
+        replCoord->getReplicationMode() == repl::ReplicationCoordinator::modeReplSet) {
+        auto rsConfig = replCoord->getConfig();
+
+        if (rsConfig.isInitialized()) {
+            log() << "Replica Set Config: " << rsConfig.toBSON();
+            log() << "Replica Set Member State: " << (replCoord->getMemberState()).toString();
+        } else {
+            log() << "Node currently has no Replica Set Config.";
+        }
+    }
+
     logProcessDetails();
 }
 
diff --git a/src/mongo/db/matcher/expression_leaf.cpp b/src/mongo/db/matcher/expression_leaf.cpp
index d562bff141e..0bf8d8685bc 100644
--- a/src/mongo/db/matcher/expression_leaf.cpp
+++ b/src/mongo/db/matcher/expression_leaf.cpp
@@ -202,7 +202,7 @@ void ComparisonMatchExpression::debugString(StringBuilder& debug, int level) con
 }
 
 void ComparisonMatchExpression::serialize(BSONObjBuilder* out) const {
-    string opString = "";
+    std::string opString = "";
     switch (matchType()) {
         case LT:
             opString = "$lt";
@@ -656,29 +656,31 @@ bool InMatchExpression::equivalent(const MatchExpression* other) const {
 
 void InMatchExpression::_doSetCollator(const CollatorInterface* collator) {
     _collator = collator;
+    _eltCmp = BSONElementComparator(BSONElementComparator::FieldNamesMode::kIgnore, _collator);
 
     // We need to re-compute '_equalitySet', since our set comparator has changed.
-    BSONElementSet equalitiesWithNewComparator(
-        _originalEqualityVector.begin(), _originalEqualityVector.end(), collator);
-    _equalitySet = std::move(equalitiesWithNewComparator);
+    _equalitySet = _eltCmp.makeBSONEltFlatSet(_originalEqualityVector);
 }
 
-Status InMatchExpression::addEquality(const BSONElement& elt) {
-    if (elt.type() == BSONType::RegEx) {
-        return Status(ErrorCodes::BadValue, "InMatchExpression equality cannot be a regex");
-    }
-    if (elt.type() == BSONType::Undefined) {
-        return Status(ErrorCodes::BadValue, "InMatchExpression equality cannot be undefined");
-    }
+Status InMatchExpression::setEqualities(std::vector<BSONElement> equalities) {
+    for (auto&& equality : equalities) {
+        if (equality.type() == BSONType::RegEx) {
+            return Status(ErrorCodes::BadValue, "InMatchExpression equality cannot be a regex");
+        }
+        if (equality.type() == BSONType::Undefined) {
+            return Status(ErrorCodes::BadValue, "InMatchExpression equality cannot be undefined");
+        }
 
-    if (elt.type() == BSONType::jstNULL) {
-        _hasNull = true;
-    }
-    if (elt.type() == BSONType::Array && elt.Obj().isEmpty()) {
-        _hasEmptyArray = true;
+        if (equality.type() == BSONType::jstNULL) {
+            _hasNull = true;
+        } else if (equality.type() == BSONType::Array && equality.Obj().isEmpty()) {
+            _hasEmptyArray = true;
+        }
     }
-    _equalitySet.insert(elt);
-    _originalEqualityVector.push_back(elt);
+    _originalEqualityVector = std::move(equalities);
+
+    _equalitySet = _eltCmp.makeBSONEltFlatSet(_originalEqualityVector);
+
     return Status::OK();
 }
 
@@ -884,7 +886,7 @@ void BitTestMatchExpression::debugString(StringBuilder& debug, int level) const
 }
 
 void BitTestMatchExpression::serialize(BSONObjBuilder* out) const {
-    string opString = "";
+    std::string opString = "";
 
     switch (matchType()) {
         case BITS_ALL_SET:
diff --git a/src/mongo/db/matcher/expression_leaf.h b/src/mongo/db/matcher/expression_leaf.h
index 128d3070737..5761f2ca2a2 100644
--- a/src/mongo/db/matcher/expression_leaf.h
+++ b/src/mongo/db/matcher/expression_leaf.h
@@ -30,9 +30,11 @@
 
 #pragma once
 
+#include "mongo/bson/bsonelement_comparator.h"
 #include "mongo/bson/bsonmisc.h"
 #include "mongo/bson/bsonobj.h"
 #include "mongo/db/matcher/expression.h"
+#include "mongo/db/query/collation/collator_interface.h"
 #include "mongo/stdx/memory.h"
 #include "mongo/stdx/unordered_map.h"
 
@@ -330,7 +332,10 @@ public:
  */
 class InMatchExpression : public LeafMatchExpression {
 public:
-    InMatchExpression() : LeafMatchExpression(MATCH_IN) {}
+    InMatchExpression()
+        : LeafMatchExpression(MATCH_IN),
+          _eltCmp(BSONElementComparator::FieldNamesMode::kIgnore, _collator),
+          _equalitySet(_eltCmp.makeBSONEltFlatSet(_originalEqualityVector)) {}
 
     Status init(StringData path);
 
@@ -349,11 +354,11 @@ public:
      */
     virtual void _doSetCollator(const CollatorInterface* collator);
 
-    Status addEquality(const BSONElement& elt);
+    Status setEqualities(std::vector<BSONElement> equalities);
 
     Status addRegex(std::unique_ptr<RegexMatchExpression> expr);
 
-    const BSONElementSet& getEqualities() const {
+    const BSONEltFlatSet& getEqualities() const {
         return _equalitySet;
     }
 
@@ -380,17 +385,20 @@ private:
     // Whether or not '_equalities' has an empty array element in it.
     bool _hasEmptyArray = false;
 
-    // Collator used to compare elements. By default, simple binary comparison will be used.
+    // Collator used to construct '_eltCmp';
     const CollatorInterface* _collator = nullptr;
 
-    // Set of equality elements associated with this expression. '_collator' is used as a comparator
-    // for this set.
-    BSONElementSet _equalitySet;
+    // Comparator used to compare elements. By default, simple binary comparison will be used.
+    BSONElementComparator _eltCmp;
 
     // Original container of equality elements, including duplicates. Needed for re-computing
     // '_equalitySet' in case '_collator' changes after elements have been added.
     std::vector<BSONElement> _originalEqualityVector;
 
+    // Set of equality elements associated with this expression. '_eltCmp' is used as a comparator
+    // for this set.
+    BSONEltFlatSet _equalitySet;
+
     // Container of regex elements this object owns.
     std::vector<std::unique_ptr<RegexMatchExpression>> _regexes;
 };
diff --git a/src/mongo/db/matcher/expression_leaf_test.cpp b/src/mongo/db/matcher/expression_leaf_test.cpp
index 293ed59c85c..1c3f7795a95 100644
--- a/src/mongo/db/matcher/expression_leaf_test.cpp
+++ b/src/mongo/db/matcher/expression_leaf_test.cpp
@@ -1501,7 +1501,8 @@ TEST(InMatchExpression, MatchesElementSingle) {
     BSONObj match = BSON("a" << 1);
     BSONObj notMatch = BSON("a" << 2);
     InMatchExpression in;
-    in.addEquality(operand.firstElement());
+    std::vector<BSONElement> equalities{operand.firstElement()};
+    ASSERT_OK(in.setEqualities(std::move(equalities)));
     ASSERT(in.matchesSingleElement(match["a"]));
     ASSERT(!in.matchesSingleElement(notMatch["a"]));
 }
@@ -1519,10 +1520,8 @@ TEST(InMatchExpression, MatchesEmpty) {
 TEST(InMatchExpression, MatchesElementMultiple) {
     BSONObj operand = BSON_ARRAY(1 << "r" << true << 1);
     InMatchExpression in;
-    in.addEquality(operand[0]);
-    in.addEquality(operand[1]);
-    in.addEquality(operand[2]);
-    in.addEquality(operand[3]);
+    std::vector<BSONElement> equalities{operand[0], operand[1], operand[2], operand[3]};
+    ASSERT_OK(in.setEqualities(std::move(equalities)));
 
     BSONObj matchFirst = BSON("a" << 1);
     BSONObj matchSecond = BSON("a"
@@ -1540,7 +1539,8 @@ TEST(InMatchExpression, MatchesScalar) {
     BSONObj operand = BSON_ARRAY(5);
     InMatchExpression in;
     in.init("a");
-    in.addEquality(operand.firstElement());
+    std::vector<BSONElement> equalities{operand.firstElement()};
+    ASSERT_OK(in.setEqualities(std::move(equalities)));
 
     ASSERT(in.matchesBSON(BSON("a" << 5.0), NULL));
     ASSERT(!in.matchesBSON(BSON("a" << 4), NULL));
@@ -1550,7 +1550,8 @@ TEST(InMatchExpression, MatchesArrayValue) {
     BSONObj operand = BSON_ARRAY(5);
     InMatchExpression in;
     in.init("a");
-    in.addEquality(operand.firstElement());
+    std::vector<BSONElement> equalities{operand.firstElement()};
+    ASSERT_OK(in.setEqualities(std::move(equalities)));
 
     ASSERT(in.matchesBSON(BSON("a" << BSON_ARRAY(5.0 << 6)), NULL));
     ASSERT(!in.matchesBSON(BSON("a" << BSON_ARRAY(6 << 7)), NULL));
@@ -1562,7 +1563,8 @@ TEST(InMatchExpression, MatchesNull) {
 
     InMatchExpression in;
     in.init("a");
-    in.addEquality(operand.firstElement());
+    std::vector<BSONElement> equalities{operand.firstElement()};
+    ASSERT_OK(in.setEqualities(std::move(equalities)));
 
     ASSERT(in.matchesBSON(BSONObj(), NULL));
     ASSERT(in.matchesBSON(BSON("a" << BSONNULL), NULL));
@@ -1576,15 +1578,16 @@ TEST(InMatchExpression, MatchesUndefined) {
 
     InMatchExpression in;
     in.init("a");
-    Status s = in.addEquality(operand.firstElement());
-    ASSERT_NOT_OK(s);
+    std::vector<BSONElement> equalities{operand.firstElement()};
+    ASSERT_NOT_OK(in.setEqualities(std::move(equalities)));
 }
 
 TEST(InMatchExpression, MatchesMinKey) {
     BSONObj operand = BSON_ARRAY(MinKey);
     InMatchExpression in;
     in.init("a");
-    in.addEquality(operand.firstElement());
+    std::vector<BSONElement> equalities{operand.firstElement()};
+    ASSERT_OK(in.setEqualities(std::move(equalities)));
 
     ASSERT(in.matchesBSON(BSON("a" << MinKey), NULL));
     ASSERT(!in.matchesBSON(BSON("a" << MaxKey), NULL));
@@ -1595,7 +1598,8 @@ TEST(InMatchExpression, MatchesMaxKey) {
     BSONObj operand = BSON_ARRAY(MaxKey);
     InMatchExpression in;
     in.init("a");
-    in.addEquality(operand.firstElement());
+    std::vector<BSONElement> equalities{operand.firstElement()};
+    ASSERT_OK(in.setEqualities(std::move(equalities)));
 
     ASSERT(in.matchesBSON(BSON("a" << MaxKey), NULL));
     ASSERT(!in.matchesBSON(BSON("a" << MinKey), NULL));
@@ -1606,9 +1610,8 @@ TEST(InMatchExpression, MatchesFullArray) {
     BSONObj operand = BSON_ARRAY(BSON_ARRAY(1 << 2) << 4 << 5);
     InMatchExpression in;
     in.init("a");
-    in.addEquality(operand[0]);
-    in.addEquality(operand[1]);
-    in.addEquality(operand[2]);
+    std::vector<BSONElement> equalities{operand[0], operand[1], operand[2]};
+    ASSERT_OK(in.setEqualities(std::move(equalities)));
 
     ASSERT(in.matchesBSON(BSON("a" << BSON_ARRAY(1 << 2)), NULL));
     ASSERT(!in.matchesBSON(BSON("a" << BSON_ARRAY(1 << 2 << 3)), NULL));
@@ -1620,8 +1623,8 @@ TEST(InMatchExpression, ElemMatchKey) {
     BSONObj operand = BSON_ARRAY(5 << 2);
     InMatchExpression in;
     in.init("a");
-    in.addEquality(operand[0]);
-    in.addEquality(operand[1]);
+    std::vector<BSONElement> equalities{operand[0], operand[1]};
+    ASSERT_OK(in.setEqualities(std::move(equalities)));
 
     MatchDetails details;
     details.requestElemMatchKey();
@@ -1639,7 +1642,8 @@ TEST(InMatchExpression, InMatchExpressionsWithDifferentNumbersOfElementsAreUnequ
                        << "string");
     InMatchExpression eq1;
     InMatchExpression eq2;
-    eq1.addEquality(obj.firstElement());
+    std::vector<BSONElement> equalities{obj.firstElement()};
+    ASSERT_OK(eq1.setEqualities(std::move(equalities)));
     ASSERT(!eq1.equivalent(&eq2));
 }
 
@@ -1675,8 +1679,12 @@ TEST(InMatchExpression, InMatchExpressionsWithCollationEquivalentElementsAreEqua
     InMatchExpression eq2;
     eq2.setCollator(&collator2);
 
-    eq1.addEquality(obj1.firstElement());
-    eq2.addEquality(obj2.firstElement());
+    std::vector<BSONElement> equalities1{obj1.firstElement()};
+    ASSERT_OK(eq1.setEqualities(std::move(equalities1)));
+
+    std::vector<BSONElement> equalities2{obj2.firstElement()};
+    ASSERT_OK(eq2.setEqualities(std::move(equalities2)));
+
     ASSERT(eq1.equivalent(&eq2));
 }
 
@@ -1692,8 +1700,12 @@ TEST(InMatchExpression, InMatchExpressionsWithCollationNonEquivalentElementsAreU
     InMatchExpression eq2;
     eq2.setCollator(&collator2);
 
-    eq1.addEquality(obj1.firstElement());
-    eq2.addEquality(obj2.firstElement());
+    std::vector<BSONElement> equalities1{obj1.firstElement()};
+    ASSERT_OK(eq1.setEqualities(std::move(equalities1)));
+
+    std::vector<BSONElement> equalities2{obj2.firstElement()};
+    ASSERT_OK(eq2.setEqualities(std::move(equalities2)));
+
     ASSERT(!eq1.equivalent(&eq2));
 }
 
@@ -1702,7 +1714,8 @@ TEST(InMatchExpression, StringMatchingWithNullCollatorUsesBinaryComparison) {
     BSONObj notMatch = BSON("a"
                             << "string2");
     InMatchExpression in;
-    in.addEquality(operand.firstElement());
+    std::vector<BSONElement> equalities{operand.firstElement()};
+    ASSERT_OK(in.setEqualities(std::move(equalities)));
     ASSERT(!in.matchesSingleElement(notMatch["a"]));
 }
 
@@ -1713,7 +1726,8 @@ TEST(InMatchExpression, StringMatchingRespectsCollation) {
     CollatorInterfaceMock collator(CollatorInterfaceMock::MockType::kAlwaysEqual);
     InMatchExpression in;
     in.setCollator(&collator);
-    in.addEquality(operand.firstElement());
+    std::vector<BSONElement> equalities{operand.firstElement()};
+    ASSERT_OK(in.setEqualities(std::move(equalities)));
     ASSERT(in.matchesSingleElement(match["a"]));
 }
 
@@ -1726,8 +1740,8 @@ TEST(InMatchExpression, ChangingCollationAfterAddingEqualitiesPreservesEqualitie
     CollatorInterfaceMock collatorReverseString(CollatorInterfaceMock::MockType::kReverseString);
     InMatchExpression in;
     in.setCollator(&collatorAlwaysEqual);
-    in.addEquality(obj1.firstElement());
-    in.addEquality(obj2.firstElement());
+    std::vector<BSONElement> equalities{obj1.firstElement(), obj2.firstElement()};
+    ASSERT_OK(in.setEqualities(std::move(equalities)));
     ASSERT(in.getEqualities().size() == 1);
     in.setCollator(&collatorReverseString);
     ASSERT(in.getEqualities().size() == 2);
diff --git a/src/mongo/db/matcher/expression_parser.cpp b/src/mongo/db/matcher/expression_parser.cpp
index 2d1399edcf9..9bfc2d75f3e 100644
--- a/src/mongo/db/matcher/expression_parser.cpp
+++ b/src/mongo/db/matcher/expression_parser.cpp
@@ -588,6 +588,7 @@ Status MatchExpressionParser::_parseInExpression(InMatchExpression* inExpression
                                                  const BSONObj& theArray,
                                                  const CollatorInterface* collator) {
     inExpression->setCollator(collator);
+    std::vector<BSONElement> equalities;
     BSONObjIterator i(theArray);
     while (i.more()) {
         BSONElement e = i.next();
@@ -606,12 +607,10 @@ Status MatchExpressionParser::_parseInExpression(InMatchExpression* inExpression
             if (!s.isOK())
                 return s;
         } else {
-            Status s = inExpression->addEquality(e);
-            if (!s.isOK())
-                return s;
+            equalities.push_back(e);
         }
     }
-    return Status::OK();
+    return inExpression->setEqualities(std::move(equalities));
 }
 
 StatusWithMatchExpression MatchExpressionParser::_parseType(const char* name,
diff --git a/src/mongo/db/mongod_options.cpp b/src/mongo/db/mongod_options.cpp
index 307d6ace41a..c4b2283ef40 100644
--- a/src/mongo/db/mongod_options.cpp
+++ b/src/mongo/db/mongod_options.cpp
@@ -58,6 +58,7 @@ using std::cout;
 using std::endl;
 using std::string;
 
+
 MongodGlobalParams mongodGlobalParams;
 
 extern DiagLog _diaglog;
@@ -685,14 +686,13 @@ Status validateMongodOptions(const moe::Environment& params) {
             }
         }
 
-        bool isClusterRoleShard = false;
+        bool isClusterRoleShard = params.count("shardsvr");
         if (params.count("sharding.clusterRole")) {
             auto clusterRole = params["sharding.clusterRole"].as<std::string>();
-            isClusterRoleShard = (clusterRole == "shardsvr");
+            isClusterRoleShard = isClusterRoleShard || (clusterRole == "shardsvr");
         }
 
-        if ((isClusterRoleShard || params.count("shardsvr")) &&
-            !params.count("sharding._overrideShardIdentity")) {
+        if (isClusterRoleShard && !params.count("sharding._overrideShardIdentity")) {
             return Status(
                 ErrorCodes::BadValue,
                 "shardsvr cluster role with queryableBackupMode requires _overrideShardIdentity");
@@ -1328,6 +1328,21 @@ Status storeMongodOptions(const moe::Environment& params) {
         log() << endl;
     }
 
+    bool isClusterRoleShard = params.count("shardsvr");
+    bool isClusterRoleConfig = params.count("configsvr");
+    if (params.count("sharding.clusterRole")) {
+        auto clusterRole = params["sharding.clusterRole"].as<std::string>();
+        isClusterRoleShard = isClusterRoleShard || (clusterRole == "shardsvr");
+        isClusterRoleConfig = isClusterRoleConfig || (clusterRole == "configsvr");
+    }
+
+    if ((isClusterRoleShard || isClusterRoleConfig) && skipShardingConfigurationChecks) {
+        auto clusterRoleStr = isClusterRoleConfig ? "--configsvr" : "--shardsvr";
+        return Status(ErrorCodes::BadValue,
+                      str::stream() << "Can not specify " << clusterRoleStr
+                                    << " and set skipShardingConfigurationChecks=true");
+    }
+
     setGlobalReplSettings(replSettings);
     return Status::OK();
 }
diff --git a/src/mongo/db/mongod_options.h b/src/mongo/db/mongod_options.h
index 27fb778cfdf..481edacc96f 100644
--- a/src/mongo/db/mongod_options.h
+++ b/src/mongo/db/mongod_options.h
@@ -50,6 +50,8 @@ struct MongodGlobalParams {
     MongodGlobalParams() : scriptingEnabled(true) {}
 };
 
+extern bool skipShardingConfigurationChecks;
+
 extern MongodGlobalParams mongodGlobalParams;
 
 Status addMongodOptions(moe::OptionSection* options);
diff --git a/src/mongo/db/namespace_string-inl.h b/src/mongo/db/namespace_string-inl.h
index 0ff45fca39e..7ad830f2481 100644
--- a/src/mongo/db/namespace_string-inl.h
+++ b/src/mongo/db/namespace_string-inl.h
@@ -118,7 +118,7 @@ inline bool NamespaceString::validCollectionName(StringData coll) {
     return true;
 }
 
-inline NamespaceString::NamespaceString() : _ns(), _dotIndex(0) {}
+inline NamespaceString::NamespaceString() : _ns(), _dotIndex(std::string::npos) {}
 inline NamespaceString::NamespaceString(StringData nsIn) {
     _ns = nsIn.toString();  // copy to our buffer
     _dotIndex = _ns.find('.');
diff --git a/src/mongo/db/namespace_string_test.cpp b/src/mongo/db/namespace_string_test.cpp
index 69884bd5ca9..69742c797bf 100644
--- a/src/mongo/db/namespace_string_test.cpp
+++ b/src/mongo/db/namespace_string_test.cpp
@@ -272,4 +272,16 @@ TEST(NamespaceStringTest, makeListIndexesNSIsCorrect) {
     ASSERT(ns.isListIndexesCursorNS());
     ASSERT_EQUALS(NamespaceString("DB.COLL"), ns.getTargetNSForListIndexes());
 }
+
+TEST(NamespaceStringTest, EmptyNSStringReturnsEmptyColl) {
+    NamespaceString nss{};
+    ASSERT_TRUE(nss.toString().empty());
+    ASSERT_EQ(nss.coll(), StringData{});
+}
+
+TEST(NamespaceStringTest, EmptyNSStringReturnsEmptyDb) {
+    NamespaceString nss{};
+    ASSERT_TRUE(nss.toString().empty());
+    ASSERT_EQ(nss.db(), StringData{});
+}
 }
diff --git a/src/mongo/db/op_observer_impl.cpp b/src/mongo/db/op_observer_impl.cpp
index 00560851a7a..91bc3317d48 100644
--- a/src/mongo/db/op_observer_impl.cpp
+++ b/src/mongo/db/op_observer_impl.cpp
@@ -42,9 +42,12 @@
 #include "mongo/db/server_options.h"
 #include "mongo/db/views/durable_view_catalog.h"
 #include "mongo/scripting/engine.h"
+#include "mongo/util/fail_point_service.h"
 
 namespace mongo {
 
+MONGO_FP_DECLARE(failCollectionUpdates);
+
 void OpObserverImpl::onCreateIndex(OperationContext* txn,
                                    const std::string& ns,
                                    BSONObj indexDoc,
@@ -80,7 +83,7 @@ void OpObserverImpl::onInserts(OperationContext* txn,
 
     if (nss.ns() == FeatureCompatibilityVersion::kCollection) {
         for (auto it = begin; it != end; it++) {
-            FeatureCompatibilityVersion::onInsertOrUpdate(*it);
+            FeatureCompatibilityVersion::onInsertOrUpdate(txn, *it);
         }
     }
 
@@ -94,6 +97,20 @@ void OpObserverImpl::onInserts(OperationContext* txn,
 }
 
 void OpObserverImpl::onUpdate(OperationContext* txn, const OplogUpdateEntryArgs& args) {
+    MONGO_FAIL_POINT_BLOCK(failCollectionUpdates, extraData) {
+        auto collElem = extraData.getData()["collectionNS"];
+        // If the failpoint specifies no collection or matches the existing one, fail.
+        if (!collElem || args.ns == collElem.String()) {
+            uasserted(40654,
+                      str::stream() << "failCollectionUpdates failpoint enabled, namespace: "
+                                    << args.ns
+                                    << ", update: "
+                                    << args.update
+                                    << " on document with "
+                                    << args.criteria);
+        }
+    }
+
     // Do not log a no-op operation; see SERVER-21738
     if (args.update.isEmpty()) {
         return;
@@ -119,7 +136,7 @@ void OpObserverImpl::onUpdate(OperationContext* txn, const OplogUpdateEntryArgs&
     }
 
     if (args.ns == FeatureCompatibilityVersion::kCollection) {
-        FeatureCompatibilityVersion::onInsertOrUpdate(args.updatedDoc);
+        FeatureCompatibilityVersion::onInsertOrUpdate(txn, args.updatedDoc);
     }
 }
 
@@ -162,7 +179,7 @@ void OpObserverImpl::onDelete(OperationContext* txn,
         DurableViewCatalog::onExternalChange(txn, ns);
     }
     if (ns.ns() == FeatureCompatibilityVersion::kCollection) {
-        FeatureCompatibilityVersion::onDelete(deleteState.idDoc);
+        FeatureCompatibilityVersion::onDelete(txn, deleteState.idDoc);
     }
 }
 
@@ -221,7 +238,7 @@ void OpObserverImpl::onDropDatabase(OperationContext* txn, const std::string& db
     repl::logOp(txn, "c", dbName.c_str(), cmdObj, nullptr, false);
 
     if (NamespaceString(dbName).db() == FeatureCompatibilityVersion::kDatabase) {
-        FeatureCompatibilityVersion::onDropCollection();
+        FeatureCompatibilityVersion::onDropCollection(txn);
     }
 
     getGlobalAuthorizationManager()->logOp(txn, "c", dbName.c_str(), cmdObj, nullptr);
@@ -243,7 +260,7 @@ void OpObserverImpl::onDropCollection(OperationContext* txn,
     }
 
     if (collectionName.ns() == FeatureCompatibilityVersion::kCollection) {
-        FeatureCompatibilityVersion::onDropCollection();
+        FeatureCompatibilityVersion::onDropCollection(txn);
     }
 
     getGlobalAuthorizationManager()->logOp(txn, "c", dbName.c_str(), cmdObj, nullptr);
@@ -276,11 +293,10 @@ void OpObserverImpl::onRenameCollection(OperationContext* txn,
                                 << dropTarget);
 
     repl::logOp(txn, "c", dbName.c_str(), cmdObj, nullptr, false);
-    if (fromCollection.coll() == DurableViewCatalog::viewsCollectionName() ||
-        toCollection.coll() == DurableViewCatalog::viewsCollectionName()) {
-        DurableViewCatalog::onExternalChange(
-            txn, NamespaceString(DurableViewCatalog::viewsCollectionName()));
-    }
+    if (fromCollection.isSystemDotViews())
+        DurableViewCatalog::onExternalChange(txn, fromCollection);
+    if (toCollection.isSystemDotViews())
+        DurableViewCatalog::onExternalChange(txn, toCollection);
 
     getGlobalAuthorizationManager()->logOp(txn, "c", dbName.c_str(), cmdObj, nullptr);
     logOpForDbHash(txn, dbName.c_str());
diff --git a/src/mongo/db/operation_context.cpp b/src/mongo/db/operation_context.cpp
index 8151fe97dfa..41b54eb6611 100644
--- a/src/mongo/db/operation_context.cpp
+++ b/src/mongo/db/operation_context.cpp
@@ -325,7 +325,7 @@ StatusWith<stdx::cv_status> OperationContext::waitForConditionOrInterruptNoAsser
         deadline = std::min(deadline, getDeadline());
     }
 
-    const auto waitStatus = [&] {
+    const auto waitStatus = [&]() -> stdx::cv_status {
         if (Date_t::max() == deadline) {
             cv.wait(m);
             return stdx::cv_status::no_timeout;
diff --git a/src/mongo/db/operation_context.h b/src/mongo/db/operation_context.h
index 2c8be360a77..79455e47309 100644
--- a/src/mongo/db/operation_context.h
+++ b/src/mongo/db/operation_context.h
@@ -202,6 +202,40 @@ public:
     }
 
     /**
+     * Waits on condition "cv" for "pred" until "pred" returns true, or the given "deadline"
+     * expires, or this operation is interrupted, or this operation's own deadline expires.
+     *
+     *
+     * If the operation deadline expires or the operation is interrupted, throws a DBException.  If
+     * the given "deadline" expires, returns pred. Otherwise, returns true.
+     */
+    template <typename Pred>
+    bool waitForConditionOrInterruptUntilPred(stdx::condition_variable& cv,
+                                              stdx::unique_lock<stdx::mutex>& m,
+                                              Date_t deadline,
+                                              Pred pred) {
+        while (!pred()) {
+            if (stdx::cv_status::timeout == waitForConditionOrInterruptUntil(cv, m, deadline)) {
+                return pred();
+            }
+        }
+        return true;
+    }
+
+    /**
+     * Same as the predicate form of waitForConditionOrInterruptUntilPred, but takes a relative
+     * amount of time to wait instead of an absolute time point.
+     */
+    template <typename Pred>
+    bool waitForConditionOrInterruptFor(stdx::condition_variable& cv,
+                                        stdx::unique_lock<stdx::mutex>& m,
+                                        Milliseconds duration,
+                                        Pred pred) {
+        return waitForConditionOrInterruptUntilPred(
+            cv, m, getServiceContext()->getPreciseClockSource()->now() + duration, pred);
+    }
+
+    /**
      * Same as waitForConditionOrInterruptUntil, except returns StatusWith<stdx::cv_status> and
      * non-ok status indicates the error instead of a DBException.
      */
diff --git a/src/mongo/db/operation_context_test.cpp b/src/mongo/db/operation_context_test.cpp
index 06c6575208a..ca77b8c316b 100644
--- a/src/mongo/db/operation_context_test.cpp
+++ b/src/mongo/db/operation_context_test.cpp
@@ -231,7 +231,7 @@ public:
                                                                  Date_t maxTime) {
 
         auto barrier = std::make_shared<unittest::Barrier>(2);
-        auto task = stdx::packaged_task<stdx::cv_status()>([=] {
+        auto task = stdx::packaged_task<stdx::cv_status()>([=]() -> stdx::cv_status {
             if (maxTime < Date_t::max()) {
                 txn->setDeadlineByDate(maxTime);
             }
diff --git a/src/mongo/db/ops/delete.cpp b/src/mongo/db/ops/delete.cpp
index 7f509308ad4..6a7da15ee74 100644
--- a/src/mongo/db/ops/delete.cpp
+++ b/src/mongo/db/ops/delete.cpp
@@ -63,19 +63,11 @@ long long deleteObjects(OperationContext* txn,
     ParsedDelete parsedDelete(txn, &request);
     uassertStatusOK(parsedDelete.parseRequest());
 
-    auto client = txn->getClient();
-    auto lastOpAtOperationStart = repl::ReplClientInfo::forClient(client).getLastOp();
-
     std::unique_ptr<PlanExecutor> exec = uassertStatusOK(
         getExecutorDelete(txn, &CurOp::get(txn)->debug(), collection, &parsedDelete));
 
     uassertStatusOK(exec->executePlan());
 
-    // No-ops need to reset lastOp in the client, for write concern.
-    if (repl::ReplClientInfo::forClient(client).getLastOp() == lastOpAtOperationStart) {
-        repl::ReplClientInfo::forClient(client).setLastOpToSystemLastOpTime(txn);
-    }
-
     return DeleteStage::getNumDeleted(*exec);
 }
 
diff --git a/src/mongo/db/ops/modifier_add_to_set.cpp b/src/mongo/db/ops/modifier_add_to_set.cpp
index a565800d749..416d487e0f4 100644
--- a/src/mongo/db/ops/modifier_add_to_set.cpp
+++ b/src/mongo/db/ops/modifier_add_to_set.cpp
@@ -225,6 +225,8 @@ Status ModifierAddToSet::prepare(mb::Element root, StringData matchedField, Exec
     // Locate the field name in 'root'.
     Status status = pathsupport::findLongestPrefix(
         _fieldRef, root, &_preparedState->idxFound, &_preparedState->elemFound);
+    const auto elemFoundIsArray =
+        _preparedState->elemFound.ok() && _preparedState->elemFound.getType() == BSONType::Array;
 
     // FindLongestPrefix may say the path does not exist at all, which is fine here, or
     // that the path was not viable or otherwise wrong, in which case, the mod cannot
@@ -248,6 +250,12 @@ Status ModifierAddToSet::prepare(mb::Element root, StringData matchedField, Exec
     if (!_preparedState->elemFound.ok() || _preparedState->idxFound < (_fieldRef.numParts() - 1)) {
         // If no target element exists, we will simply be creating a new array.
         _preparedState->addAll = true;
+
+        if (elemFoundIsArray) {
+            // Report that an existing array will gain a new element as a result of this mod.
+            execInfo->indexOfArrayWithNewElement[0] = _preparedState->idxFound;
+        }
+
         return Status::OK();
     }
 
diff --git a/src/mongo/db/ops/modifier_add_to_set_test.cpp b/src/mongo/db/ops/modifier_add_to_set_test.cpp
index 235a65239e6..63d5a7136d4 100644
--- a/src/mongo/db/ops/modifier_add_to_set_test.cpp
+++ b/src/mongo/db/ops/modifier_add_to_set_test.cpp
@@ -459,4 +459,41 @@ TEST(Collation, AddToSetWithEachRespectsCollation) {
 
     ASSERT_EQUALS(doc, fromjson("{ a : ['abc', 'bdc'] }"));
 }
+
+TEST(IndexedMod, PrepareReportCreatedArrayElement) {
+    Document doc(fromjson("{a: [{b: 0}]}"));
+    Mod mod(fromjson("{$addToSet: {'a.1.c': 2}}"));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(mod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.1.c");
+    ASSERT_TRUE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_EQUALS(*execInfo.indexOfArrayWithNewElement[0], 0u);
+    ASSERT_FALSE(execInfo.noOp);
+}
+
+TEST(IndexedMod, PrepareDoNotReportModifiedArrayElement) {
+    Document doc(fromjson("{a: [{b: 0}]}"));
+    Mod mod(fromjson("{$addToSet: {'a.0.c': 2}}"));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(mod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.0.c");
+    ASSERT_FALSE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_FALSE(execInfo.noOp);
+}
+
+TEST(IndexedMod, PrepareDoNotReportCreatedNumericObjectField) {
+    Document doc(fromjson("{a: {'0': {b: 0}}}"));
+    Mod mod(fromjson("{$addToSet: {'a.1.c': 2}}"));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(mod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.1.c");
+    ASSERT_FALSE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_FALSE(execInfo.noOp);
+}
 }  // namespace
diff --git a/src/mongo/db/ops/modifier_bit.cpp b/src/mongo/db/ops/modifier_bit.cpp
index fb7ae3f45eb..99296911e95 100644
--- a/src/mongo/db/ops/modifier_bit.cpp
+++ b/src/mongo/db/ops/modifier_bit.cpp
@@ -164,6 +164,8 @@ Status ModifierBit::prepare(mutablebson::Element root,
     // Locate the field name in 'root'.
     Status status = pathsupport::findLongestPrefix(
         _fieldRef, root, &_preparedState->idxFound, &_preparedState->elemFound);
+    const auto elemFoundIsArray =
+        _preparedState->elemFound.ok() && _preparedState->elemFound.getType() == BSONType::Array;
 
 
     // FindLongestPrefix may say the path does not exist at all, which is fine here, or
@@ -189,6 +191,12 @@ Status ModifierBit::prepare(mutablebson::Element root,
         // If no target element exists, the value we will write is the result of applying
         // the operation to a zero-initialized integer element.
         _preparedState->newValue = apply(SafeNum(static_cast<int32_t>(0)));
+
+        if (elemFoundIsArray) {
+            // Report that an existing array will gain a new element as a result of this mod.
+            execInfo->indexOfArrayWithNewElement[0] = _preparedState->idxFound;
+        }
+
         return Status::OK();
     }
 
diff --git a/src/mongo/db/ops/modifier_bit_test.cpp b/src/mongo/db/ops/modifier_bit_test.cpp
index dbc4d7ba0b2..5b8b0131ea7 100644
--- a/src/mongo/db/ops/modifier_bit_test.cpp
+++ b/src/mongo/db/ops/modifier_bit_test.cpp
@@ -755,4 +755,41 @@ TEST(DbUpdateTests, Bit1_4_Combined) {
     ASSERT_EQUALS(BSON("$set" << BSON("x" << ((3 | 2) & 8))), logDoc);
 }
 
+TEST(IndexedMod, PrepareReportCreatedArrayElement) {
+    Document doc(fromjson("{a: [{b: 0}]}"));
+    Mod mod(fromjson("{$bit: {'a.1.c': {and: NumberInt(1)}}}"));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(mod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.1.c");
+    ASSERT_TRUE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_EQUALS(*execInfo.indexOfArrayWithNewElement[0], 0u);
+    ASSERT_FALSE(execInfo.noOp);
+}
+
+TEST(IndexedMod, PrepareDoNotReportModifiedArrayElement) {
+    Document doc(fromjson("{a: [{b: NumberInt(0)}]}"));
+    Mod mod(fromjson("{$bit: {'a.0.c': {or: NumberInt(1)}}}"));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(mod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.0.c");
+    ASSERT_FALSE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_FALSE(execInfo.noOp);
+}
+
+TEST(IndexedMod, PrepareDoNotReportCreatedNumericObjectField) {
+    Document doc(fromjson("{a: {'0': {b: 0}}}"));
+    Mod mod(fromjson("{$bit: {'a.1.c': {and: NumberInt(1)}}}"));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(mod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.1.c");
+    ASSERT_FALSE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_FALSE(execInfo.noOp);
+}
+
 }  // namespace
diff --git a/src/mongo/db/ops/modifier_compare.cpp b/src/mongo/db/ops/modifier_compare.cpp
index 4e366f90ee2..7db7abd09ca 100644
--- a/src/mongo/db/ops/modifier_compare.cpp
+++ b/src/mongo/db/ops/modifier_compare.cpp
@@ -110,6 +110,8 @@ Status ModifierCompare::prepare(mutablebson::Element root,
     // be created during the apply.
     Status status = pathsupport::findLongestPrefix(
         _updatePath, root, &_preparedState->idxFound, &_preparedState->elemFound);
+    const auto elemFoundIsArray =
+        _preparedState->elemFound.ok() && _preparedState->elemFound.getType() == BSONType::Array;
 
     // FindLongestPrefix may say the path does not exist at all, which is fine here, or
     // that the path was not viable or otherwise wrong, in which case, the mod cannot
@@ -128,6 +130,11 @@ Status ModifierCompare::prepare(mutablebson::Element root,
                              _preparedState->idxFound == (_updatePath.numParts() - 1));
     if (!destExists) {
         execInfo->noOp = false;
+
+        if (elemFoundIsArray) {
+            // Report that an existing array will gain a new element as a result of this mod.
+            execInfo->indexOfArrayWithNewElement[0] = _preparedState->idxFound;
+        }
     } else {
         const int compareVal =
             _preparedState->elemFound.compareWithBSONElement(_val, _collator, false);
diff --git a/src/mongo/db/ops/modifier_compare_test.cpp b/src/mongo/db/ops/modifier_compare_test.cpp
index 9bc036b776b..edaba6a2a68 100644
--- a/src/mongo/db/ops/modifier_compare_test.cpp
+++ b/src/mongo/db/ops/modifier_compare_test.cpp
@@ -341,4 +341,41 @@ TEST(Collation, MaxRespectsCollationFromSetCollator) {
     ASSERT_OK(mod.apply());
     ASSERT_EQUALS(fromjson("{a : 'abd'}"), doc);
 }
+
+TEST(IndexedMod, PrepareReportCreatedArrayElement) {
+    Document doc(fromjson("{a: [{b: 0}]}"));
+    Mod mod(fromjson("{$min: {'a.1.c': 2}}"));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(mod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.1.c");
+    ASSERT_TRUE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_EQUALS(*execInfo.indexOfArrayWithNewElement[0], 0u);
+    ASSERT_FALSE(execInfo.noOp);
+}
+
+TEST(IndexedMod, PrepareDoNotReportModifiedArrayElement) {
+    Document doc(fromjson("{a: [{b: 0}]}"));
+    Mod mod(fromjson("{$min: {'a.0.c': 2}}"));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(mod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.0.c");
+    ASSERT_FALSE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_FALSE(execInfo.noOp);
+}
+
+TEST(IndexedMod, PrepareDoNotReportCreatedNumericObjectField) {
+    Document doc(fromjson("{a: {'0': {b: 0}}}"));
+    Mod mod(fromjson("{$min: {'a.1.c': 2}}"));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(mod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.1.c");
+    ASSERT_FALSE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_FALSE(execInfo.noOp);
+}
 }  // namespace
diff --git a/src/mongo/db/ops/modifier_current_date.cpp b/src/mongo/db/ops/modifier_current_date.cpp
index cd328f5fe94..4496e0cd2cd 100644
--- a/src/mongo/db/ops/modifier_current_date.cpp
+++ b/src/mongo/db/ops/modifier_current_date.cpp
@@ -161,6 +161,8 @@ Status ModifierCurrentDate::prepare(mutablebson::Element root,
     // be created during the apply.
     Status status = pathsupport::findLongestPrefix(
         _updatePath, root, &_preparedState->idxFound, &_preparedState->elemFound);
+    const auto elemFoundIsArray =
+        _preparedState->elemFound.ok() && _preparedState->elemFound.getType() == BSONType::Array;
 
     // FindLongestPrefix may say the path does not exist at all, which is fine here, or
     // that the path was not viable or otherwise wrong, in which case, the mod cannot
@@ -175,6 +177,14 @@ Status ModifierCurrentDate::prepare(mutablebson::Element root,
     // there is any conflict among mods.
     execInfo->fieldRef[0] = &_updatePath;
 
+    if (!_preparedState->elemFound.ok() ||
+        _preparedState->idxFound < (_updatePath.numParts() - 1)) {
+        if (elemFoundIsArray) {
+            // Report that an existing array will gain a new element as a result of this mod.
+            execInfo->indexOfArrayWithNewElement[0] = _preparedState->idxFound;
+        }
+    }
+
     return Status::OK();
 }
 
diff --git a/src/mongo/db/ops/modifier_current_date_test.cpp b/src/mongo/db/ops/modifier_current_date_test.cpp
index 81bcd4d04e0..a3746e445da 100644
--- a/src/mongo/db/ops/modifier_current_date_test.cpp
+++ b/src/mongo/db/ops/modifier_current_date_test.cpp
@@ -359,4 +359,41 @@ TEST(DottedTimestampInput, EmptyStartDoc) {
     validateOplogEntry(oplogFormat, logDoc);
 }
 
+TEST(BoolInput, PrepareReportCreatedArrayElement) {
+    Document doc(fromjson("{a: [{b: 0}]}"));
+    Mod mod(fromjson("{$currentDate: {'a.1.c': true}}"));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(mod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.1.c");
+    ASSERT_TRUE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_EQUALS(*execInfo.indexOfArrayWithNewElement[0], 0u);
+    ASSERT_FALSE(execInfo.noOp);
+}
+
+TEST(BoolInput, PrepareDoNotReportModifiedArrayElement) {
+    Document doc(fromjson("{a: [{b: 0}]}"));
+    Mod mod(fromjson("{$currentDate: {'a.0.c': true}}"));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(mod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.0.c");
+    ASSERT_FALSE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_FALSE(execInfo.noOp);
+}
+
+TEST(BoolInput, PrepareDoNotReportCreatedNumericObjectField) {
+    Document doc(fromjson("{a: {'0': {b: 0}}}"));
+    Mod mod(fromjson("{$currentDate: {'a.1.c': true}}"));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(mod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.1.c");
+    ASSERT_FALSE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_FALSE(execInfo.noOp);
+}
+
 }  // namespace
diff --git a/src/mongo/db/ops/modifier_inc.cpp b/src/mongo/db/ops/modifier_inc.cpp
index 314ac6a5024..f2d51b390da 100644
--- a/src/mongo/db/ops/modifier_inc.cpp
+++ b/src/mongo/db/ops/modifier_inc.cpp
@@ -135,6 +135,8 @@ Status ModifierInc::prepare(mutablebson::Element root,
     // be created during the apply.
     Status status = pathsupport::findLongestPrefix(
         _fieldRef, root, &_preparedState->idxFound, &_preparedState->elemFound);
+    const auto elemFoundIsArray =
+        _preparedState->elemFound.ok() && _preparedState->elemFound.getType() == BSONType::Array;
 
     // FindLongestPrefix may say the path does not exist at all, which is fine here, or
     // that the path was not viable or otherwise wrong, in which case, the mod cannot
@@ -166,6 +168,11 @@ Status ModifierInc::prepare(mutablebson::Element root,
         if (_mode == MODE_MUL)
             _preparedState->newValue *= SafeNum(static_cast<int32_t>(0));
 
+        if (elemFoundIsArray) {
+            // Report that an existing array will gain a new element as a result of this mod.
+            execInfo->indexOfArrayWithNewElement[0] = _preparedState->idxFound;
+        }
+
         return Status::OK();
     }
 
diff --git a/src/mongo/db/ops/modifier_inc_test.cpp b/src/mongo/db/ops/modifier_inc_test.cpp
index 02fbbd4474a..3784abbcdff 100644
--- a/src/mongo/db/ops/modifier_inc_test.cpp
+++ b/src/mongo/db/ops/modifier_inc_test.cpp
@@ -630,4 +630,41 @@ TEST(Multiplication, ApplyMissingElementDouble) {
     ASSERT_EQUALS(mongo::NumberDouble, doc.root().rightChild().getType());
 }
 
+TEST(IndexedMod, PrepareReportCreatedArrayElement) {
+    Document doc(fromjson("{a: [{b: 0}]}"));
+    Mod mod(fromjson("{$inc: {'a.1.c': 2}}"));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(mod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.1.c");
+    ASSERT_TRUE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_EQUALS(*execInfo.indexOfArrayWithNewElement[0], 0u);
+    ASSERT_FALSE(execInfo.noOp);
+}
+
+TEST(IndexedMod, PrepareDoNotReportModifiedArrayElement) {
+    Document doc(fromjson("{a: [{b: 0}]}"));
+    Mod mod(fromjson("{$inc: {'a.0.c': 2}}"));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(mod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.0.c");
+    ASSERT_FALSE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_FALSE(execInfo.noOp);
+}
+
+TEST(IndexedMod, PrepareDoNotReportCreatedNumericObjectField) {
+    Document doc(fromjson("{a: {'0': {b: 0}}}"));
+    Mod mod(fromjson("{$inc: {'a.1.c': 2}}"));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(mod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.1.c");
+    ASSERT_FALSE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_FALSE(execInfo.noOp);
+}
+
 }  // namespace
diff --git a/src/mongo/db/ops/modifier_interface.h b/src/mongo/db/ops/modifier_interface.h
index d833dcf07a1..2294802e18d 100644
--- a/src/mongo/db/ops/modifier_interface.h
+++ b/src/mongo/db/ops/modifier_interface.h
@@ -205,11 +205,17 @@ struct ModifierInterface::ExecInfo {
     ExecInfo() : noOp(false), context(ANY_CONTEXT) {
         for (int i = 0; i < MAX_NUM_FIELDS; i++) {
             fieldRef[i] = NULL;
+            indexOfArrayWithNewElement[i] = boost::none;
         }
     }
 
     // The fields of concern to the driver: no other op may modify the fields listed here.
     FieldRef* fieldRef[MAX_NUM_FIELDS];  // not owned here
+
+    // For each modified field ref, the index of the path component representing an existing array
+    // that gained a new element.
+    boost::optional<size_t> indexOfArrayWithNewElement[MAX_NUM_FIELDS];
+
     bool noOp;
     UpdateContext context;
 };
diff --git a/src/mongo/db/ops/modifier_push.cpp b/src/mongo/db/ops/modifier_push.cpp
index 8bc671ecf46..ef19e45c8a2 100644
--- a/src/mongo/db/ops/modifier_push.cpp
+++ b/src/mongo/db/ops/modifier_push.cpp
@@ -434,6 +434,8 @@ Status ModifierPush::prepare(mutablebson::Element root,
     // be created during the apply.
     Status status = pathsupport::findLongestPrefix(
         _fieldRef, root, &_preparedState->idxFound, &_preparedState->elemFound);
+    const auto elemFoundIsArray =
+        _preparedState->elemFound.ok() && _preparedState->elemFound.getType() == BSONType::Array;
 
     // FindLongestPrefix may say the path does not exist at all, which is fine here, or
     // that the path was not viable or otherwise wrong, in which case, the mod cannot
@@ -463,6 +465,13 @@ Status ModifierPush::prepare(mutablebson::Element root,
     // there is any conflict among mods.
     execInfo->fieldRef[0] = &_fieldRef;
 
+    if (!_preparedState->elemFound.ok() || _preparedState->idxFound < (_fieldRef.numParts() - 1)) {
+        if (elemFoundIsArray) {
+            // Report that an existing array will gain a new element as a result of this mod.
+            execInfo->indexOfArrayWithNewElement[0] = _preparedState->idxFound;
+        }
+    }
+
     return Status::OK();
 }
 
diff --git a/src/mongo/db/ops/modifier_push_test.cpp b/src/mongo/db/ops/modifier_push_test.cpp
index 49b9d01b002..a7de5d9f0f0 100644
--- a/src/mongo/db/ops/modifier_push_test.cpp
+++ b/src/mongo/db/ops/modifier_push_test.cpp
@@ -1473,4 +1473,87 @@ TEST(ToPosition, Back) {
     ASSERT_EQUALS(fromjson("{$set: {'a.1':1}}"), logDoc);
 }
 
+TEST(IndexedMod, PrepareReportCreatedArrayElement) {
+    Document doc(fromjson("{a: [{b: 0}]}"));
+    Mod mod(fromjson("{$push: {'a.1.c': 2}}"));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(mod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.1.c");
+    ASSERT_TRUE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_EQUALS(*execInfo.indexOfArrayWithNewElement[0], 0u);
+    ASSERT_FALSE(execInfo.noOp);
+}
+
+TEST(IndexedMod, PrepareReportCreatedArrayElementPushAll) {
+    Document doc(fromjson("{a: [{b: 0}]}"));
+    auto modObj = fromjson("{$pushAll: {'a.1.c': [2]}}");
+    ModifierPush mod(ModifierPush::PUSH_ALL);
+    ASSERT_OK(mod.init(modObj["$pushAll"].embeddedObject().firstElement(),
+                       ModifierInterface::Options::normal()));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(mod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.1.c");
+    ASSERT_TRUE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_EQUALS(*execInfo.indexOfArrayWithNewElement[0], 0u);
+    ASSERT_FALSE(execInfo.noOp);
+}
+
+TEST(IndexedMod, PrepareDoNotReportModifiedArrayElement) {
+    Document doc(fromjson("{a: [{b: 0}]}"));
+    Mod mod(fromjson("{$push: {'a.0.c': 2}}"));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(mod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.0.c");
+    ASSERT_FALSE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_FALSE(execInfo.noOp);
+}
+
+TEST(IndexedMod, PrepareDoNotReportModifiedArrayElementPushAll) {
+    Document doc(fromjson("{a: [{b: 0}]}"));
+    auto modObj = fromjson("{$pushAll: {'a.0.c': [2]}}");
+    ModifierPush mod(ModifierPush::PUSH_ALL);
+    ASSERT_OK(mod.init(modObj["$pushAll"].embeddedObject().firstElement(),
+                       ModifierInterface::Options::normal()));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(mod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.0.c");
+    ASSERT_FALSE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_FALSE(execInfo.noOp);
+}
+
+TEST(IndexedMod, PrepareDoNotReportCreatedNumericObjectField) {
+    Document doc(fromjson("{a: {'0': {b: 0}}}"));
+    Mod mod(fromjson("{$push: {'a.1.c': 2}}"));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(mod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.1.c");
+    ASSERT_FALSE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_FALSE(execInfo.noOp);
+}
+
+TEST(IndexedMod, PrepareDoNotReportCreatedNumericObjectFieldPushAll) {
+    Document doc(fromjson("{a: {'0': {b: 0}}}"));
+    auto modObj = fromjson("{$pushAll: {'a.1.c': [2]}}");
+    ModifierPush mod(ModifierPush::PUSH_ALL);
+    ASSERT_OK(mod.init(modObj["$pushAll"].embeddedObject().firstElement(),
+                       ModifierInterface::Options::normal()));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(mod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.1.c");
+    ASSERT_FALSE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_FALSE(execInfo.noOp);
+}
+
 }  // unnamed namespace
diff --git a/src/mongo/db/ops/modifier_set.cpp b/src/mongo/db/ops/modifier_set.cpp
index 59f59c555ce..03b703b860c 100644
--- a/src/mongo/db/ops/modifier_set.cpp
+++ b/src/mongo/db/ops/modifier_set.cpp
@@ -127,6 +127,8 @@ Status ModifierSet::prepare(mutablebson::Element root,
     // be created during the apply.
     Status status = pathsupport::findLongestPrefix(
         _fieldRef, root, &_preparedState->idxFound, &_preparedState->elemFound);
+    const auto elemFoundIsArray =
+        _preparedState->elemFound.ok() && _preparedState->elemFound.getType() == BSONType::Array;
 
     // FindLongestPrefix may say the path does not exist at all, which is fine here, or
     // that the path was not viable or otherwise wrong, in which case, the mod cannot
@@ -155,6 +157,10 @@ Status ModifierSet::prepare(mutablebson::Element root,
 
     // If the field path is not fully present, then this mod cannot be in place, nor is it a noOp.
     if (!_preparedState->elemFound.ok() || _preparedState->idxFound < (_fieldRef.numParts() - 1)) {
+        if (elemFoundIsArray) {
+            // Report that an existing array will gain a new element as a result of this mod.
+            execInfo->indexOfArrayWithNewElement[0] = _preparedState->idxFound;
+        }
         return Status::OK();
     }
 
diff --git a/src/mongo/db/ops/modifier_set_test.cpp b/src/mongo/db/ops/modifier_set_test.cpp
index 1579b096722..b83270da957 100644
--- a/src/mongo/db/ops/modifier_set_test.cpp
+++ b/src/mongo/db/ops/modifier_set_test.cpp
@@ -429,6 +429,43 @@ TEST(IndexedMod, PrepareNonViablePath) {
     ASSERT_NOT_OK(setMod.prepare(doc.root(), "", &execInfo));
 }
 
+TEST(IndexedMod, PrepareReportCreatedArrayElement) {
+    Document doc(fromjson("{a: [{b: 0}]}"));
+    Mod setMod(fromjson("{$set: {'a.1.c': 2}}"));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(setMod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.1.c");
+    ASSERT_TRUE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_EQUALS(*execInfo.indexOfArrayWithNewElement[0], 0u);
+    ASSERT_FALSE(execInfo.noOp);
+}
+
+TEST(IndexedMod, PrepareDoNotReportModifiedArrayElement) {
+    Document doc(fromjson("{a: [{b: 0}]}"));
+    Mod setMod(fromjson("{$set: {'a.0.c': 2}}"));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(setMod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.0.c");
+    ASSERT_FALSE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_FALSE(execInfo.noOp);
+}
+
+TEST(IndexedMod, PrepareDoNotReportCreatedNumericObjectField) {
+    Document doc(fromjson("{a: {'0': {b: 0}}}"));
+    Mod setMod(fromjson("{$set: {'a.1.c': 2}}"));
+
+    ModifierInterface::ExecInfo execInfo;
+    ASSERT_OK(setMod.prepare(doc.root(), "", &execInfo));
+
+    ASSERT_EQUALS(execInfo.fieldRef[0]->dottedField(), "a.1.c");
+    ASSERT_FALSE(execInfo.indexOfArrayWithNewElement[0]);
+    ASSERT_FALSE(execInfo.noOp);
+}
+
 TEST(IndexedMod, PrepareApplyInPlace) {
     Document doc(fromjson("{a: [{b: 0},{b: 1},{b: 1}]}"));
     Mod setMod(fromjson("{$set: {'a.2.b': 2}}"));
diff --git a/src/mongo/db/ops/update.cpp b/src/mongo/db/ops/update.cpp
index 5e0763f9eac..bc3c595dc6c 100644
--- a/src/mongo/db/ops/update.cpp
+++ b/src/mongo/db/ops/update.cpp
@@ -63,20 +63,9 @@ UpdateResult update(OperationContext* txn, Database* db, const UpdateRequest& re
     // Explain should never use this helper.
     invariant(!request.isExplain());
 
-    auto client = txn->getClient();
-    auto lastOpAtOperationStart = repl::ReplClientInfo::forClient(client).getLastOp();
-    ScopeGuard lastOpSetterGuard = MakeObjGuard(repl::ReplClientInfo::forClient(client),
-                                                &repl::ReplClientInfo::setLastOpToSystemLastOpTime,
-                                                txn);
-
     const NamespaceString& nsString = request.getNamespaceString();
     Collection* collection = db->getCollection(nsString.ns());
 
-    // If this is the local database, don't set last op.
-    if (db->name() == "local") {
-        lastOpSetterGuard.Dismiss();
-    }
-
     // The update stage does not create its own collection.  As such, if the update is
     // an upsert, create the collection that the update stage inserts into beforehand.
     if (!collection && request.isUpsert()) {
@@ -116,12 +105,6 @@ UpdateResult update(OperationContext* txn, Database* db, const UpdateRequest& re
         uassertStatusOK(getExecutorUpdate(txn, nullOpDebug, collection, &parsedUpdate));
 
     uassertStatusOK(exec->executePlan());
-    if (repl::ReplClientInfo::forClient(client).getLastOp() != lastOpAtOperationStart) {
-        // If this operation has already generated a new lastOp, don't bother setting it here.
-        // No-op updates will not generate a new lastOp, so we still need the guard to fire in that
-        // case.
-        lastOpSetterGuard.Dismiss();
-    }
 
     const UpdateStats* updateStats = UpdateStage::getUpdateStats(exec.get());
 
diff --git a/src/mongo/db/ops/update_driver.cpp b/src/mongo/db/ops/update_driver.cpp
index e5a63d64ccb..dc47ca4175f 100644
--- a/src/mongo/db/ops/update_driver.cpp
+++ b/src/mongo/db/ops/update_driver.cpp
@@ -291,11 +291,21 @@ Status UpdateDriver::update(StringData matchedField,
             // is not a no-op and it is in a valid context -- then we switch back to a
             // non-in-place mode.
             //
-            // TODO: make mightBeIndexed and fieldRef like each other.
-            if (!_affectIndices && !execInfo.noOp && _indexedFields &&
-                _indexedFields->mightBeIndexed(execInfo.fieldRef[i]->dottedField())) {
-                _affectIndices = true;
-                doc->disableInPlaceUpdates();
+            // To determine if indexes are affected: If we did not create a new element in an array,
+            // check whether the full path affects indexes. If we did create a new element in an
+            // array, check whether the array itself might affect any indexes. This is necessary
+            // because if there is an index {"a.b": 1}, and we set "a.1.c" and implicitly create an
+            // array element in "a", then we may need to add a null key to the index {"a.b": 1},
+            // even though "a.1.c" does not appear to affect the index.
+            if (!_affectIndices && !execInfo.noOp && _indexedFields) {
+                auto pathLengthForIndexCheck = execInfo.indexOfArrayWithNewElement[i]
+                    ? *execInfo.indexOfArrayWithNewElement[i] + 1
+                    : execInfo.fieldRef[i]->numParts();
+                if (_indexedFields->mightBeIndexed(
+                        execInfo.fieldRef[i]->dottedSubstring(0, pathLengthForIndexCheck))) {
+                    _affectIndices = true;
+                    doc->disableInPlaceUpdates();
+                }
             }
         }
 
diff --git a/src/mongo/db/pipeline/SConscript b/src/mongo/db/pipeline/SConscript
index 98a66a02f48..ca8617f3d3d 100644
--- a/src/mongo/db/pipeline/SConscript
+++ b/src/mongo/db/pipeline/SConscript
@@ -252,8 +252,8 @@ docSourceEnv.Library(
         '$BUILD_DIR/mongo/db/matcher/expression_algo',
         '$BUILD_DIR/mongo/db/service_context',
         '$BUILD_DIR/mongo/db/stats/top',
+        '$BUILD_DIR/mongo/db/storage/encryption_hooks',
         '$BUILD_DIR/mongo/db/storage/storage_options',
-        '$BUILD_DIR/mongo/db/storage/wiredtiger/storage_wiredtiger_customization_hooks',
         '$BUILD_DIR/third_party/shim_snappy',
     ],
     LIBDEPS_TAGS=[
diff --git a/src/mongo/db/pipeline/document_source_geo_near.cpp b/src/mongo/db/pipeline/document_source_geo_near.cpp
index 17e9997bbf5..a385eeeb07e 100644
--- a/src/mongo/db/pipeline/document_source_geo_near.cpp
+++ b/src/mongo/db/pipeline/document_source_geo_near.cpp
@@ -150,10 +150,12 @@ BSONObj DocumentSourceGeoNear::buildGeoNearCmd() const {
         geoNear.append("minDistance", minDistance);
 
     geoNear.append("query", query);
-    if (pExpCtx->getCollator()) {
-        geoNear.append("collation", pExpCtx->getCollator()->getSpec().toBSON());
-    } else {
-        geoNear.append("collation", CollationSpec::kSimpleSpec);
+    if (!pExpCtx->collation.isEmpty()) {
+        if (pExpCtx->getCollator()) {
+            geoNear.append("collation", pExpCtx->getCollator()->getSpec().toBSON());
+        } else {
+            geoNear.append("collation", CollationSpec::kSimpleSpec);
+        }
     }
 
     geoNear.append("spherical", spherical);
diff --git a/src/mongo/db/pipeline/parsed_aggregation_projection.cpp b/src/mongo/db/pipeline/parsed_aggregation_projection.cpp
index 69b196faab3..c8d7bf06d7f 100644
--- a/src/mongo/db/pipeline/parsed_aggregation_projection.cpp
+++ b/src/mongo/db/pipeline/parsed_aggregation_projection.cpp
@@ -46,6 +46,8 @@
 namespace mongo {
 namespace parsed_aggregation_projection {
 
+using expression::isPathPrefixOf;
+
 //
 // ProjectionSpecValidator
 //
@@ -54,22 +56,36 @@ Status ProjectionSpecValidator::validate(const BSONObj& spec) {
     return ProjectionSpecValidator(spec).validate();
 }
 
-Status ProjectionSpecValidator::ensurePathDoesNotConflictOrThrow(StringData path) {
-    for (auto&& seenPath : _seenPaths) {
-        if ((path == seenPath) || (expression::isPathPrefixOf(path, seenPath)) ||
-            (expression::isPathPrefixOf(seenPath, path))) {
-            return Status(ErrorCodes::FailedToParse,
-                          str::stream() << "specification contains two conflicting paths. "
-                                           "Cannot specify both '"
-                                        << path
-                                        << "' and '"
-                                        << seenPath
-                                        << "': "
-                                        << _rawObj.toString(),
-                          40176);
-        }
+Status ProjectionSpecValidator::ensurePathDoesNotConflictOrThrow(const std::string& path) {
+    auto result = _seenPaths.emplace(path);
+    auto pos = result.first;
+
+    // Check whether the path was a duplicate of an existing path.
+    auto conflictingPath = boost::make_optional(!result.second, *pos);
+
+    // Check whether the preceding path prefixes this path.
+    if (!conflictingPath && pos != _seenPaths.begin()) {
+        conflictingPath =
+            boost::make_optional(isPathPrefixOf(*std::prev(pos), path), *std::prev(pos));
+    }
+
+    // Check whether this path prefixes the subsequent path.
+    if (!conflictingPath && std::next(pos) != _seenPaths.end()) {
+        conflictingPath =
+            boost::make_optional(isPathPrefixOf(path, *std::next(pos)), *std::next(pos));
+    }
+
+    if (conflictingPath) {
+        return Status(ErrorCodes::FailedToParse,
+                      str::stream() << "specification contains two conflicting paths. "
+                                       "Cannot specify both '"
+                                    << path
+                                    << "' and '"
+                                    << *conflictingPath
+                                    << "': "
+                                    << _rawObj.toString(),
+                      40176);
     }
-    _seenPaths.emplace_back(path.toString());
     return Status::OK();
 }
 
diff --git a/src/mongo/db/pipeline/parsed_aggregation_projection.h b/src/mongo/db/pipeline/parsed_aggregation_projection.h
index 8d92e4ab768..4fd9731774e 100644
--- a/src/mongo/db/pipeline/parsed_aggregation_projection.h
+++ b/src/mongo/db/pipeline/parsed_aggregation_projection.h
@@ -69,7 +69,7 @@ private:
      * For example, a user is not allowed to specify {'a': 1, 'a.b': 1}, or some similar conflicting
      * paths.
      */
-    Status ensurePathDoesNotConflictOrThrow(StringData path);
+    Status ensurePathDoesNotConflictOrThrow(const std::string& path);
 
     /**
      * Returns the relevant error if an invalid projection specification is detected.
@@ -99,9 +99,40 @@ private:
     // The original object. Used to generate more helpful error messages.
     const BSONObj& _rawObj;
 
+    // Custom comparator that orders fieldpath strings by path prefix first, then by field.
+    struct PathPrefixComparator {
+        static constexpr char dot = '.';
+
+        // Returns true if the lhs value should sort before the rhs, false otherwise.
+        bool operator()(const std::string& lhs, const std::string& rhs) const {
+            for (size_t pos = 0, len = std::min(lhs.size(), rhs.size()); pos < len; ++pos) {
+                auto &lchar = lhs[pos], &rchar = rhs[pos];
+                if (lchar == rchar) {
+                    continue;
+                }
+
+                // Consider the path delimiter '.' as being less than all other characters, so that
+                // paths sort directly before any paths they prefix and directly after any paths
+                // which prefix them.
+                if (lchar == dot) {
+                    return true;
+                } else if (rchar == dot) {
+                    return false;
+                }
+
+                // Otherwise, default to normal character comparison.
+                return lchar < rchar;
+            }
+
+            // If we get here, then we have reached the end of lhs and/or rhs and all of their path
+            // segments up to this point match. If lhs is shorter than rhs, then lhs prefixes rhs
+            // and should sort before it.
+            return lhs.size() < rhs.size();
+        }
+    };
+
     // Tracks which paths we've seen to ensure no two paths conflict with each other.
-    // Can be a vector since we iterate through it.
-    std::vector<std::string> _seenPaths;
+    std::set<std::string, PathPrefixComparator> _seenPaths;
 };
 
 /**
diff --git a/src/mongo/db/pipeline/parsed_aggregation_projection_test.cpp b/src/mongo/db/pipeline/parsed_aggregation_projection_test.cpp
index 71dd1c2378f..cb5d50b2771 100644
--- a/src/mongo/db/pipeline/parsed_aggregation_projection_test.cpp
+++ b/src/mongo/db/pipeline/parsed_aggregation_projection_test.cpp
@@ -126,6 +126,75 @@ TEST(ParsedAggregationProjectionErrors, ShouldRejectFieldsWithSharedPrefix) {
                   UserException);
 }
 
+TEST(ParsedAggregationProjectionErrors, ShouldRejectPathConflictsWithNonAlphaNumericCharacters) {
+    const boost::intrusive_ptr<ExpressionContextForTest> expCtx(new ExpressionContextForTest());
+    // Include/exclude non-alphanumeric fields with a shared prefix. First assert that the non-
+    // alphanumeric fields are accepted when no prefixes are present.
+    ASSERT(ParsedAggregationProjection::create(
+        expCtx, BSON("a.b-c" << true << "a.b" << true << "a.b?c" << true << "a.b c" << true)));
+    ASSERT(ParsedAggregationProjection::create(
+        expCtx, BSON("a.b c" << false << "a.b?c" << false << "a.b" << false << "a.b-c" << false)));
+
+    // Then assert that we throw when we introduce a prefixed field.
+    ASSERT_THROWS(
+        ParsedAggregationProjection::create(
+            expCtx,
+            BSON("a.b-c" << true << "a.b" << true << "a.b?c" << true << "a.b c" << true << "a.b.d"
+                         << true)),
+        AssertionException);
+    ASSERT_THROWS(
+        ParsedAggregationProjection::create(
+            expCtx,
+            BSON("a.b.d" << false << "a.b c" << false << "a.b?c" << false << "a.b" << false
+                         << "a.b-c"
+                         << false)),
+        AssertionException);
+
+    // Adding the same field twice.
+    ASSERT_THROWS(ParsedAggregationProjection::create(
+                      expCtx, BSON("a.b?c" << wrapInLiteral(1) << "a.b?c" << wrapInLiteral(0))),
+                  AssertionException);
+    ASSERT_THROWS(ParsedAggregationProjection::create(
+                      expCtx, BSON("a.b c" << wrapInLiteral(0) << "a.b c" << wrapInLiteral(1))),
+                  AssertionException);
+
+    // Mix of include/exclude and adding a shared prefix.
+    ASSERT_THROWS(
+        ParsedAggregationProjection::create(
+            expCtx,
+            BSON("a.b-c" << true << "a.b" << wrapInLiteral(1) << "a.b?c" << true << "a.b c" << true
+                         << "a.b.d"
+                         << true)),
+        AssertionException);
+    ASSERT_THROWS(ParsedAggregationProjection::create(
+                      expCtx,
+                      BSON("a.b.d" << false << "a.b c" << false << "a.b?c" << false << "a.b"
+                                   << wrapInLiteral(0)
+                                   << "a.b-c"
+                                   << false)),
+                  AssertionException);
+
+    // Adding a shared prefix twice.
+    ASSERT_THROWS(ParsedAggregationProjection::create(
+                      expCtx,
+                      BSON("a.b-c" << wrapInLiteral(1) << "a.b" << wrapInLiteral(1) << "a.b?c"
+                                   << wrapInLiteral(1)
+                                   << "a.b c"
+                                   << wrapInLiteral(1)
+                                   << "a.b.d"
+                                   << wrapInLiteral(0))),
+                  AssertionException);
+    ASSERT_THROWS(ParsedAggregationProjection::create(
+                      expCtx,
+                      BSON("a.b.d" << wrapInLiteral(1) << "a.b c" << wrapInLiteral(1) << "a.b?c"
+                                   << wrapInLiteral(1)
+                                   << "a.b"
+                                   << wrapInLiteral(0)
+                                   << "a.b-c"
+                                   << wrapInLiteral(1))),
+                  AssertionException);
+}
+
 TEST(ParsedAggregationProjectionErrors, ShouldRejectMixOfIdAndSubFieldsOfId) {
     const boost::intrusive_ptr<ExpressionContextForTest> expCtx(new ExpressionContextForTest());
     // Include/exclude _id twice.
diff --git a/src/mongo/db/query/SConscript b/src/mongo/db/query/SConscript
index a6c14a5f7dc..521f44cecb8 100644
--- a/src/mongo/db/query/SConscript
+++ b/src/mongo/db/query/SConscript
@@ -364,47 +364,11 @@ env.CppUnitTest(
 env.CppUnitTest(
     target="query_planner_test",
     source=[
-        "query_planner_test.cpp"
-    ],
-    LIBDEPS=[
-        "query_planner_test_fixture",
-    ],
-)
-
-env.CppUnitTest(
-    target="query_planner_array_test",
-    source=[
-        "query_planner_array_test.cpp"
-    ],
-    LIBDEPS=[
-        "query_planner_test_fixture",
-    ],
-)
-
-env.CppUnitTest(
-    target="query_planner_geo_test",
-    source=[
-        "query_planner_geo_test.cpp"
-    ],
-    LIBDEPS=[
-        "query_planner_test_fixture",
-    ],
-)
-
-env.CppUnitTest(
-    target="query_planner_partialidx_test",
-    source=[
-        "query_planner_partialidx_test.cpp"
-    ],
-    LIBDEPS=[
-        "query_planner_test_fixture",
-    ],
-)
-
-env.CppUnitTest(
-    target="query_planner_collation_test",
-    source=[
+        "query_planner_array_test.cpp",
         "query_planner_collation_test.cpp",
+        "query_planner_geo_test.cpp",
+        "query_planner_partialidx_test.cpp",
+        "query_planner_test.cpp",
     ],
     LIBDEPS=[
         "collation/collator_interface_mock",
diff --git a/src/mongo/db/query/canonical_query_test.cpp b/src/mongo/db/query/canonical_query_test.cpp
index c7a1f2a05a2..31ba2737ee9 100644
--- a/src/mongo/db/query/canonical_query_test.cpp
+++ b/src/mongo/db/query/canonical_query_test.cpp
@@ -548,7 +548,8 @@ TEST(CanonicalQueryTest, NormalizeWithInPreservesCollator) {
     BSONObj obj = fromjson("{'': 'string'}");
     auto inMatchExpression = stdx::make_unique<InMatchExpression>();
     inMatchExpression->setCollator(&collator);
-    inMatchExpression->addEquality(obj.firstElement());
+    std::vector<BSONElement> equalities{obj.firstElement()};
+    ASSERT_OK(inMatchExpression->setEqualities(std::move(equalities)));
     unique_ptr<MatchExpression> matchExpression(
         CanonicalQuery::normalizeTree(inMatchExpression.release()));
     ASSERT(matchExpression->matchType() == MatchExpression::MatchType::EQ);
diff --git a/src/mongo/db/query/index_bounds_builder.cpp b/src/mongo/db/query/index_bounds_builder.cpp
index b16ad43dc5e..eeb1c2f1114 100644
--- a/src/mongo/db/query/index_bounds_builder.cpp
+++ b/src/mongo/db/query/index_bounds_builder.cpp
@@ -249,6 +249,14 @@ bool typeMatch(const BSONObj& obj) {
     return first.canonicalType() == second.canonicalType();
 }
 
+bool IndexBoundsBuilder::canUseCoveredMatching(const MatchExpression* expr,
+                                               const IndexEntry& index) {
+    IndexBoundsBuilder::BoundsTightness tightness;
+    OrderedIntervalList oil;
+    translate(expr, BSONElement{}, index, &oil, &tightness);
+    return tightness >= IndexBoundsBuilder::INEXACT_COVERED;
+}
+
 // static
 void IndexBoundsBuilder::translate(const MatchExpression* expr,
                                    const BSONElement& elt,
diff --git a/src/mongo/db/query/index_bounds_builder.h b/src/mongo/db/query/index_bounds_builder.h
index b397f81b295..311b3ee61f6 100644
--- a/src/mongo/db/query/index_bounds_builder.h
+++ b/src/mongo/db/query/index_bounds_builder.h
@@ -70,6 +70,17 @@ public:
     static void allValuesForField(const BSONElement& elt, OrderedIntervalList* out);
 
     /**
+     * Returns true if 'expr' can correctly be assigned as an INEXACT_COVERED predicate to an index
+     * scan over 'index'.
+     *
+     * The result of this function is not meaningful when the predicate applies to special fields
+     * such as "hashed", "2d", or "2dsphere". That is, the caller is responsible for ensuring that
+     * 'expr' is a candidate for covered matching over a regular ascending/descending field of the
+     * index.
+     */
+    static bool canUseCoveredMatching(const MatchExpression* expr, const IndexEntry& index);
+
+    /**
      * Turn the MatchExpression in 'expr' into a set of index bounds.  The field that 'expr' is
      * concerned with is indexed according to the keypattern element 'elt' from index 'index'.
      *
diff --git a/src/mongo/db/query/index_bounds_builder_test.cpp b/src/mongo/db/query/index_bounds_builder_test.cpp
index 45f11a2cc1d..a8ae3651d32 100644
--- a/src/mongo/db/query/index_bounds_builder_test.cpp
+++ b/src/mongo/db/query/index_bounds_builder_test.cpp
@@ -2137,4 +2137,54 @@ TEST(IndexBoundsBuilderTest, InWithStringAgainstHashedIndexWithCollatorUsesHashO
     ASSERT_EQUALS(tightness, IndexBoundsBuilder::INEXACT_FETCH);
 }
 
+TEST(IndexBoundsBuilderTest, CanUseCoveredMatchingForEqualityPredicate) {
+    IndexEntry testIndex = IndexEntry(BSONObj());
+    BSONObj obj = fromjson("{a: {$eq: 3}}");
+    unique_ptr<MatchExpression> expr(parseMatchExpression(obj));
+    ASSERT_TRUE(IndexBoundsBuilder::canUseCoveredMatching(expr.get(), testIndex));
+}
+
+TEST(IndexBoundsBuilderTest, CannotUseCoveredMatchingForEqualityToArrayPredicate) {
+    IndexEntry testIndex = IndexEntry(BSONObj());
+    BSONObj obj = fromjson("{a: {$eq: [1, 2, 3]}}");
+    unique_ptr<MatchExpression> expr(parseMatchExpression(obj));
+    ASSERT_FALSE(IndexBoundsBuilder::canUseCoveredMatching(expr.get(), testIndex));
+}
+
+TEST(IndexBoundsBuilderTest, CannotUseCoveredMatchingForEqualityToNullPredicate) {
+    IndexEntry testIndex = IndexEntry(BSONObj());
+    BSONObj obj = fromjson("{a: null}");
+    unique_ptr<MatchExpression> expr(parseMatchExpression(obj));
+    ASSERT_FALSE(IndexBoundsBuilder::canUseCoveredMatching(expr.get(), testIndex));
+}
+
+TEST(IndexBoundsBuilderTest, CannotUseCoveredMatchingForTypeArrayPredicate) {
+    IndexEntry testIndex = IndexEntry(BSONObj());
+    BSONObj obj = fromjson("{a: {$type: 'array'}}");
+    unique_ptr<MatchExpression> expr(parseMatchExpression(obj));
+    ASSERT_FALSE(IndexBoundsBuilder::canUseCoveredMatching(expr.get(), testIndex));
+}
+
+TEST(IndexBoundsBuilderTest, CannotUseCoveredMatchingForExistsTruePredicate) {
+    IndexEntry testIndex = IndexEntry(BSONObj());
+    BSONObj obj = fromjson("{a: {$exists: true}}");
+    unique_ptr<MatchExpression> expr(parseMatchExpression(obj));
+    ASSERT_FALSE(IndexBoundsBuilder::canUseCoveredMatching(expr.get(), testIndex));
+}
+
+TEST(IndexBoundsBuilderTest, CannotUseCoveredMatchingForExistsFalsePredicate) {
+    IndexEntry testIndex = IndexEntry(BSONObj());
+    BSONObj obj = fromjson("{a: {$exists: false}}");
+    unique_ptr<MatchExpression> expr(parseMatchExpression(obj));
+    ASSERT_FALSE(IndexBoundsBuilder::canUseCoveredMatching(expr.get(), testIndex));
+}
+
+TEST(IndexBoundsBuilderTest, CanUseCoveredMatchingForExistsTrueWithSparseIndex) {
+    IndexEntry testIndex = IndexEntry(BSONObj());
+    testIndex.sparse = true;
+    BSONObj obj = fromjson("{a: {$exists: true}}");
+    unique_ptr<MatchExpression> expr(parseMatchExpression(obj));
+    ASSERT_TRUE(IndexBoundsBuilder::canUseCoveredMatching(expr.get(), testIndex));
+}
+
 }  // namespace
diff --git a/src/mongo/db/query/plan_enumerator.cpp b/src/mongo/db/query/plan_enumerator.cpp
index 789cd405607..de80606df8d 100644
--- a/src/mongo/db/query/plan_enumerator.cpp
+++ b/src/mongo/db/query/plan_enumerator.cpp
@@ -579,6 +579,32 @@ bool PlanEnumerator::prepMemo(MatchExpression* node, PrepMemoContext context) {
     return false;
 }
 
+void PlanEnumerator::assignToNonMultikeyMandatoryIndex(
+    const IndexEntry& index,
+    const std::vector<MatchExpression*>& predsOverLeadingField,
+    const IndexToPredMap& idxToNotFirst,
+    OneIndexAssignment* indexAssign) {
+    // Text indexes are typically multikey because there is an index key for each token in the
+    // source text. However, the leading and trailing non-text fields of the index cannot be
+    // multikey. As a result, we should use non-multikey predicate assignment rules for such
+    // indexes.
+    invariant(!index.multikey || index.type == IndexType::INDEX_TEXT);
+
+    // Since the index is not multikey, all predicates over the leading field can be assigned.
+    indexAssign->preds = predsOverLeadingField;
+
+    // Since everything in assign.preds prefixes the index, they all go at position '0' in the
+    // index, the first position.
+    indexAssign->positions.resize(indexAssign->preds.size(), 0);
+
+    // And now we begin compound analysis. Find everything that could use assign.index but isn't a
+    // pred over the first field of that index.
+    auto compIt = idxToNotFirst.find(indexAssign->index);
+    if (compIt != idxToNotFirst.end()) {
+        compound(compIt->second, index, indexAssign);
+    }
+}
+
 bool PlanEnumerator::enumerateMandatoryIndex(const IndexToPredMap& idxToFirst,
                                              const IndexToPredMap& idxToNotFirst,
                                              MatchExpression* mandatoryPred,
@@ -613,7 +639,12 @@ bool PlanEnumerator::enumerateMandatoryIndex(const IndexToPredMap& idxToFirst,
 
         const vector<MatchExpression*>& predsOverLeadingField = it->second;
 
-        if (thisIndex.multikey && !thisIndex.multikeyPaths.empty()) {
+        // Text indexes should be treated like non-multikey indexes, since the non-text fields are
+        // prohibited from containing arrays.
+        if (thisIndex.type == IndexType::INDEX_TEXT) {
+            assignToNonMultikeyMandatoryIndex(
+                thisIndex, predsOverLeadingField, idxToNotFirst, &indexAssign);
+        } else if (thisIndex.multikey && !thisIndex.multikeyPaths.empty()) {
             // 2dsphere indexes are the only special index type that should ever have path-level
             // multikey information.
             invariant(INDEX_2DSPHERE == thisIndex.type);
@@ -719,22 +750,9 @@ bool PlanEnumerator::enumerateMandatoryIndex(const IndexToPredMap& idxToFirst,
                 }
             }
         } else {
-            // For non-multikey, we don't have to do anything too special.
-            // Just assign all "first" predicates and try to compound like usual.
-            indexAssign.preds = it->second;
-
-            // Since everything in assign.preds prefixes the index, they all go
-            // at position '0' in the index, the first position.
-            indexAssign.positions.resize(indexAssign.preds.size(), 0);
-
-            // And now we begin compound analysis.
-
-            // Find everything that could use assign.index but isn't a pred over
-            // the first field of that index.
-            IndexToPredMap::const_iterator compIt = idxToNotFirst.find(indexAssign.index);
-            if (compIt != idxToNotFirst.end()) {
-                compound(compIt->second, thisIndex, &indexAssign);
-            }
+            // The index is not multikey.
+            assignToNonMultikeyMandatoryIndex(
+                thisIndex, predsOverLeadingField, idxToNotFirst, &indexAssign);
         }
 
         // The mandatory predicate must be assigned.
diff --git a/src/mongo/db/query/plan_enumerator.h b/src/mongo/db/query/plan_enumerator.h
index 5543bc68ca9..a29c1dd4e76 100644
--- a/src/mongo/db/query/plan_enumerator.h
+++ b/src/mongo/db/query/plan_enumerator.h
@@ -434,6 +434,17 @@ private:
                                  AndAssignment* andAssignment);
 
     /**
+     * Assigns predicates in 'predsOverLeadingField' and 'idxToNotFirst' to 'indexAssign'. Assumes
+     * that the index is not multikey. Also assumes that that the index is of a type used to answer
+     * "mandatory predicates" such as text or geoNear.
+     */
+    void assignToNonMultikeyMandatoryIndex(
+        const IndexEntry& index,
+        const std::vector<MatchExpression*>& predsOverLeadingField,
+        const IndexToPredMap& idxToNotFirst,
+        OneIndexAssignment* indexAssign);
+
+    /**
      * Try to assign predicates in 'tryCompound' to 'thisIndex' as compound assignments.
      * Output the assignments in 'assign'.
      */
diff --git a/src/mongo/db/query/planner_access.cpp b/src/mongo/db/query/planner_access.cpp
index 9a6f52e18d0..90c4873ffbb 100644
--- a/src/mongo/db/query/planner_access.cpp
+++ b/src/mongo/db/query/planner_access.cpp
@@ -224,6 +224,17 @@ QuerySolutionNode* QueryPlannerAccess::makeLeafNode(
         TextMatchExpressionBase* textExpr = static_cast<TextMatchExpressionBase*>(expr);
         TextNode* ret = new TextNode(index);
         ret->ftsQuery = textExpr->getFTSQuery().clone();
+
+        // Count the number of prefix fields before the "text" field.
+        for (auto&& keyPatternElt : ret->index.keyPattern) {
+            // We know that the only key pattern with a type of String is the _fts field
+            // which is immediately after all prefix fields.
+            if (BSONType::String == keyPatternElt.type()) {
+                break;
+            }
+            ++(ret->numPrefixFields);
+        }
+
         return ret;
     } else {
         // Note that indexKeyPattern.firstElement().fieldName() may not equal expr->path()
@@ -335,10 +346,24 @@ void QueryPlannerAccess::mergeWithLeafNode(MatchExpression* expr, ScanBuildingSt
 
     const StageType type = node->getType();
 
-    // Text data is covered, but not exactly.  Text covering is unlike any other covering
-    // so we deal with it in addFilterToSolutionNode.
     if (STAGE_TEXT == type) {
-        scanState->tightness = IndexBoundsBuilder::INEXACT_COVERED;
+        auto textNode = static_cast<TextNode*>(node);
+
+        if (pos < textNode->numPrefixFields) {
+            // This predicate is assigned to one of the prefix fields of the text index. Such
+            // predicates must always be equalities and must always be attached to the TEXT node. In
+            // order to ensure this happens, we assign INEXACT_COVERED tightness.
+            scanState->tightness = IndexBoundsBuilder::INEXACT_COVERED;
+        } else {
+            // The predicate is assigned to one of the trailing fields of the text index. We
+            // currently don't generate bounds for predicates assigned to trailing fields of a text
+            // index, but rather attempt to attach a covered filter. However, certain predicates can
+            // never be correctly covered (e.g. $exists), so we assign the tightness accordingly.
+            scanState->tightness = IndexBoundsBuilder::canUseCoveredMatching(expr, index)
+                ? IndexBoundsBuilder::INEXACT_COVERED
+                : IndexBoundsBuilder::INEXACT_FETCH;
+        }
+
         return;
     }
 
@@ -347,20 +372,23 @@ void QueryPlannerAccess::mergeWithLeafNode(MatchExpression* expr, ScanBuildingSt
     if (STAGE_GEO_NEAR_2D == type) {
         invariant(INDEX_2D == index.type);
 
-        // 2D indexes are weird - the "2d" field stores a normally-indexed BinData field, but
-        // additional array fields are *not* exploded into multi-keys - they are stored directly
-        // as arrays in the index.  Also, no matter what the index expression, the "2d" field is
-        // always first.
-        // This means that we can only generically accumulate bounds for 2D indexes over the
-        // first "2d" field (pos == 0) - MatchExpressions over other fields in the 2D index may
-        // be covered (can be evaluated using only the 2D index key).  The additional fields
-        // must not affect the index scan bounds, since they are not stored in an
-        // IndexScan-compatible format.
+        // 2D indexes have a special format - the "2d" field stores a normally-indexed BinData
+        // field, but additional array fields are *not* exploded into multi-keys - they are stored
+        // directly as arrays in the index.  Also, no matter what the index expression, the "2d"
+        // field is always first.
+        //
+        // This means that we can only generically accumulate bounds for 2D indexes over the first
+        // "2d" field (pos == 0) - MatchExpressions over other fields in the 2D index may be covered
+        // (can be evaluated using only the 2D index key).  The additional fields must not affect
+        // the index scan bounds, since they are not stored in an IndexScan-compatible format.
 
         if (pos > 0) {
-            // Marking this field as covered allows the planner to accumulate a MatchExpression
-            // over the returned 2D index keys instead of adding to the index bounds.
-            scanState->tightness = IndexBoundsBuilder::INEXACT_COVERED;
+            // The predicate is over a trailing field of the "2d" index. If possible, we assign it
+            // as a covered filter (the INEXACT_COVERED case). Otherwise, the filter must be
+            // evaluated after fetching the full documents.
+            scanState->tightness = IndexBoundsBuilder::canUseCoveredMatching(expr, index)
+                ? IndexBoundsBuilder::INEXACT_COVERED
+                : IndexBoundsBuilder::INEXACT_FETCH;
             return;
         }
 
@@ -374,10 +402,15 @@ void QueryPlannerAccess::mergeWithLeafNode(MatchExpression* expr, ScanBuildingSt
         verify(type == STAGE_IXSCAN);
         IndexScanNode* scan = static_cast<IndexScanNode*>(node);
 
-        // See STAGE_GEO_NEAR_2D above - 2D indexes can only accumulate scan bounds over the
-        // first "2d" field (pos == 0)
+        // See STAGE_GEO_NEAR_2D above - 2D indexes can only accumulate scan bounds over the first
+        // "2d" field (pos == 0).
         if (INDEX_2D == index.type && pos > 0) {
-            scanState->tightness = IndexBoundsBuilder::INEXACT_COVERED;
+            // The predicate is over a trailing field of the "2d" index. If possible, we assign it
+            // as a covered filter (the INEXACT_COVERED case). Otherwise, the filter must be
+            // evaluated after fetching the full documents.
+            scanState->tightness = IndexBoundsBuilder::canUseCoveredMatching(expr, index)
+                ? IndexBoundsBuilder::INEXACT_COVERED
+                : IndexBoundsBuilder::INEXACT_FETCH;
             return;
         }
 
@@ -416,25 +449,9 @@ void QueryPlannerAccess::mergeWithLeafNode(MatchExpression* expr, ScanBuildingSt
 void QueryPlannerAccess::finishTextNode(QuerySolutionNode* node, const IndexEntry& index) {
     TextNode* tn = static_cast<TextNode*>(node);
 
-    // Figure out what positions are prefix positions.  We build an index key prefix from
-    // the predicates over the text index prefix keys.
-    // For example, say keyPattern = { a: 1, _fts: "text", _ftsx: 1, b: 1 }
-    // prefixEnd should be 1.
-    size_t prefixEnd = 0;
-    BSONObjIterator it(tn->index.keyPattern);
-    // Count how many prefix terms we have.
-    while (it.more()) {
-        // We know that the only key pattern with a type of String is the _fts field
-        // which is immediately after all prefix fields.
-        if (String == it.next().type()) {
-            break;
-        }
-        ++prefixEnd;
-    }
-
     // If there's no prefix, the filter is already on the node and the index prefix is null.
     // We can just return.
-    if (!prefixEnd) {
+    if (!tn->numPrefixFields) {
         return;
     }
 
@@ -448,7 +465,7 @@ void QueryPlannerAccess::finishTextNode(QuerySolutionNode* node, const IndexEntr
 
     if (MatchExpression::AND != textFilterMe->matchType()) {
         // Only one prefix term.
-        invariant(1 == prefixEnd);
+        invariant(1u == tn->numPrefixFields);
         // Sanity check: must be an EQ.
         invariant(MatchExpression::EQ == textFilterMe->matchType());
 
@@ -460,10 +477,10 @@ void QueryPlannerAccess::finishTextNode(QuerySolutionNode* node, const IndexEntr
 
         // Indexed by the keyPattern position index assignment.  We want to add
         // prefixes in order but we must order them first.
-        vector<MatchExpression*> prefixExprs(prefixEnd, NULL);
+        vector<MatchExpression*> prefixExprs(tn->numPrefixFields, nullptr);
 
         AndMatchExpression* amExpr = static_cast<AndMatchExpression*>(textFilterMe);
-        invariant(amExpr->numChildren() >= prefixEnd);
+        invariant(amExpr->numChildren() >= tn->numPrefixFields);
 
         // Look through the AND children.  The prefix children we want to
         // stash in prefixExprs.
@@ -474,7 +491,7 @@ void QueryPlannerAccess::finishTextNode(QuerySolutionNode* node, const IndexEntr
             invariant(NULL != ixtag);
             // Skip this child if it's not part of a prefix, or if we've already assigned a
             // predicate to this prefix position.
-            if (ixtag->pos >= prefixEnd || prefixExprs[ixtag->pos] != NULL) {
+            if (ixtag->pos >= tn->numPrefixFields || prefixExprs[ixtag->pos] != NULL) {
                 ++curChild;
                 continue;
             }
diff --git a/src/mongo/db/query/planner_ixselect.cpp b/src/mongo/db/query/planner_ixselect.cpp
index 321e856d1b7..2bf2eedc25e 100644
--- a/src/mongo/db/query/planner_ixselect.cpp
+++ b/src/mongo/db/query/planner_ixselect.cpp
@@ -48,6 +48,28 @@
 
 namespace mongo {
 
+namespace {
+
+/**
+ * Checks whether the given index is compatible with each child of the given $elemMatch expression.
+ * Assumes that the match expression is of type ELEM_MATCH_VALUE.
+ */
+bool elemMatchValueCompatible(const BSONElement& elt,
+                              const IndexEntry& index,
+                              MatchExpression* elemMatch,
+                              const CollatorInterface* collator) {
+    invariant(elemMatch->matchType() == MatchExpression::ELEM_MATCH_VALUE);
+    for (size_t child = 0; child < elemMatch->numChildren(); ++child) {
+        if (!QueryPlannerIXSelect::compatible(
+                elt, index, elemMatch->getChild(child), collator, true)) {
+            return false;
+        }
+    }
+    return true;
+}
+
+}  // namespace
+
 static double fieldWithDefault(const BSONObj& infoObj, const string& name, double def) {
     BSONElement e = infoObj[name];
     if (e.isNumber()) {
@@ -168,7 +190,8 @@ void QueryPlannerIXSelect::findRelevantIndices(const unordered_set<string>& fiel
 bool QueryPlannerIXSelect::compatible(const BSONElement& elt,
                                       const IndexEntry& index,
                                       MatchExpression* node,
-                                      const CollatorInterface* collator) {
+                                      const CollatorInterface* collator,
+                                      bool elemMatchChild) {
     if ((boundsGeneratingNodeContainsComparisonToType(node, BSONType::String) ||
          boundsGeneratingNodeContainsComparisonToType(node, BSONType::Array) ||
          boundsGeneratingNodeContainsComparisonToType(node, BSONType::Object)) &&
@@ -195,16 +218,18 @@ bool QueryPlannerIXSelect::compatible(const BSONElement& elt,
     MatchExpression::MatchType exprtype = node->matchType();
 
     if (indexedFieldType.empty()) {
-        // Can't check for null w/a sparse index.
-        if (exprtype == MatchExpression::EQ && index.sparse) {
+        // Can't use a sparse index for $eq with a null element, unless the equality is within a
+        // $elemMatch expression since the latter implies a match on the literal element 'null'.
+        if (exprtype == MatchExpression::EQ && index.sparse && !elemMatchChild) {
             const EqualityMatchExpression* expr = static_cast<const EqualityMatchExpression*>(node);
             if (expr->getData().isNull()) {
                 return false;
             }
         }
 
-        // Can't check for $in w/ null element w/a sparse index.
-        if (exprtype == MatchExpression::MATCH_IN && index.sparse) {
+        // Can't use a sparse index for $in with a null element, unless the $eq is within a
+        // $elemMatch expression since the latter implies a match on the literal element 'null'.
+        if (exprtype == MatchExpression::MATCH_IN && index.sparse && !elemMatchChild) {
             const InMatchExpression* expr = static_cast<const InMatchExpression*>(node);
             if (expr->hasNull()) {
                 return false;
@@ -220,7 +245,10 @@ bool QueryPlannerIXSelect::compatible(const BSONElement& elt,
         // the expression is a NOT.
         if (exprtype == MatchExpression::NOT) {
             // Don't allow indexed NOT on special index types such as geo or text indices.
-            if (INDEX_BTREE != index.type) {
+            // TODO: SERVER-30994 should remove this check entirely and allow $not on the
+            // 'non-special' fields of non-btree indices.
+            // (e.g. {a: 1, geo: "2dsphere"})
+            if (INDEX_BTREE != index.type && !elemMatchChild) {
                 return false;
             }
 
@@ -386,12 +414,19 @@ void QueryPlannerIXSelect::rateIndices(MatchExpression* node,
             BSONObjIterator it(indices[i].keyPattern);
             BSONElement elt = it.next();
             if (elt.fieldName() == fullPath && compatible(elt, indices[i], node, collator)) {
-                rt->first.push_back(i);
+                if (node->matchType() != MatchExpression::ELEM_MATCH_VALUE ||
+                    elemMatchValueCompatible(elt, indices[i], node, collator)) {
+                    rt->first.push_back(i);
+                }
             }
             while (it.more()) {
                 elt = it.next();
-                if (elt.fieldName() == fullPath && compatible(elt, indices[i], node, collator)) {
-                    rt->notFirst.push_back(i);
+                if (elt.fieldName() == fullPath &&
+                    compatible(elt, indices[i], node, collator, false)) {
+                    if (node->matchType() != MatchExpression::ELEM_MATCH_VALUE ||
+                        elemMatchValueCompatible(elt, indices[i], node, collator)) {
+                        rt->notFirst.push_back(i);
+                    }
                 }
             }
         }
diff --git a/src/mongo/db/query/planner_ixselect.h b/src/mongo/db/query/planner_ixselect.h
index 914e9265d9b..7cb34973810 100644
--- a/src/mongo/db/query/planner_ixselect.h
+++ b/src/mongo/db/query/planner_ixselect.h
@@ -72,7 +72,8 @@ public:
     static bool compatible(const BSONElement& elt,
                            const IndexEntry& index,
                            MatchExpression* node,
-                           const CollatorInterface* collator);
+                           const CollatorInterface* collator,
+                           bool elemMatchChild = false);
 
     /**
      * Determine how useful all of our relevant 'indices' are to all predicates in the subtree
diff --git a/src/mongo/db/query/planner_ixselect_test.cpp b/src/mongo/db/query/planner_ixselect_test.cpp
index 1c273db8716..0db164a68be 100644
--- a/src/mongo/db/query/planner_ixselect_test.cpp
+++ b/src/mongo/db/query/planner_ixselect_test.cpp
@@ -45,6 +45,8 @@ using namespace mongo;
 
 namespace {
 
+constexpr CollatorInterface* kSimpleCollator = nullptr;
+
 using std::unique_ptr;
 using std::string;
 using std::vector;
@@ -304,6 +306,83 @@ TEST(QueryPlannerIXSelectTest, RateIndicesTaggedNodePathArrayNegation) {
 }
 
 /**
+ * $not within $elemMatch should not attempt to use a sparse index for $exists:false.
+ */
+TEST(QueryPlannerIXSelectTest, ElemMatchNotExistsShouldNotUseSparseIndex) {
+    std::vector<IndexEntry> indices;
+    auto idxEntry = IndexEntry(BSON("a" << 1));
+    idxEntry.sparse = true;
+    indices.push_back(idxEntry);
+    std::set<size_t> expectedIndices;
+    testRateIndices("{a: {$elemMatch: {$not: {$exists: true}}}}",
+                    "",
+                    kSimpleCollator,
+                    indices,
+                    "",
+                    expectedIndices);
+}
+
+/**
+ * $in with a null value within $elemMatch can use a sparse index.
+ */
+TEST(QueryPlannerIXSelectTest, ElemMatchInNullValueShouldUseSparseIndex) {
+    std::vector<IndexEntry> indices;
+    auto idxEntry = IndexEntry(BSON("a" << 1));
+    idxEntry.sparse = true;
+    indices.push_back(idxEntry);
+    std::set<size_t> expectedIndices = {0};
+    testRateIndices(
+        "{a: {$elemMatch: {$in: [null]}}}", "", kSimpleCollator, indices, "a", expectedIndices);
+}
+
+/**
+ * $geo queries within $elemMatch should not use a normal B-tree index.
+ */
+TEST(QueryPlannerIXSelectTest, ElemMatchGeoShouldNotUseBtreeIndex) {
+    std::vector<IndexEntry> indices;
+    auto idxEntry = IndexEntry(BSON("a" << 1));
+    indices.push_back(idxEntry);
+    std::set<size_t> expectedIndices;
+    testRateIndices(R"({a: {$elemMatch: {$geoWithin: {$geometry: {type: 'Polygon', 
+                      coordinates: [[[0,0],[0,1],[1,0],[0,0]]]}}}}})",
+                    "",
+                    kSimpleCollator,
+                    indices,
+                    "a",
+                    expectedIndices);
+}
+
+/**
+ * $eq with a null value within $elemMatch can use a sparse index.
+ */
+TEST(QueryPlannerIXSelectTest, ElemMatchEqNullValueShouldUseSparseIndex) {
+    std::vector<IndexEntry> indices;
+    auto idxEntry = IndexEntry(BSON("a" << 1));
+    idxEntry.sparse = true;
+    indices.push_back(idxEntry);
+    std::set<size_t> expectedIndices = {0};
+    testRateIndices(
+        "{a: {$elemMatch: {$eq: null}}}", "", kSimpleCollator, indices, "a", expectedIndices);
+}
+
+/**
+ * $elemMatch with multiple children will not use an index if any child is incompatible.
+ */
+TEST(QueryPlannerIXSelectTest, ElemMatchMultipleChildrenShouldRequireAllToBeCompatible) {
+    std::vector<IndexEntry> indices;
+    auto idxEntry = IndexEntry(BSON("a" << 1));
+    idxEntry.sparse = true;
+    indices.push_back(idxEntry);
+    std::set<size_t> expectedIndices;
+    testRateIndices("{a: {$elemMatch: {$eq: null, $not: {$exists: true}}}}",
+                    "",
+                    kSimpleCollator,
+                    indices,
+                    "",
+                    expectedIndices);
+}
+
+/**
  * If the collator is null, we select the relevant index with a null collator.
  */
 TEST(QueryPlannerIXSelectTest, NullCollatorsMatch) {
diff --git a/src/mongo/db/query/query_planner_geo_test.cpp b/src/mongo/db/query/query_planner_geo_test.cpp
index 9e087444079..0421546a60d 100644
--- a/src/mongo/db/query/query_planner_geo_test.cpp
+++ b/src/mongo/db/query/query_planner_geo_test.cpp
@@ -1531,4 +1531,66 @@ TEST_F(QueryPlanner2dsphereVersionTest, NegationWithoutGeoPredCannotUseGeoIndex)
     testMultiple2dsphereIndexVersions(versions, keyPatterns, predicate, solutions);
 }
 
+TEST_F(QueryPlannerTest, 2dInexactFetchPredicateOverTrailingFieldHandledCorrectly) {
+    params.options = QueryPlannerParams::NO_TABLE_SCAN;
+
+    addIndex(BSON("a"
+                  << "2d"
+                  << "b"
+                  << 1));
+
+    runQuery(fromjson("{a: {$geoWithin: {$center: [[0, 0], 1]}}, b: {$exists: true}}"));
+    assertNumSolutions(1U);
+    assertSolutionExists(
+        "{fetch: {filter: {a: {$geoWithin: {$center: [[0, 0], 1]}}, b: {$exists: true}}, node: "
+        "{ixscan: {filter: null, pattern: {a: '2d', b: 1}}}}}");
+}
+
+TEST_F(QueryPlannerTest, 2dInexactFetchPredicateOverTrailingFieldHandledCorrectlyMultikey) {
+    params.options = QueryPlannerParams::NO_TABLE_SCAN;
+
+    const bool multikey = true;
+    addIndex(BSON("a"
+                  << "2d"
+                  << "b"
+                  << 1),
+             multikey);
+
+    runQuery(fromjson("{a: {$geoWithin: {$center: [[0, 0], 1]}}, b: {$exists: true}}"));
+    assertNumSolutions(1U);
+    assertSolutionExists(
+        "{fetch: {filter: {a: {$geoWithin: {$center: [[0, 0], 1]}}, b: {$exists: true}}, node: "
+        "{ixscan: {filter: null, pattern: {a: '2d', b: 1}}}}}");
+}
+
+TEST_F(QueryPlannerTest, 2dNearInexactFetchPredicateOverTrailingFieldHandledCorrectly) {
+    params.options = QueryPlannerParams::NO_TABLE_SCAN;
+
+    addIndex(BSON("a"
+                  << "2d"
+                  << "b"
+                  << 1));
+
+    runQuery(fromjson("{a: {$near: [0, 0]}, b: {$exists: true}}"));
+    assertNumSolutions(1U);
+    assertSolutionExists(
+        "{fetch: {filter: {b: {$exists: true}}, node: {geoNear2d: {a: '2d', b: 1}}}}");
+}
+
+TEST_F(QueryPlannerTest, 2dNearInexactFetchPredicateOverTrailingFieldMultikey) {
+    params.options = QueryPlannerParams::NO_TABLE_SCAN;
+
+    const bool multikey = true;
+    addIndex(BSON("a"
+                  << "2d"
+                  << "b"
+                  << 1),
+             multikey);
+
+    runQuery(fromjson("{a: {$near: [0, 0]}, b: {$exists: true}}"));
+    assertNumSolutions(1U);
+    assertSolutionExists(
+        "{fetch: {filter: {b: {$exists: true}}, node: {geoNear2d: {a: '2d', b: 1}}}}");
+}
+
 }  // namespace
diff --git a/src/mongo/db/query/query_planner_text_test.cpp b/src/mongo/db/query/query_planner_text_test.cpp
index 5050653292a..cbb7fcdbd39 100644
--- a/src/mongo/db/query/query_planner_text_test.cpp
+++ b/src/mongo/db/query/query_planner_text_test.cpp
@@ -461,4 +461,77 @@ TEST_F(QueryPlannerTest, SortKeyMetaProjectionWithTextScoreMetaSort) {
         "{sortKeyGen: {node: {text: {search: 'foo'}}}}}}}}");
 }
 
+TEST_F(QueryPlannerTest, PredicatesOverLeadingFieldsWithSharedPathPrefixHandledCorrectly) {
+    const bool multikey = true;
+    addIndex(BSON("a.x" << 1 << "a.y" << 1 << "b.x" << 1 << "b.y" << 1 << "_fts"
+                        << "text"
+                        << "_ftsx"
+                        << 1),
+             multikey);
+
+    runQuery(fromjson("{'a.x': 1, 'a.y': 2, 'b.x': 3, 'b.y': 4, $text: {$search: 'foo'}}"));
+
+    assertNumSolutions(1U);
+    assertSolutionExists(
+        "{text: {search: 'foo', prefix: {'a.x': 1, 'a.y': 2, 'b.x': 3, 'b.y': 4}}}");
+}
+
+TEST_F(QueryPlannerTest, EqualityToArrayOverLeadingFieldHandledCorrectly) {
+    addIndex(BSON("a" << 1 << "_fts"
+                      << "text"
+                      << "_ftsx"
+                      << 1));
+
+    runQuery(fromjson("{a: [1, 2, 3], $text: {$search: 'foo'}}"));
+
+    assertNumSolutions(1U);
+    assertSolutionExists("{text: {search: 'foo', prefix: {a: [1, 2, 3]}}}");
+}
+
+TEST_F(QueryPlannerTest, EqualityToArrayOverLeadingFieldHandledCorrectlyWithMultikeyTrue) {
+    const bool multikey = true;
+    addIndex(BSON("a" << 1 << "_fts"
+                      << "text"
+                      << "_ftsx"
+                      << 1),
+             multikey);
+
+    runQuery(fromjson("{a: [1, 2, 3], $text: {$search: 'foo'}}"));
+
+    assertNumSolutions(1U);
+    assertSolutionExists("{text: {search: 'foo', prefix: {a: [1, 2, 3]}}}");
+}
+
+TEST_F(QueryPlannerTest, InexactFetchPredicateOverTrailingFieldHandledCorrectly) {
+    addIndex(BSON("a" << 1 << "_fts"
+                      << "text"
+                      << "_ftsx"
+                      << 1
+                      << "b"
+                      << 1));
+
+    runQuery(fromjson("{a: 3, $text: {$search: 'foo'}, b: {$exists: true}}"));
+
+    assertNumSolutions(1U);
+    assertSolutionExists(
+        "{fetch: {filter: {b: {$exists: true}}, node: {text: {search: 'foo', prefix: {a: 3}}}}}");
+}
+
+TEST_F(QueryPlannerTest, InexactFetchPredicateOverTrailingFieldHandledCorrectlyMultikeyTrue) {
+    const bool multikey = true;
+    addIndex(BSON("a" << 1 << "_fts"
+                      << "text"
+                      << "_ftsx"
+                      << 1
+                      << "b"
+                      << 1),
+             multikey);
+
+    runQuery(fromjson("{a: 3, $text: {$search: 'foo'}, b: {$exists: true}}"));
+
+    assertNumSolutions(1U);
+    assertSolutionExists(
+        "{fetch: {filter: {b: {$exists: true}}, node: {text: {search: 'foo', prefix: {a: 3}}}}}");
+}
+
 }  // namespace
diff --git a/src/mongo/db/query/query_solution.h b/src/mongo/db/query/query_solution.h
index 7e68378042d..ff02e0df3a5 100644
--- a/src/mongo/db/query/query_solution.h
+++ b/src/mongo/db/query/query_solution.h
@@ -250,6 +250,13 @@ struct TextNode : public QuerySolutionNode {
     IndexEntry index;
     std::unique_ptr<fts::FTSQuery> ftsQuery;
 
+    // The number of fields in the prefix of the text index. For example, if the key pattern is
+    //
+    //   { a: 1, b: 1, _fts: "text", _ftsx: 1, c: 1 }
+    //
+    // then the number of prefix fields is 2, because of "a" and "b".
+    size_t numPrefixFields = 0u;
+
     // "Prefix" fields of a text index can handle equality predicates.  We group them with the
     // text node while creating the text leaf node and convert them into a BSONObj index prefix
     // when we finish the text leaf node.
diff --git a/src/mongo/db/range_deleter.cpp b/src/mongo/db/range_deleter.cpp
index b97d9fae216..e0d1d5c5aa0 100644
--- a/src/mongo/db/range_deleter.cpp
+++ b/src/mongo/db/range_deleter.cpp
@@ -354,16 +354,13 @@ bool RangeDeleter::deleteNow(OperationContext* txn,
     taskDetails.stats.queueEndTS = jsTime();
 
     taskDetails.stats.deleteStartTS = jsTime();
-    bool result = _env->deleteRange(txn, taskDetails, &taskDetails.stats.deletedDocCount, errMsg);
-
+    bool result = _env->deleteRange(txn,
+                                    taskDetails,
+                                    &taskDetails.stats.deletedDocCount,
+                                    taskDetails.stats.waitForReplDurationMs,
+                                    errMsg);
     taskDetails.stats.deleteEndTS = jsTime();
 
-    if (result) {
-        taskDetails.stats.waitForReplStartTS = jsTime();
-        result = _waitForMajority(txn, errMsg);
-        taskDetails.stats.waitForReplEndTS = jsTime();
-    }
-
     {
         stdx::lock_guard<stdx::mutex> sl(_queueMutex);
         _deleteSet.erase(&deleteRange);
@@ -485,8 +482,11 @@ void RangeDeleter::doWork() {
         {
             auto txn = client->makeOperationContext();
             nextTask->stats.deleteStartTS = jsTime();
-            bool delResult =
-                _env->deleteRange(txn.get(), *nextTask, &nextTask->stats.deletedDocCount, &errMsg);
+            bool delResult = _env->deleteRange(txn.get(),
+                                               *nextTask,
+                                               &nextTask->stats.deletedDocCount,
+                                               nextTask->stats.waitForReplDurationMs,
+                                               &errMsg);
             nextTask->stats.deleteEndTS = jsTime();
 
             if (delResult) {
diff --git a/src/mongo/db/range_deleter.h b/src/mongo/db/range_deleter.h
index 776f8a825fc..b6acafe3112 100644
--- a/src/mongo/db/range_deleter.h
+++ b/src/mongo/db/range_deleter.h
@@ -254,6 +254,7 @@ struct DeleteJobStats {
     Date_t deleteEndTS;
     Date_t waitForReplStartTS;
     Date_t waitForReplEndTS;
+    Milliseconds waitForReplDurationMs;
 
     long long int deletedDocCount;
 
@@ -314,6 +315,7 @@ struct RangeDeleterEnv {
     virtual bool deleteRange(OperationContext* txn,
                              const RangeDeleteEntry& taskDetails,
                              long long int* deletedDocs,
+                             Milliseconds& replWaitDuration,
                              std::string* errMsg) = 0;
 
     /**
diff --git a/src/mongo/db/range_deleter_db_env.cpp b/src/mongo/db/range_deleter_db_env.cpp
index 2f692111c0a..421cafc9692 100644
--- a/src/mongo/db/range_deleter_db_env.cpp
+++ b/src/mongo/db/range_deleter_db_env.cpp
@@ -61,6 +61,7 @@ using std::string;
 bool RangeDeleterDBEnv::deleteRange(OperationContext* txn,
                                     const RangeDeleteEntry& taskDetails,
                                     long long int* deletedDocs,
+                                    Milliseconds& replWaitDuration,
                                     std::string* errMsg) {
     const string ns(taskDetails.options.range.ns);
     const BSONObj inclusiveLower(taskDetails.options.range.minKey);
@@ -92,6 +93,7 @@ bool RangeDeleterDBEnv::deleteRange(OperationContext* txn,
                                  KeyRange(ns, inclusiveLower, exclusiveUpper, keyPattern),
                                  BoundInclusion::kIncludeStartKeyOnly,
                                  writeConcern,
+                                 replWaitDuration,
                                  removeSaverPtr,
                                  fromMigrate,
                                  onlyRemoveOrphans);
diff --git a/src/mongo/db/range_deleter_db_env.h b/src/mongo/db/range_deleter_db_env.h
index 0bca8c6618c..b0e5a11d0cc 100644
--- a/src/mongo/db/range_deleter_db_env.h
+++ b/src/mongo/db/range_deleter_db_env.h
@@ -46,13 +46,16 @@ struct RangeDeleterDBEnv : public RangeDeleterEnv {
      * Note that secondaryThrottle will be ignored if current process is not part
      * of a replica set.
      *
-     * docsDeleted would contain the number of docs deleted if the deletion was successful.
+     * deletedDocs would contain the number of docs deleted if the deletion was successful.
+     *
+     * Returns time spent waiting for majority replication in replWaitDuration.
      *
      * Does not throw Exceptions.
      */
     virtual bool deleteRange(OperationContext* txn,
                              const RangeDeleteEntry& taskDetails,
                              long long int* deletedDocs,
+                             Milliseconds& replWaitDuration,
                              std::string* errMsg);
 
     /**
diff --git a/src/mongo/db/range_deleter_mock_env.cpp b/src/mongo/db/range_deleter_mock_env.cpp
index 1ee01d6a19a..9c50185ba07 100644
--- a/src/mongo/db/range_deleter_mock_env.cpp
+++ b/src/mongo/db/range_deleter_mock_env.cpp
@@ -104,6 +104,7 @@ DeletedRange RangeDeleterMockEnv::getLastDelete() const {
 bool RangeDeleterMockEnv::deleteRange(OperationContext* txn,
                                       const RangeDeleteEntry& taskDetails,
                                       long long int* deletedDocs,
+                                      Milliseconds& replWaitDuration,
                                       string* errMsg) {
     {
         stdx::unique_lock<stdx::mutex> sl(_pauseDeleteMutex);
diff --git a/src/mongo/db/range_deleter_mock_env.h b/src/mongo/db/range_deleter_mock_env.h
index 307dbe2bd53..5792b24499c 100644
--- a/src/mongo/db/range_deleter_mock_env.h
+++ b/src/mongo/db/range_deleter_mock_env.h
@@ -127,6 +127,7 @@ public:
     bool deleteRange(OperationContext* txn,
                      const RangeDeleteEntry& taskDetails,
                      long long int* deletedDocs,
+                     Milliseconds& replWaitDuration,
                      std::string* errMsg);
 
     /**
diff --git a/src/mongo/db/repl/SConscript b/src/mongo/db/repl/SConscript
index 05dbeff18a0..31f35d384f8 100644
--- a/src/mongo/db/repl/SConscript
+++ b/src/mongo/db/repl/SConscript
@@ -631,12 +631,12 @@ env.Library('replica_set_messages',
                 'is_master_response.cpp',
                 'member_config.cpp',
                 'old_update_position_args.cpp',
+                'repl_set_config.cpp',
                 'repl_set_heartbeat_args.cpp',
                 'repl_set_heartbeat_args_v1.cpp',
                 'repl_set_heartbeat_response.cpp',
                 'repl_set_html_summary.cpp',
                 'repl_set_request_votes_args.cpp',
-                'repl_set_config.cpp',
                 'repl_set_tag.cpp',
                 'update_position_args.cpp',
                 'last_vote.cpp',
diff --git a/src/mongo/db/repl/bgsync.cpp b/src/mongo/db/repl/bgsync.cpp
index 2c04c40bf55..fbbf1c56efd 100644
--- a/src/mongo/db/repl/bgsync.cpp
+++ b/src/mongo/db/repl/bgsync.cpp
@@ -225,15 +225,16 @@ void BackgroundSync::_runProducer() {
     }
     // we want to start when we're no longer primary
     // start() also loads _lastOpTimeFetched, which we know is set from the "if"
-    auto txn = cc().makeOperationContext();
-    if (getState() == ProducerState::Starting) {
-        start(txn.get());
+    {
+        auto opCtx = cc().makeOperationContext();
+        if (getState() == ProducerState::Starting) {
+            start(opCtx.get());
+        }
     }
-
-    _produce(txn.get());
+    _produce();
 }
 
-void BackgroundSync::_produce(OperationContext* opCtx) {
+void BackgroundSync::_produce() {
     if (MONGO_FAIL_POINT(stopReplProducer)) {
         // This log output is used in js tests so please leave it.
         log() << "bgsync - stopReplProducer fail point "
@@ -266,15 +267,18 @@ void BackgroundSync::_produce(OperationContext* opCtx) {
         }
     }
 
-    auto storageInterface = StorageInterface::get(opCtx);
     // find a target to sync from the last optime fetched
     OpTime lastOpTimeFetched;
     HostAndPort source;
     HostAndPort oldSource = _syncSourceHost;
     SyncSourceResolverResponse syncSourceResp;
     {
-        const OpTime minValidSaved = storageInterface->getMinValid(opCtx);
-
+        OpTime minValidSaved;
+        {
+            auto opCtx = cc().makeOperationContext();
+            auto storageInterface = StorageInterface::get(opCtx.get());
+            minValidSaved = storageInterface->getMinValid(opCtx.get());
+        }
         stdx::lock_guard<stdx::mutex> lock(_mutex);
         if (_state != ProducerState::Running) {
             return;
@@ -397,8 +401,12 @@ void BackgroundSync::_produce(OperationContext* opCtx) {
     // Set the applied point if unset. This is most likely the first time we've established a sync
     // source since stepping down or otherwise clearing the applied point. We need to set this here,
     // before the OplogWriter gets a chance to append to the oplog.
-    if (storageInterface->getAppliedThrough(opCtx).isNull()) {
-        storageInterface->setAppliedThrough(opCtx, _replCoord->getMyLastAppliedOpTime());
+    {
+        auto opCtx = cc().makeOperationContext();
+        auto storageInterface = StorageInterface::get(opCtx.get());
+        if (storageInterface->getAppliedThrough(opCtx.get()).isNull()) {
+            storageInterface->setAppliedThrough(opCtx.get(), _replCoord->getMyLastAppliedOpTime());
+        }
     }
 
     // "lastFetched" not used. Already set in _enqueueDocuments.
@@ -521,10 +529,18 @@ void BackgroundSync::_produce(OperationContext* opCtx) {
             }
         }
 
-        OplogInterfaceLocal localOplog(opCtx, rsOplogName);
-        RollbackSourceImpl rollbackSource(getConnection, source, rsOplogName);
-        rollback(
-            opCtx, localOplog, rollbackSource, syncSourceResp.rbid, _replCoord, storageInterface);
+        {
+            auto opCtx = cc().makeOperationContext();
+            OplogInterfaceLocal localOplog(opCtx.get(), rsOplogName);
+            RollbackSourceImpl rollbackSource(getConnection, source, rsOplogName);
+            auto storageInterface = StorageInterface::get(opCtx.get());
+            rollback(opCtx.get(),
+                     localOplog,
+                     rollbackSource,
+                     syncSourceResp.rbid,
+                     _replCoord,
+                     storageInterface);
+        }
 
         // Reset the producer to clear the sync source and the last optime fetched.
         stop(true);
diff --git a/src/mongo/db/repl/bgsync.h b/src/mongo/db/repl/bgsync.h
index 6d068967723..00f23e59b1d 100644
--- a/src/mongo/db/repl/bgsync.h
+++ b/src/mongo/db/repl/bgsync.h
@@ -148,7 +148,7 @@ private:
     void _run();
     // Production thread inner loop.
     void _runProducer();
-    void _produce(OperationContext* txn);
+    void _produce();
 
     /**
      * Checks current background sync state before pushing operations into blocking queue and
diff --git a/src/mongo/db/repl/collection_cloner.cpp b/src/mongo/db/repl/collection_cloner.cpp
index 9dc3674efa0..bbaf04ba3dc 100644
--- a/src/mongo/db/repl/collection_cloner.cpp
+++ b/src/mongo/db/repl/collection_cloner.cpp
@@ -114,7 +114,8 @@ CollectionCloner::CollectionCloner(executor::TaskExecutor* executor,
                                      stdx::placeholders::_2,
                                      stdx::placeholders::_3),
                           rpc::ServerSelectionMetadata(true, boost::none).toBSON(),
-                          RemoteCommandRequest::kNoTimeout,
+                          RemoteCommandRequest::kNoTimeout /* find network timeout */,
+                          RemoteCommandRequest::kNoTimeout /* getMore network timeout */,
                           RemoteCommandRetryScheduler::makeRetryPolicy(
                               numInitialSyncListIndexesAttempts,
                               executor::RemoteCommandRequest::kNoTimeout,
@@ -518,7 +519,8 @@ void CollectionCloner::_beginCollectionCallback(const executor::TaskExecutor::Ca
                    stdx::placeholders::_3,
                    onCompletionGuard),
         rpc::ServerSelectionMetadata(true, boost::none).toBSON(),
-        RemoteCommandRequest::kNoTimeout,
+        RemoteCommandRequest::kNoTimeout /* find network timeout */,
+        RemoteCommandRequest::kNoTimeout /* getMore network timeout */,
         RemoteCommandRetryScheduler::makeRetryPolicy(
             numInitialSyncCollectionFindAttempts.load(),
             executor::RemoteCommandRequest::kNoTimeout,
diff --git a/src/mongo/db/repl/database_cloner.cpp b/src/mongo/db/repl/database_cloner.cpp
index 461c4dfe1b7..d0211fdf058 100644
--- a/src/mongo/db/repl/database_cloner.cpp
+++ b/src/mongo/db/repl/database_cloner.cpp
@@ -114,7 +114,8 @@ DatabaseCloner::DatabaseCloner(executor::TaskExecutor* executor,
                                          stdx::placeholders::_2,
                                          stdx::placeholders::_3),
                               rpc::ServerSelectionMetadata(true, boost::none).toBSON(),
-                              RemoteCommandRequest::kNoTimeout,
+                              RemoteCommandRequest::kNoTimeout /* find network timeout */,
+                              RemoteCommandRequest::kNoTimeout /* getMore network timeout */,
                               RemoteCommandRetryScheduler::makeRetryPolicy(
                                   numInitialSyncListCollectionsAttempts,
                                   executor::RemoteCommandRequest::kNoTimeout,
diff --git a/src/mongo/db/repl/databases_cloner.cpp b/src/mongo/db/repl/databases_cloner.cpp
index 4a18e1be1b8..c9016c322f4 100644
--- a/src/mongo/db/repl/databases_cloner.cpp
+++ b/src/mongo/db/repl/databases_cloner.cpp
@@ -231,6 +231,44 @@ void DatabasesCloner::setScheduleDbWorkFn_forTest(const CollectionCloner::Schedu
     _scheduleDbWorkFn = work;
 }
 
+StatusWith<std::vector<BSONElement>> DatabasesCloner::parseListDatabasesResponse_forTest(
+    BSONObj dbResponse) {
+    return _parseListDatabasesResponse(dbResponse);
+}
+
+void DatabasesCloner::setAdminAsFirst_forTest(std::vector<BSONElement>& dbsArray) {
+    _setAdminAsFirst(dbsArray);
+}
+
+StatusWith<std::vector<BSONElement>> DatabasesCloner::_parseListDatabasesResponse(
+    BSONObj dbResponse) {
+    if (!dbResponse.hasField("databases")) {
+        return Status(ErrorCodes::BadValue,
+                      "The 'listDatabases' response does not contain a 'databases' field.");
+    }
+    BSONElement response = dbResponse["databases"];
+    try {
+        return response.Array();
+    } catch (const MsgAssertionException& e) {
+        return Status(ErrorCodes::BadValue,
+                      "The 'listDatabases' response is unable to be transformed into an array.");
+    }
+}
+
+void DatabasesCloner::_setAdminAsFirst(std::vector<BSONElement>& dbsArray) {
+    auto adminIter = std::find_if(dbsArray.begin(), dbsArray.end(), [](BSONElement elem) {
+        if (!elem.isABSONObj()) {
+            return false;
+        }
+        auto bsonObj = elem.Obj();
+        std::string databaseName = bsonObj.getStringField("name");
+        return (databaseName == "admin");
+    });
+    if (adminIter != dbsArray.end()) {
+        std::iter_swap(adminIter, dbsArray.begin());
+    }
+}
+
 void DatabasesCloner::_onListDatabaseFinish(const CommandCallbackArgs& cbd) {
     Status respStatus = cbd.response.status;
     if (respStatus.isOK()) {
@@ -239,24 +277,44 @@ void DatabasesCloner::_onListDatabaseFinish(const CommandCallbackArgs& cbd) {
 
     UniqueLock lk(_mutex);
     if (!respStatus.isOK()) {
-        LOG(1) << "listDatabases failed: " << respStatus;
+        LOG(1) << "'listDatabases' failed: " << respStatus;
         _fail_inlock(&lk, respStatus);
         return;
     }
 
-    const auto respBSON = cbd.response.data;
-    // There should not be any cloners yet
+    // There should not be any cloners yet.
     invariant(_databaseCloners.size() == 0);
-    const auto dbsElem = respBSON["databases"].Obj();
-    BSONForEach(arrayElement, dbsElem) {
+    const auto respBSON = cbd.response.data;
+
+    auto databasesArray = _parseListDatabasesResponse(respBSON);
+    if (!databasesArray.isOK()) {
+        LOG(1) << "'listDatabases' returned a malformed response: "
+               << databasesArray.getStatus().toString();
+        _fail_inlock(&lk, databasesArray.getStatus());
+        return;
+    }
+
+    auto dbsArray = databasesArray.getValue();
+    // Ensure that the 'admin' database is the first element in the array of databases so that it
+    // will be the first to be cloned. This allows users to authenticate against a database while
+    // initial sync is occurring.
+    _setAdminAsFirst(dbsArray);
+
+    for (BSONElement arrayElement : dbsArray) {
         const BSONObj dbBSON = arrayElement.Obj();
 
         // Check to see if we want to exclude this db from the clone.
         if (!_includeDbFn(dbBSON)) {
-            LOG(1) << "excluding db: " << dbBSON;
+            LOG(1) << "Excluding database from the 'listDatabases' response: " << dbBSON;
             continue;
         }
 
+        if (!dbBSON.hasField("name")) {
+            LOG(1) << "Excluding database due to the 'listDatabases' response not containing a "
+                      "'name' field for this entry: "
+                   << dbBSON;
+        }
+
         const std::string dbName = dbBSON["name"].str();
         std::shared_ptr<DatabaseCloner> dbCloner{nullptr};
 
@@ -321,7 +379,6 @@ void DatabasesCloner::_onListDatabaseFinish(const CommandCallbackArgs& cbd) {
         // add cloner to list.
         _databaseCloners.push_back(dbCloner);
     }
-
     if (_databaseCloners.size() == 0) {
         if (_status.isOK()) {
             _succeed_inlock(&lk);
diff --git a/src/mongo/db/repl/databases_cloner.h b/src/mongo/db/repl/databases_cloner.h
index 53cced9ee4f..13f11f1e2a1 100644
--- a/src/mongo/db/repl/databases_cloner.h
+++ b/src/mongo/db/repl/databases_cloner.h
@@ -34,6 +34,7 @@
 
 #include "mongo/base/disallow_copying.h"
 #include "mongo/base/status.h"
+#include "mongo/base/status_with.h"
 #include "mongo/bson/bsonobj.h"
 #include "mongo/client/fetcher.h"
 #include "mongo/db/namespace_string.h"
@@ -104,6 +105,18 @@ public:
      */
     void setScheduleDbWorkFn_forTest(const CollectionCloner::ScheduleDbWorkFn& scheduleDbWorkFn);
 
+    /**
+     * Calls DatabasesCloner::_setAdminAsFirst.
+     * For testing only.
+     */
+    void setAdminAsFirst_forTest(std::vector<BSONElement>& dbsArray);
+
+    /**
+     * Calls DatabasesCloner::_parseListDatabasesResponse.
+     * For testing only.
+     */
+    StatusWith<std::vector<BSONElement>> parseListDatabasesResponse_forTest(BSONObj dbResponse);
+
 private:
     bool _isActive_inlock() const;
 
@@ -135,6 +148,23 @@ private:
 
     void _onListDatabaseFinish(const CommandCallbackArgs& cbd);
 
+    /**
+     * Takes a vector of BSONElements and scans for an element that contains a 'name' field with the
+     * value 'admin'. If found, the element is swapped with the first element in the vector.
+     * Otherwise, return.
+     *
+     * Used to parse the BSONResponse returned by listDatabases.
+     */
+    void _setAdminAsFirst(std::vector<BSONElement>& dbsArray);
+
+    /**
+     * Takes a 'listDatabases' command response and parses the response into a
+     * vector of BSON elements.
+     *
+     * If the input response is malformed, Status ErrorCodes::BadValue will be returned.
+     */
+    StatusWith<std::vector<BSONElement>> _parseListDatabasesResponse(BSONObj dbResponse);
+
     //
     // All member variables are labeled with one of the following codes indicating the
     // synchronization rules for accessing them.
diff --git a/src/mongo/db/repl/databases_cloner_test.cpp b/src/mongo/db/repl/databases_cloner_test.cpp
index 81552201485..e44114e5d58 100644
--- a/src/mongo/db/repl/databases_cloner_test.cpp
+++ b/src/mongo/db/repl/databases_cloner_test.cpp
@@ -314,6 +314,15 @@ protected:
         ASSERT_OK(result);
     };
 
+    std::unique_ptr<DatabasesCloner> makeDummyDatabasesCloner() {
+        return stdx::make_unique<DatabasesCloner>(&getStorage(),
+                                                  &getExecutor(),
+                                                  &getDbWorkThreadPool(),
+                                                  HostAndPort{"local:1234"},
+                                                  [](const BSONObj&) { return true; },
+                                                  [](const Status&) {});
+    }
+
 private:
     executor::ThreadPoolMock::Options makeThreadPoolMockOptions() const override;
 
@@ -450,6 +459,87 @@ TEST_F(DBsClonerTest, StartupReturnsInternalErrorAfterSuccessfulStartup) {
     ASSERT_TRUE(cloner.isActive());
 }
 
+TEST_F(DBsClonerTest, ParseAndSetAdminFirstWhenAdminInListDatabasesResponse) {
+    const Responses responsesWithAdmin = {
+        {"listDatabases", fromjson("{ok:1, databases:[{name:'a'}, {name:'aab'}, {name:'admin'}]}")},
+        {"listDatabases", fromjson("{ok:1, databases:[{name:'admin'}, {name:'a'}, {name:'b'}]}")},
+    };
+    std::unique_ptr<DatabasesCloner> cloner = makeDummyDatabasesCloner();
+
+    for (auto&& resp : responsesWithAdmin) {
+        auto parseResponseStatus = cloner->parseListDatabasesResponse_forTest(resp.second);
+        ASSERT_TRUE(parseResponseStatus.isOK());
+        std::vector<BSONElement> dbNamesArray = parseResponseStatus.getValue();
+        cloner->setAdminAsFirst_forTest(dbNamesArray);
+        ASSERT_EQUALS("admin", dbNamesArray[0].Obj().firstElement().str());
+    }
+}
+
+TEST_F(DBsClonerTest, ParseAndAttemptSetAdminFirstWhenAdminNotInListDatabasesResponse) {
+    const Responses responsesWithoutAdmin = {
+        {"listDatabases", fromjson("{ok:1, databases:[{name:'a'}, {name:'aab'}, {name:'abc'}]}")},
+        {"listDatabases", fromjson("{ok:1, databases:[{name:'foo'}, {name:'a'}, {name:'b'}]}")},
+        {"listDatabases", fromjson("{ok:1, databases:[{name:1}, {name:2}, {name:3}]}")},
+    };
+    std::unique_ptr<DatabasesCloner> cloner = makeDummyDatabasesCloner();
+
+    for (auto&& resp : responsesWithoutAdmin) {
+        auto parseResponseStatus = cloner->parseListDatabasesResponse_forTest(resp.second);
+        ASSERT_TRUE(parseResponseStatus.isOK());
+        std::vector<BSONElement> dbNamesArray = parseResponseStatus.getValue();
+        std::string expectedResult = dbNamesArray[0].Obj().firstElement().str();
+        cloner->setAdminAsFirst_forTest(dbNamesArray);
+        ASSERT_EQUALS(expectedResult, dbNamesArray[0].Obj().firstElement().str());
+    }
+}
+
+
+TEST_F(DBsClonerTest, ParseListDatabasesResponseWithMalformedResponses) {
+    Status expectedResultForNoDatabasesField{
+        ErrorCodes::BadValue,
+        "The 'listDatabases' command response does not contain a databases field."};
+    Status expectedResultForNoArrayOfDatabases{
+        ErrorCodes::BadValue,
+        "The 'listDatabases' command response is unable to be transformed into an array."};
+
+    const Responses responsesWithoutDatabasesField = {
+        {"listDatabases", fromjson("{ok:1, fake:[{name:'a'}, {name:'aab'}, {name:'foo'}]}")},
+        {"listDatabases", fromjson("{ok:1, fake:[{name:'admin'}, {name:'a'}, {name:'b'}]}")},
+    };
+
+    const Responses responsesWithoutArrayOfDatabases = {
+        {"listDatabases", fromjson("{ok:1, databases:1}")},
+        {"listDatabases", fromjson("{ok:1, databases:'abc'}")},
+    };
+
+    const Responses responsesWithInvalidAdminNameField = {
+        {"listDatabases", fromjson("{ok:1, databases:[{name:'a'}, {name:'aab'}, {fake:'admin'}]}")},
+        {"listDatabases", fromjson("{ok:1, databases:[{fake:'admin'}, {name:'a'}, {name:'b'}]}")},
+    };
+
+    std::unique_ptr<DatabasesCloner> cloner = makeDummyDatabasesCloner();
+
+    for (auto&& resp : responsesWithoutDatabasesField) {
+        auto parseResponseStatus = cloner->parseListDatabasesResponse_forTest(resp.second);
+        ASSERT_EQ(parseResponseStatus.getStatus(), expectedResultForNoDatabasesField);
+    }
+
+    for (auto&& resp : responsesWithoutArrayOfDatabases) {
+        auto parseResponseStatus = cloner->parseListDatabasesResponse_forTest(resp.second);
+        ASSERT_EQ(parseResponseStatus.getStatus(), expectedResultForNoArrayOfDatabases);
+    }
+
+    for (auto&& resp : responsesWithInvalidAdminNameField) {
+        auto parseResponseStatus = cloner->parseListDatabasesResponse_forTest(resp.second);
+        ASSERT_TRUE(parseResponseStatus.isOK());
+        // We expect no elements to be swapped.
+        std::vector<BSONElement> dbNamesArray = parseResponseStatus.getValue();
+        std::string expectedResult = dbNamesArray[0].Obj().firstElement().str();
+        cloner->setAdminAsFirst_forTest(dbNamesArray);
+        ASSERT_EQUALS(expectedResult, dbNamesArray[0].Obj().firstElement().str());
+    }
+}
+
 TEST_F(DBsClonerTest, FailsOnListDatabases) {
     Status result{Status::OK()};
     Status expectedResult{ErrorCodes::BadValue, "foo"};
diff --git a/src/mongo/db/repl/initial_syncer.cpp b/src/mongo/db/repl/initial_syncer.cpp
index e9a920bb37b..f3bc1a8acd3 100644
--- a/src/mongo/db/repl/initial_syncer.cpp
+++ b/src/mongo/db/repl/initial_syncer.cpp
@@ -1189,7 +1189,8 @@ Status InitialSyncer::_scheduleLastOplogEntryFetcher_inlock(Fetcher::CallbackFn
                                    query,
                                    callback,
                                    rpc::ServerSelectionMetadata(true, boost::none).toBSON(),
-                                   RemoteCommandRequest::kNoTimeout,
+                                   RemoteCommandRequest::kNoTimeout /* find network timeout */,
+                                   RemoteCommandRequest::kNoTimeout /* getMore network timeout */,
                                    RemoteCommandRetryScheduler::makeRetryPolicy(
                                        numInitialSyncOplogFindAttempts,
                                        executor::RemoteCommandRequest::kNoTimeout,
diff --git a/src/mongo/db/repl/master_slave.cpp b/src/mongo/db/repl/master_slave.cpp
index 96cfe58efdc..6929d5be297 100644
--- a/src/mongo/db/repl/master_slave.cpp
+++ b/src/mongo/db/repl/master_slave.cpp
@@ -168,7 +168,7 @@ BSONObj ReplSource::jsobj() {
 
     BSONObjBuilder dbsNextPassBuilder;
     int n = 0;
-    for (set<string>::iterator i = addDbNextPass.begin(); i != addDbNextPass.end(); i++) {
+    for (set<std::string>::iterator i = addDbNextPass.begin(); i != addDbNextPass.end(); i++) {
         n++;
         dbsNextPassBuilder.appendBool(*i, 1);
     }
@@ -177,7 +177,8 @@ BSONObj ReplSource::jsobj() {
 
     BSONObjBuilder incompleteCloneDbsBuilder;
     n = 0;
-    for (set<string>::iterator i = incompleteCloneDbs.begin(); i != incompleteCloneDbs.end(); i++) {
+    for (set<std::string>::iterator i = incompleteCloneDbs.begin(); i != incompleteCloneDbs.end();
+         i++) {
         n++;
         incompleteCloneDbsBuilder.appendBool(*i, 1);
     }
@@ -188,7 +189,7 @@ BSONObj ReplSource::jsobj() {
 }
 
 void ReplSource::ensureMe(OperationContext* txn) {
-    string myname = getHostName();
+    std::string myname = getHostName();
 
     // local.me is an identifier for a server for getLastError w:2+
     bool exists = Helpers::getSingleton(txn, "local.me", _me);
@@ -378,10 +379,10 @@ public:
     }
 
     virtual bool run(OperationContext* txn,
-                     const string& ns,
+                     const std::string& ns,
                      BSONObj& cmdObj,
                      int options,
-                     string& errmsg,
+                     std::string& errmsg,
                      BSONObjBuilder& result) {
         HandshakeArgs handshake;
         Status status = handshake.initialize(cmdObj);
@@ -398,7 +399,7 @@ public:
 } handshakeCmd;
 
 bool replHandshake(DBClientConnection* conn, const OID& myRID) {
-    string myname = getHostName();
+    std::string myname = getHostName();
 
     BSONObjBuilder cmd;
     cmd.append("handshake", myRID);
@@ -450,7 +451,7 @@ void ReplSource::forceResync(OperationContext* txn, const char* requester) {
         BSONElement e = i.next();
         if (e.eoo())
             break;
-        string name = e.embeddedObject().getField("name").valuestr();
+        std::string name = e.embeddedObject().getField("name").valuestr();
         if (!e.embeddedObject().getBoolField("empty")) {
             if (name != "local") {
                 if (only.empty() || only == name) {
@@ -481,7 +482,7 @@ Status ReplSource::_updateIfDoneWithInitialSync() {
     return Status::OK();
 }
 
-void ReplSource::resyncDrop(OperationContext* txn, const string& dbName) {
+void ReplSource::resyncDrop(OperationContext* txn, const std::string& dbName) {
     log() << "resync: dropping database " << dbName;
     invariant(txn->lockState()->isW());
 
@@ -531,13 +532,13 @@ void ReplSource::resync(OperationContext* txn, const std::string& dbName) {
 
 static DatabaseIgnorer ___databaseIgnorer;
 
-void DatabaseIgnorer::doIgnoreUntilAfter(const string& db, const Timestamp& futureOplogTime) {
+void DatabaseIgnorer::doIgnoreUntilAfter(const std::string& db, const Timestamp& futureOplogTime) {
     if (futureOplogTime > _ignores[db]) {
         _ignores[db] = futureOplogTime;
     }
 }
 
-bool DatabaseIgnorer::ignoreAt(const string& db, const Timestamp& currentOplogTime) {
+bool DatabaseIgnorer::ignoreAt(const std::string& db, const Timestamp& currentOplogTime) {
     if (_ignores[db].isNull()) {
         return false;
     }
@@ -627,7 +628,7 @@ bool ReplSource::handleDuplicateDbName(OperationContext* txn,
 
     // The database is present on the master and no conflicting databases
     // are present on the master.  Drop any local conflicts.
-    for (set<string>::const_iterator i = duplicates.begin(); i != duplicates.end(); ++i) {
+    for (set<std::string>::const_iterator i = duplicates.begin(); i != duplicates.end(); ++i) {
         ___databaseIgnorer.doIgnoreUntilAfter(*i, lastTime);
         incompleteCloneDbs.erase(*i);
         addDbNextPass.erase(*i);
@@ -822,10 +823,10 @@ void ReplSource::_sync_pullOpLog_applyOperation(OperationContext* txn,
 }
 
 void ReplSource::syncToTailOfRemoteLog() {
-    string _ns = ns();
+    std::string _ns = ns();
     BSONObjBuilder b;
     if (!only.empty()) {
-        b.appendRegex("ns", string("^") + pcrecpp::RE::QuoteMeta(only));
+        b.appendRegex("ns", std::string("^") + pcrecpp::RE::QuoteMeta(only));
     }
     BSONObj last = oplogReader.findOne(_ns.c_str(), Query(b.done()).sort(BSON("$natural" << -1)));
     if (!last.isEmpty()) {
@@ -873,7 +874,7 @@ public:
 */
 int ReplSource::_sync_pullOpLog(OperationContext* txn, int& nApplied) {
     int okResultCode = restartSyncAfterSleep;
-    string ns = string("local.oplog.$") + sourceName();
+    std::string ns = std::string("local.oplog.$") + sourceName();
     LOG(2) << "sync_pullOpLog " << ns << " syncedTo:" << syncedTo.toStringLong() << '\n';
 
     bool tailing = true;
@@ -893,7 +894,7 @@ int ReplSource::_sync_pullOpLog(OperationContext* txn, int& nApplied) {
                 BSONElement e = i.next();
                 if (e.eoo())
                     break;
-                string name = e.embeddedObject().getField("name").valuestr();
+                std::string name = e.embeddedObject().getField("name").valuestr();
                 if (!e.embeddedObject().getBoolField("empty")) {
                     if (name != "local") {
                         if (only.empty() || only == name) {
@@ -917,7 +918,7 @@ int ReplSource::_sync_pullOpLog(OperationContext* txn, int& nApplied) {
         if (!only.empty()) {
             // note we may here skip a LOT of data table scanning, a lot of work for the master.
             // maybe append "\\." here?
-            query.appendRegex("ns", string("^") + pcrecpp::RE::QuoteMeta(only));
+            query.appendRegex("ns", std::string("^") + pcrecpp::RE::QuoteMeta(only));
         }
         BSONObj queryObj = query.done();
         // e.g. queryObj = { ts: { $gte: syncedTo } }
@@ -936,7 +937,7 @@ int ReplSource::_sync_pullOpLog(OperationContext* txn, int& nApplied) {
 
     // show any deferred database creates from a previous pass
     {
-        set<string>::iterator i = addDbNextPass.begin();
+        set<std::string>::iterator i = addDbNextPass.begin();
         if (i != addDbNextPass.end()) {
             BSONObjBuilder b;
             b.append("ns", *i + '.');
@@ -980,7 +981,7 @@ int ReplSource::_sync_pullOpLog(OperationContext* txn, int& nApplied) {
         BSONObj op = oplogReader.nextSafe();
         BSONElement ts = op.getField("ts");
         if (ts.type() != Date && ts.type() != bsonTimestamp) {
-            string err = op.getStringField("$err");
+            std::string err = op.getStringField("$err");
             if (!err.empty()) {
                 // 13051 is "tailable cursor requested on non capped collection"
                 if (op.getIntField("code") == 13051) {
@@ -1148,7 +1149,7 @@ int ReplSource::sync(OperationContext* txn, int& nApplied) {
 
     // FIXME Handle cases where this db isn't on default port, or default port is spec'd in
     // hostName.
-    if ((string("localhost") == hostName || string("127.0.0.1") == hostName) &&
+    if ((std::string("localhost") == hostName || std::string("127.0.0.1") == hostName) &&
         serverGlobalParams.port == ServerGlobalParams::DefaultDBPort) {
         log() << "can't sync from self (localhost). sources configuration may be wrong." << endl;
         sleepsecs(5);
@@ -1293,7 +1294,7 @@ static void replMain(OperationContext* txn) {
         if (s) {
             stringstream ss;
             ss << "sleep " << s << " sec before next pass";
-            string msg = ss.str();
+            std::string msg = ss.str();
             if (!serverGlobalParams.quiet)
                 log() << msg << endl;
             ReplInfo r(msg.c_str());
diff --git a/src/mongo/db/repl/oplog.cpp b/src/mongo/db/repl/oplog.cpp
index 43a1eba5042..76df271ed78 100644
--- a/src/mongo/db/repl/oplog.cpp
+++ b/src/mongo/db/repl/oplog.cpp
@@ -116,6 +116,11 @@ namespace {
 // cached copy...so don't rename, drop, etc.!!!
 Collection* _localOplogCollection = nullptr;
 
+// Specifies whether we abort initial sync when attempting to apply a renameCollection operation.
+// If set to true, users risk corrupting their data. This should only be enabled by expert users
+// of the server who understand the risks this poses.
+MONGO_EXPORT_SERVER_PARAMETER(allowUnsafeRenamesDuringInitialSync, bool, false);
+
 PseudoRandom hashGenerator(std::unique_ptr<SecureRandom>(SecureRandom::create())->nextInt64());
 
 // Synchronizes the section where a new Timestamp is generated and when it actually
@@ -663,6 +668,45 @@ std::map<std::string, ApplyOpMetadata> opsMap = {
 
 }  // namespace
 
+std::pair<BSONObj, NamespaceString> prepForApplyOpsIndexInsert(const BSONElement& fieldO,
+                                                               const BSONObj& op,
+                                                               const NamespaceString& requestNss) {
+    uassert(ErrorCodes::NoSuchKey,
+            str::stream() << "Missing expected index spec in field 'o': " << op,
+            !fieldO.eoo());
+    uassert(ErrorCodes::TypeMismatch,
+            str::stream() << "Expected object for index spec in field 'o': " << op,
+            fieldO.isABSONObj());
+    BSONObj indexSpec = fieldO.embeddedObject();
+
+    std::string indexNs;
+    uassertStatusOK(bsonExtractStringField(indexSpec, "ns", &indexNs));
+    const NamespaceString indexNss(indexNs);
+    uassert(ErrorCodes::InvalidNamespace,
+            str::stream() << "Invalid namespace in index spec: " << op,
+            indexNss.isValid());
+    uassert(ErrorCodes::InvalidNamespace,
+            str::stream() << "Database name mismatch for database (" << requestNss.db()
+                          << ") while creating index: "
+                          << op,
+            requestNss.db() == indexNss.db());
+
+    if (!indexSpec["v"]) {
+        // If the "v" field isn't present in the index specification, then we assume it is a
+        // v=1 index from an older version of MongoDB. This is because
+        //   (1) we haven't built v=0 indexes as the default for a long time, and
+        //   (2) the index version has been included in the corresponding oplog entry since
+        //       v=2 indexes were introduced.
+        BSONObjBuilder bob;
+
+        bob.append("v", static_cast<int>(IndexVersion::kV1));
+        bob.appendElements(indexSpec);
+
+        indexSpec = bob.obj();
+    }
+
+    return std::make_pair(indexSpec, indexNss);
+}
 // @return failure status if an update should have happened and the document DNE.
 // See replset initial sync code.
 Status applyOperation_inlock(OperationContext* txn,
@@ -688,6 +732,7 @@ Status applyOperation_inlock(OperationContext* txn,
         o = fieldO.embeddedObject();
 
     const StringData ns = fieldNs.valueStringData();
+    NamespaceString requestNss{ns};
 
     BSONObj o2;
     if (fieldO2.isABSONObj())
@@ -718,27 +763,11 @@ Status applyOperation_inlock(OperationContext* txn,
     invariant(*opType != 'c');  // commands are processed in applyCommand_inlock()
 
     if (*opType == 'i') {
-        if (nsToCollectionSubstring(ns) == "system.indexes") {
-            uassert(ErrorCodes::NoSuchKey,
-                    str::stream() << "Missing expected index spec in field 'o': " << op,
-                    !fieldO.eoo());
-            uassert(ErrorCodes::TypeMismatch,
-                    str::stream() << "Expected object for index spec in field 'o': " << op,
-                    fieldO.isABSONObj());
-            BSONObj indexSpec = fieldO.embeddedObject();
-
-            std::string indexNs;
-            uassertStatusOK(bsonExtractStringField(indexSpec, "ns", &indexNs));
-            const NamespaceString indexNss(indexNs);
-            uassert(ErrorCodes::InvalidNamespace,
-                    str::stream() << "Invalid namespace in index spec: " << op,
-                    indexNss.isValid());
-            uassert(ErrorCodes::InvalidNamespace,
-                    str::stream() << "Database name mismatch for database ("
-                                  << nsToDatabaseSubstring(ns)
-                                  << ") while creating index: "
-                                  << op,
-                    nsToDatabaseSubstring(ns) == indexNss.db());
+        if (requestNss.isSystemDotIndexes()) {
+            BSONObj indexSpec;
+            NamespaceString indexNss;
+            std::tie(indexSpec, indexNss) =
+                repl::prepForApplyOpsIndexInsert(fieldO, op, requestNss);
 
             // Check if collection exists.
             auto indexCollection = db->getCollection(indexNss);
@@ -749,20 +778,6 @@ Status applyOperation_inlock(OperationContext* txn,
 
             opCounters->gotInsert();
 
-            if (!indexSpec["v"]) {
-                // If the "v" field isn't present in the index specification, then we assume it is a
-                // v=1 index from an older version of MongoDB. This is because
-                //   (1) we haven't built v=0 indexes as the default for a long time, and
-                //   (2) the index version has been included in the corresponding oplog entry since
-                //       v=2 indexes were introduced.
-                BSONObjBuilder bob;
-
-                bob.append("v", static_cast<int>(IndexVersion::kV1));
-                bob.appendElements(indexSpec);
-
-                indexSpec = bob.obj();
-            }
-
             bool relaxIndexConstraints =
                 ReplicationCoordinator::get(txn)->shouldRelaxIndexConstraints(indexNss);
             if (indexSpec["background"].trueValue()) {
@@ -863,13 +878,12 @@ Status applyOperation_inlock(OperationContext* txn,
                 BSONObjBuilder b;
                 b.append(o.getField("_id"));
 
-                const NamespaceString requestNs(ns);
-                UpdateRequest request(requestNs);
+                UpdateRequest request(requestNss);
 
                 request.setQuery(b.done());
                 request.setUpdates(o);
                 request.setUpsert();
-                UpdateLifecycleImpl updateLifecycle(requestNs);
+                UpdateLifecycleImpl updateLifecycle(requestNss);
                 request.setLifecycle(&updateLifecycle);
 
                 UpdateResult res = update(txn, db, request);
@@ -894,13 +908,12 @@ Status applyOperation_inlock(OperationContext* txn,
                 str::stream() << "Failed to apply update due to missing _id: " << op.toString(),
                 updateCriteria.hasField("_id"));
 
-        const NamespaceString requestNs(ns);
-        UpdateRequest request(requestNs);
+        UpdateRequest request(requestNss);
 
         request.setQuery(updateCriteria);
         request.setUpdates(o);
         request.setUpsert(upsert);
-        UpdateLifecycleImpl updateLifecycle(requestNs);
+        UpdateLifecycleImpl updateLifecycle(requestNss);
         request.setLifecycle(&updateLifecycle);
 
         UpdateResult ur = update(txn, db, request);
@@ -954,7 +967,12 @@ Status applyOperation_inlock(OperationContext* txn,
                 o.hasField("_id"));
 
         if (opType[1] == 0) {
-            deleteObjects(txn, collection, ns, o, PlanExecutor::YIELD_MANUAL, /*justOne*/ valueB);
+            deleteObjects(txn,
+                          collection,
+                          requestNss.ns().c_str(),
+                          o,
+                          PlanExecutor::YIELD_MANUAL,
+                          /*justOne*/ valueB);
         } else
             verify(opType[1] == 'b');  // "db" advertisement
         if (incrementOpsAppliedStats) {
@@ -970,16 +988,6 @@ Status applyOperation_inlock(OperationContext* txn,
             14825, str::stream() << "error in applyOperation : unknown opType " << *opType);
     }
 
-    // AuthorizationManager's logOp method registers a RecoveryUnit::Change and to do so we need
-    // to a new WriteUnitOfWork, if we dont have a wrapping unit of work already. If we already
-    // have a wrapping WUOW, the extra nexting is harmless. The logOp really should have been
-    // done in the WUOW that did the write, but this won't happen because applyOps turns off
-    // observers.
-    WriteUnitOfWork wuow(txn);
-    getGlobalAuthorizationManager()->logOp(
-        txn, opType, ns.toString().c_str(), o, fieldO2.isABSONObj() ? &o2 : NULL);
-    wuow.commit();
-
     return Status::OK();
 }
 
@@ -1021,9 +1029,15 @@ Status applyCommand_inlock(OperationContext* txn,
     // Applying renameCollection during initial sync might lead to data corruption, so we restart
     // the initial sync.
     if (!inSteadyStateReplication && o.firstElementFieldName() == std::string("renameCollection")) {
-        return Status(ErrorCodes::OplogOperationUnsupported,
-                      str::stream() << "Applying renameCollection not supported in initial sync: "
-                                    << redact(op));
+        if (!allowUnsafeRenamesDuringInitialSync.load()) {
+            return Status(ErrorCodes::OplogOperationUnsupported,
+                          str::stream()
+                              << "Applying renameCollection not supported in initial sync: "
+                              << redact(op));
+        }
+        warning() << "allowUnsafeRenamesDuringInitialSync set to true. Applying renameCollection "
+                     "operation during initial sync even though it may lead to data corruption: "
+                  << redact(op);
     }
 
     // Applying commands in repl is done under Global W-lock, so it is safe to not
@@ -1071,11 +1085,8 @@ Status applyCommand_inlock(OperationContext* txn,
                 break;
             }
             default:
-                if (_oplogCollectionName == masterSlaveOplogName) {
-                    error() << "Failed command " << redact(o) << " on " << nss.db()
-                            << " with status " << status << " during oplog application";
-                } else if (curOpToApply.acceptableErrors.find(status.code()) ==
-                           curOpToApply.acceptableErrors.end()) {
+                if (curOpToApply.acceptableErrors.find(status.code()) ==
+                    curOpToApply.acceptableErrors.end()) {
                     error() << "Failed command " << redact(o) << " on " << nss.db()
                             << " with status " << status << " during oplog application";
                     return status;
diff --git a/src/mongo/db/repl/oplog.h b/src/mongo/db/repl/oplog.h
index b2078e93d28..f51aadb8dcf 100644
--- a/src/mongo/db/repl/oplog.h
+++ b/src/mongo/db/repl/oplog.h
@@ -104,6 +104,15 @@ void oplogCheckCloseDatabase(OperationContext* txn, Database* db);
 
 using IncrementOpsAppliedStatsFn = stdx::function<void()>;
 /**
+ * Take the object field of a BSONObj, the BSONObj, and the namespace of
+ * the operation and perform necessary validation to ensure the BSONObj is a
+ * properly-formed command to insert into system.indexes. This is only to
+ * be used for insert operations into system.indexes. It is called via applyOps.
+ */
+std::pair<BSONObj, NamespaceString> prepForApplyOpsIndexInsert(const BSONElement& fieldO,
+                                                               const BSONObj& op,
+                                                               const NamespaceString& requestNss);
+/**
  * Take a non-command op and apply it locally
  * Used for applying from an oplog
  * @param inSteadyStateReplication convert some updates to upserts for idempotency reasons
diff --git a/src/mongo/db/repl/oplog_fetcher.cpp b/src/mongo/db/repl/oplog_fetcher.cpp
index ece0ef266fd..c9b96ab77f7 100644
--- a/src/mongo/db/repl/oplog_fetcher.cpp
+++ b/src/mongo/db/repl/oplog_fetcher.cpp
@@ -36,6 +36,7 @@
 #include "mongo/db/commands/server_status_metric.h"
 #include "mongo/db/jsobj.h"
 #include "mongo/db/repl/replication_coordinator.h"
+#include "mongo/db/server_parameters.h"
 #include "mongo/db/stats/timer_stats.h"
 #include "mongo/rpc/metadata/oplog_query_metadata.h"
 #include "mongo/rpc/metadata/server_selection_metadata.h"
@@ -56,8 +57,12 @@ MONGO_FP_DECLARE(stopReplProducer);
 
 namespace {
 
-Seconds kOplogInitialFindMaxTime{60};
-Seconds kOplogQueryNetworkTimeout{65};  // 5 seconds past the find command's 1 minute maxTimeMs
+// Number of seconds for the `maxTimeMS` on the initial `find` command.
+MONGO_EXPORT_SERVER_PARAMETER(oplogInitialFindMaxSeconds, int, 60);
+
+// Number of milliseconds to add to the `find` and `getMore` timeouts to calculate the network
+// timeout for the requests.
+const Milliseconds kNetworkTimeoutBufferMS{5000};
 
 Counter64 readersCreatedStats;
 ServerStatusMetricField<Counter64> displayReadersCreated("repl.network.readersCreated",
@@ -91,17 +96,22 @@ Milliseconds calculateAwaitDataTimeout(const ReplSetConfig& config) {
  */
 BSONObj makeFindCommandObject(const NamespaceString& nss,
                               long long currentTerm,
-                              OpTime lastOpTimeFetched) {
+                              OpTime lastOpTimeFetched,
+                              Milliseconds fetcherMaxTimeMS) {
     BSONObjBuilder cmdBob;
     cmdBob.append("find", nss.coll());
     cmdBob.append("filter", BSON("ts" << BSON("$gte" << lastOpTimeFetched.getTimestamp())));
     cmdBob.append("tailable", true);
     cmdBob.append("oplogReplay", true);
     cmdBob.append("awaitData", true);
-    cmdBob.append("maxTimeMS", durationCount<Milliseconds>(kOplogInitialFindMaxTime));
+    cmdBob.append("maxTimeMS", durationCount<Milliseconds>(fetcherMaxTimeMS));
     if (currentTerm != OpTime::kUninitializedTerm) {
         cmdBob.append("term", currentTerm);
     }
+    if (serverGlobalParams.featureCompatibility.version.load() ==
+        ServerGlobalParams::FeatureCompatibility::Version::k34) {
+        cmdBob.append("readConcern", BSON("afterOpTime" << lastOpTimeFetched.toBSON()));
+    }
     return cmdBob.obj();
 }
 
@@ -450,6 +460,14 @@ BSONObj OplogFetcher::getMetadataObject_forTest() const {
 }
 
 Milliseconds OplogFetcher::getAwaitDataTimeout_forTest() const {
+    return _getGetMoreMaxTime();
+}
+
+Milliseconds OplogFetcher::_getFindMaxTime() const {
+    return Milliseconds(oplogInitialFindMaxSeconds.load() * 1000);
+}
+
+Milliseconds OplogFetcher::_getGetMoreMaxTime() const {
     return _awaitDataTimeout;
 }
 
@@ -660,7 +678,7 @@ void OplogFetcher::_callback(const Fetcher::QueryResponseStatus& result,
     getMoreBob->appendElements(makeGetMoreCommandObject(queryResponse.nss,
                                                         queryResponse.cursorId,
                                                         lastCommittedWithCurrentTerm,
-                                                        _awaitDataTimeout));
+                                                        _getGetMoreMaxTime()));
 }
 
 void OplogFetcher::_finishCallback(Status status) {
@@ -691,10 +709,11 @@ std::unique_ptr<Fetcher> OplogFetcher::_makeFetcher(long long currentTerm,
         _executor,
         _source,
         _nss.db().toString(),
-        makeFindCommandObject(_nss, currentTerm, lastFetchedOpTime),
+        makeFindCommandObject(_nss, currentTerm, lastFetchedOpTime, _getFindMaxTime()),
         stdx::bind(&OplogFetcher::_callback, this, stdx::placeholders::_1, stdx::placeholders::_3),
         _metadataObject,
-        kOplogQueryNetworkTimeout);
+        _getFindMaxTime() + kNetworkTimeoutBufferMS,
+        _getGetMoreMaxTime() + kNetworkTimeoutBufferMS);
 }
 
 bool OplogFetcher::_isShuttingDown() const {
diff --git a/src/mongo/db/repl/oplog_fetcher.h b/src/mongo/db/repl/oplog_fetcher.h
index ad4fde296b1..54bfbabbf8d 100644
--- a/src/mongo/db/repl/oplog_fetcher.h
+++ b/src/mongo/db/repl/oplog_fetcher.h
@@ -235,6 +235,16 @@ private:
     void _finishCallback(Status status, OpTimeWithHash opTimeWithHash);
 
     /**
+     * Returns how long the `find` command should wait before timing out.
+     */
+    virtual Milliseconds _getFindMaxTime() const;
+
+    /**
+     * Returns how long the `getMore` command should wait before timing out.
+     */
+    virtual Milliseconds _getGetMoreMaxTime() const;
+
+    /**
      * Creates a new instance of the fetcher to tail the remote oplog starting at the given optime.
      */
     std::unique_ptr<Fetcher> _makeFetcher(long long currentTerm, OpTime lastFetchedOpTime);
diff --git a/src/mongo/db/repl/repl_client_info.cpp b/src/mongo/db/repl/repl_client_info.cpp
index 3e98a0cb9d3..5b77e3aa6f3 100644
--- a/src/mongo/db/repl/repl_client_info.cpp
+++ b/src/mongo/db/repl/repl_client_info.cpp
@@ -26,6 +26,8 @@
  *    it in the license file.
  */
 
+#define MONGO_LOG_DEFAULT_COMPONENT ::mongo::logger::LogComponent::kReplication
+
 #include "mongo/platform/basic.h"
 
 #include "mongo/db/repl/repl_client_info.h"
@@ -36,6 +38,7 @@
 #include "mongo/db/operation_context.h"
 #include "mongo/db/repl/replication_coordinator_global.h"
 #include "mongo/util/decorable.h"
+#include "mongo/util/log.h"
 
 namespace mongo {
 namespace repl {
@@ -48,10 +51,23 @@ void ReplClientInfo::setLastOp(const OpTime& ot) {
     _lastOp = ot;
 }
 
+
 void ReplClientInfo::setLastOpToSystemLastOpTime(OperationContext* txn) {
     ReplicationCoordinator* replCoord = repl::ReplicationCoordinator::get(txn->getServiceContext());
     if (replCoord->isReplEnabled() && txn->writesAreReplicated()) {
-        setLastOp(replCoord->getMyLastAppliedOpTime());
+        auto systemOpTime = replCoord->getMyLastAppliedOpTime();
+
+        // If the system optime has gone backwards, that must mean that there was a rollback.
+        // This is safe, but the last op for a Client should never go backwards, so just leave
+        // the last op for this Client as it was.
+        if (systemOpTime >= _lastOp) {
+            _lastOp = systemOpTime;
+        } else {
+            log() << "Not setting the last OpTime for this Client from " << _lastOp
+                  << " to the current system time of " << systemOpTime
+                  << " as that would be moving the OpTime backwards.  This should only happen if "
+                     "there was a rollback recently";
+        }
     }
 }
 
diff --git a/src/mongo/db/repl/repl_set_config.cpp b/src/mongo/db/repl/repl_set_config.cpp
index 1ad2d245ac6..024c3ff04a5 100644
--- a/src/mongo/db/repl/repl_set_config.cpp
+++ b/src/mongo/db/repl/repl_set_config.cpp
@@ -35,11 +35,19 @@
 #include "mongo/bson/util/bson_check.h"
 #include "mongo/bson/util/bson_extract.h"
 #include "mongo/db/jsobj.h"
+#include "mongo/db/mongod_options.h"
 #include "mongo/db/server_options.h"
+#include "mongo/db/server_parameters.h"
 #include "mongo/stdx/functional.h"
 #include "mongo/util/stringutils.h"
 
 namespace mongo {
+/**
+ * Dont run any sharding validations. Can not be combined with --configsvr or shardvr. Intended to
+ * allow restarting config server or shard as an independent replica set.
+ */
+MONGO_EXPORT_STARTUP_SERVER_PARAMETER(skipShardingConfigurationChecks, bool, false);
+
 namespace repl {
 
 const size_t ReplSetConfig::kMaxMembers;
@@ -552,7 +560,8 @@ Status ReplSetConfig::validate() const {
                               "servers cannot have a non-zero slaveDelay");
             }
         }
-        if (serverGlobalParams.clusterRole != ClusterRole::ConfigServer) {
+        if (serverGlobalParams.clusterRole != ClusterRole::ConfigServer &&
+            !skipShardingConfigurationChecks) {
             return Status(ErrorCodes::BadValue,
                           "Nodes being used for config servers must be started with the "
                           "--configsvr flag");
diff --git a/src/mongo/db/repl/replication_coordinator_external_state_impl.cpp b/src/mongo/db/repl/replication_coordinator_external_state_impl.cpp
index cfc6dbb55e1..05d070c08fe 100644
--- a/src/mongo/db/repl/replication_coordinator_external_state_impl.cpp
+++ b/src/mongo/db/repl/replication_coordinator_external_state_impl.cpp
@@ -240,8 +240,11 @@ void ReplicationCoordinatorExternalStateImpl::startSteadyStateReplication(
     _applierThread->startup();
     log() << "Starting replication reporter thread";
     invariant(!_syncSourceFeedbackThread);
-    _syncSourceFeedbackThread.reset(new stdx::thread(stdx::bind(
-        &SyncSourceFeedback::run, &_syncSourceFeedback, _taskExecutor.get(), _bgSync.get())));
+    _syncSourceFeedbackThread.reset(new stdx::thread(stdx::bind(&SyncSourceFeedback::run,
+                                                                &_syncSourceFeedback,
+                                                                _taskExecutor.get(),
+                                                                _bgSync.get(),
+                                                                replCoord)));
 }
 
 void ReplicationCoordinatorExternalStateImpl::stopDataReplication(OperationContext* txn) {
diff --git a/src/mongo/db/repl/replication_coordinator_impl.cpp b/src/mongo/db/repl/replication_coordinator_impl.cpp
index 3de3d35bb61..ad1a9fd93ba 100644
--- a/src/mongo/db/repl/replication_coordinator_impl.cpp
+++ b/src/mongo/db/repl/replication_coordinator_impl.cpp
@@ -2922,6 +2922,26 @@ ReplicationCoordinatorImpl::_setCurrentRSConfig_inlock(const ReplSetConfig& newC
     const ReplSetConfig oldConfig = _rsConfig;
     _rsConfig = newConfig;
     _protVersion.store(_rsConfig.getProtocolVersion());
+
+    // Warn if running --nojournal and writeConcernMajorityJournalDefault = false
+    StorageEngine* storageEngine = getGlobalServiceContext()->getGlobalStorageEngine();
+    if (storageEngine && !storageEngine->isDurable() &&
+        (newConfig.getWriteConcernMajorityShouldJournal() &&
+         (!oldConfig.isInitialized() || !oldConfig.getWriteConcernMajorityShouldJournal()))) {
+        log() << startupWarningsLog;
+        log() << "** WARNING: This replica set is running without journaling enabled but the "
+              << startupWarningsLog;
+        log() << "**          writeConcernMajorityJournalDefault option to the replica set config "
+              << startupWarningsLog;
+        log() << "**          is set to true. The writeConcernMajorityJournalDefault "
+              << startupWarningsLog;
+        log() << "**          option to the replica set config must be set to false "
+              << startupWarningsLog;
+        log() << "**          or w:majority write concerns will never complete."
+              << startupWarningsLog;
+        log() << startupWarningsLog;
+    }
+
     log() << "New replica set config in use: " << _rsConfig.toBSON() << rsLog;
     _selfIndex = myIndex;
     if (_selfIndex >= 0) {
diff --git a/src/mongo/db/repl/replication_coordinator_impl_elect_v1.cpp b/src/mongo/db/repl/replication_coordinator_impl_elect_v1.cpp
index fbd976ec6b7..1dd9bf5b7c3 100644
--- a/src/mongo/db/repl/replication_coordinator_impl_elect_v1.cpp
+++ b/src/mongo/db/repl/replication_coordinator_impl_elect_v1.cpp
@@ -133,7 +133,9 @@ void ReplicationCoordinatorImpl::_startElectSelfV1() {
         return;
     }
 
-    log() << "conducting a dry run election to see if we could be elected";
+    long long term = _topCoord->getTerm();
+
+    log() << "conducting a dry run election to see if we could be elected. current term: " << term;
     _voteRequester.reset(new VoteRequester);
 
     // This is necessary because the voteRequester may call directly into winning an
@@ -141,12 +143,11 @@ void ReplicationCoordinatorImpl::_startElectSelfV1() {
     // _mutex again.
     lk.unlock();
 
-    long long term = _topCoord->getTerm();
     StatusWith<ReplicationExecutor::EventHandle> nextPhaseEvh =
         _voteRequester->start(&_replExecutor,
                               _rsConfig,
                               _selfIndex,
-                              _topCoord->getTerm(),
+                              term,
                               true,  // dry run
                               lastOpTime);
     if (nextPhaseEvh.getStatus() == ErrorCodes::ShutdownInProgress) {
@@ -165,7 +166,8 @@ void ReplicationCoordinatorImpl::_onDryRunComplete(long long originalTerm) {
     LockGuard lk(_topoMutex);
 
     if (_topCoord->getTerm() != originalTerm) {
-        log() << "not running for primary, we have been superceded already";
+        log() << "not running for primary, we have been superseded already during dry run. "
+              << "original term: " << originalTerm << ", current term: " << _topCoord->getTerm();
         return;
     }
 
@@ -175,23 +177,24 @@ void ReplicationCoordinatorImpl::_onDryRunComplete(long long originalTerm) {
         log() << "not running for primary, we received insufficient votes";
         return;
     } else if (endResult == VoteRequester::Result::kStaleTerm) {
-        log() << "not running for primary, we have been superceded already";
+        log() << "not running for primary, we have been superseded already";
         return;
     } else if (endResult != VoteRequester::Result::kSuccessfullyElected) {
         log() << "not running for primary, we received an unexpected problem";
         return;
     }
 
-    log() << "dry election run succeeded, running for election";
+    long long newTerm = originalTerm + 1;
+    log() << "dry election run succeeded, running for election in term " << newTerm;
     // Stepdown is impossible from this term update.
     TopologyCoordinator::UpdateTermResult updateTermResult;
-    _updateTerm_incallback(originalTerm + 1, &updateTermResult);
+    _updateTerm_incallback(newTerm, &updateTermResult);
     invariant(updateTermResult == TopologyCoordinator::UpdateTermResult::kUpdatedTerm);
     // Secure our vote for ourself first
     _topCoord->voteForMyselfV1();
 
     // Store the vote in persistent storage.
-    LastVote lastVote{originalTerm + 1, _selfIndex};
+    LastVote lastVote{newTerm, _selfIndex};
 
     auto cbStatus = _replExecutor.scheduleDBWork(
         [this, lastVote](const ReplicationExecutor::CallbackArgs& cbData) {
@@ -232,12 +235,19 @@ void ReplicationCoordinatorImpl::_startVoteRequester(long long newTerm) {
 
     LockGuard lk(_topoMutex);
 
+    if (_topCoord->getTerm() != newTerm) {
+        log() << "not running for primary, we have been superseded already while writing our last "
+                 "vote. election term: "
+              << newTerm << ", current term: " << _topCoord->getTerm();
+        return;
+    }
+
     const auto lastOpTime =
         _isDurableStorageEngine() ? getMyLastDurableOpTime() : getMyLastAppliedOpTime();
 
     _voteRequester.reset(new VoteRequester);
-    StatusWith<ReplicationExecutor::EventHandle> nextPhaseEvh = _voteRequester->start(
-        &_replExecutor, _rsConfig, _selfIndex, _topCoord->getTerm(), false, lastOpTime);
+    StatusWith<ReplicationExecutor::EventHandle> nextPhaseEvh =
+        _voteRequester->start(&_replExecutor, _rsConfig, _selfIndex, newTerm, false, lastOpTime);
     if (nextPhaseEvh.getStatus() == ErrorCodes::ShutdownInProgress) {
         return;
     }
@@ -249,14 +259,15 @@ void ReplicationCoordinatorImpl::_startVoteRequester(long long newTerm) {
     lossGuard.dismiss();
 }
 
-void ReplicationCoordinatorImpl::_onVoteRequestComplete(long long originalTerm) {
+void ReplicationCoordinatorImpl::_onVoteRequestComplete(long long newTerm) {
     invariant(_voteRequester);
     LoseElectionGuardV1 lossGuard(this);
 
     LockGuard lk(_topoMutex);
 
-    if (_topCoord->getTerm() != originalTerm) {
-        log() << "not becoming primary, we have been superceded already";
+    if (_topCoord->getTerm() != newTerm) {
+        log() << "not becoming primary, we have been superseded already during election. "
+              << "election term: " << newTerm << ", current term: " << _topCoord->getTerm();
         return;
     }
 
@@ -267,7 +278,7 @@ void ReplicationCoordinatorImpl::_onVoteRequestComplete(long long originalTerm)
             log() << "not becoming primary, we received insufficient votes";
             return;
         case VoteRequester::Result::kStaleTerm:
-            log() << "not becoming primary, we have been superceded already";
+            log() << "not becoming primary, we have been superseded already";
             return;
         case VoteRequester::Result::kSuccessfullyElected:
             log() << "election succeeded, assuming primary role in term " << _topCoord->getTerm();
diff --git a/src/mongo/db/repl/replication_coordinator_impl_elect_v1_test.cpp b/src/mongo/db/repl/replication_coordinator_impl_elect_v1_test.cpp
index 6e87f882b48..0f550869546 100644
--- a/src/mongo/db/repl/replication_coordinator_impl_elect_v1_test.cpp
+++ b/src/mongo/db/repl/replication_coordinator_impl_elect_v1_test.cpp
@@ -431,7 +431,7 @@ TEST_F(ReplCoordTest, ElectionFailsWhenDryRunResponseContainsANewerTerm) {
     getReplCoord()->waitForElectionFinish_forTest();
     stopCapturingLogMessages();
     ASSERT_EQUALS(
-        1, countLogLinesContaining("not running for primary, we have been superceded already"));
+        1, countLogLinesContaining("not running for primary, we have been superseded already"));
 }
 
 TEST_F(ReplCoordTest, NodeWillNotStandForElectionDuringHeartbeatReconfig) {
@@ -689,7 +689,7 @@ TEST_F(ReplCoordTest, ElectionFailsWhenVoteRequestResponseContainsANewerTerm) {
     getReplCoord()->waitForElectionFinish_forTest();
     stopCapturingLogMessages();
     ASSERT_EQUALS(1,
-                  countLogLinesContaining("not becoming primary, we have been superceded already"));
+                  countLogLinesContaining("not becoming primary, we have been superseded already"));
 }
 
 TEST_F(ReplCoordTest, ElectionFailsWhenTermChangesDuringDryRun) {
@@ -728,8 +728,9 @@ TEST_F(ReplCoordTest, ElectionFailsWhenTermChangesDuringDryRun) {
     simulateSuccessfulDryRun(onDryRunRequest);
 
     stopCapturingLogMessages();
-    ASSERT_EQUALS(
-        1, countLogLinesContaining("not running for primary, we have been superceded already"));
+    ASSERT_EQUALS(1,
+                  countLogLinesContaining(
+                      "not running for primary, we have been superseded already during dry run"));
 }
 
 TEST_F(ReplCoordTest, ElectionFailsWhenTermChangesDuringActualElection) {
@@ -784,7 +785,7 @@ TEST_F(ReplCoordTest, ElectionFailsWhenTermChangesDuringActualElection) {
     getReplCoord()->waitForElectionFinish_forTest();
     stopCapturingLogMessages();
     ASSERT_EQUALS(1,
-                  countLogLinesContaining("not becoming primary, we have been superceded already"));
+                  countLogLinesContaining("not becoming primary, we have been superseded already"));
 }
 
 class PriorityTakeoverTest : public ReplCoordTest {
diff --git a/src/mongo/db/repl/replication_coordinator_impl_heartbeat.cpp b/src/mongo/db/repl/replication_coordinator_impl_heartbeat.cpp
index b72d7e7dbcc..66c9799029d 100644
--- a/src/mongo/db/repl/replication_coordinator_impl_heartbeat.cpp
+++ b/src/mongo/db/repl/replication_coordinator_impl_heartbeat.cpp
@@ -778,18 +778,24 @@ void ReplicationCoordinatorImpl::_scheduleNextLivenessUpdate_inlock() {
     }
 
     auto nextTimeout = earliestDate + _rsConfig.getElectionTimeoutPeriod();
-    if (nextTimeout > _replExecutor.now()) {
-        LOG(3) << "scheduling next check at " << nextTimeout;
-        auto cbh = _scheduleWorkAt(nextTimeout,
-                                   stdx::bind(&ReplicationCoordinatorImpl::_handleLivenessTimeout,
-                                              this,
-                                              stdx::placeholders::_1));
-        if (!cbh) {
-            return;
-        }
-        _handleLivenessTimeoutCbh = cbh;
-        _earliestMemberId = earliestMemberId;
+    LOG(3) << "scheduling next check at " << nextTimeout;
+
+    // It is possible we will schedule the next timeout in the past.
+    // ReplicationExecutor::_scheduleWorkAt() schedules its work immediately if it's given a
+    // time <= now().
+    // If we missed the timeout, it means that on our last check the earliest live member was
+    // just barely fresh and it has become stale since then. We must schedule another liveness
+    // check to continue conducting liveness checks and be able to step down from primary if we
+    // lose contact with a majority of nodes.
+    auto cbh = _scheduleWorkAt(nextTimeout,
+                               stdx::bind(&ReplicationCoordinatorImpl::_handleLivenessTimeout,
+                                          this,
+                                          stdx::placeholders::_1));
+    if (!cbh) {
+        return;
     }
+    _handleLivenessTimeoutCbh = cbh;
+    _earliestMemberId = earliestMemberId;
 }
 
 void ReplicationCoordinatorImpl::_cancelAndRescheduleLivenessUpdate_inlock(int updatedMemberId) {
diff --git a/src/mongo/db/repl/replication_executor_test.cpp b/src/mongo/db/repl/replication_executor_test.cpp
index ba41df46769..2e8a66465de 100644
--- a/src/mongo/db/repl/replication_executor_test.cpp
+++ b/src/mongo/db/repl/replication_executor_test.cpp
@@ -249,6 +249,21 @@ TEST_F(ReplicationExecutorTest, ScheduleCallbackAtNow) {
     executor.waitForEvent(finishEvent);
 }
 
+TEST_F(ReplicationExecutorTest, ScheduleCallbackInPast) {
+    launchExecutorThread();
+    getNet()->exitNetwork();
+
+    ReplicationExecutor& executor = getReplExecutor();
+    auto finishEvent = assertGet(executor.makeEvent());
+    auto fn = [&executor, finishEvent](const ReplicationExecutor::CallbackArgs& cbData) {
+        ASSERT_OK(cbData.status);
+        executor.signalEvent(finishEvent);
+    };
+
+    auto cb = executor.scheduleWorkAt(getNet()->now() - Milliseconds(1000), fn);
+    executor.waitForEvent(finishEvent);
+}
+
 TEST_F(ReplicationExecutorTest, ScheduleCallbackAtAFutureTime) {
     launchExecutorThread();
     getNet()->exitNetwork();
diff --git a/src/mongo/db/repl/sync_source_feedback.cpp b/src/mongo/db/repl/sync_source_feedback.cpp
index 4d395e376e6..a82b1553212 100644
--- a/src/mongo/db/repl/sync_source_feedback.cpp
+++ b/src/mongo/db/repl/sync_source_feedback.cpp
@@ -50,27 +50,23 @@ namespace repl {
 namespace {
 
 /**
- * Calculates the keep alive interval based on the current configuration in the replication
- * coordinator.
+ * Calculates the keep alive interval based on the given ReplSetConfig.
  */
-Milliseconds calculateKeepAliveInterval(OperationContext* txn, stdx::mutex& mtx) {
-    stdx::lock_guard<stdx::mutex> lock(mtx);
-    auto replCoord = repl::ReplicationCoordinator::get(txn);
-    auto rsConfig = replCoord->getConfig();
-    auto keepAliveInterval = rsConfig.getElectionTimeoutPeriod() / 2;
-    return keepAliveInterval;
+Milliseconds calculateKeepAliveInterval(const ReplSetConfig& rsConfig) {
+    return rsConfig.getElectionTimeoutPeriod() / 2;
 }
 
 /**
  * Returns function to prepare update command
  */
 Reporter::PrepareReplSetUpdatePositionCommandFn makePrepareReplSetUpdatePositionCommandFn(
-    OperationContext* txn,
+    ReplicationCoordinator* replCoord,
     stdx::mutex& mtx,
     const HostAndPort& syncTarget,
     BackgroundSync* bgsync) {
-    return [&mtx, syncTarget, txn, bgsync](ReplicationCoordinator::ReplSetUpdatePositionCommandStyle
-                                               commandStyle) -> StatusWith<BSONObj> {
+    return [&mtx, syncTarget, replCoord, bgsync](
+               ReplicationCoordinator::ReplSetUpdatePositionCommandStyle
+                   commandStyle) -> StatusWith<BSONObj> {
         auto currentSyncTarget = bgsync->getSyncTarget();
         if (currentSyncTarget != syncTarget) {
             if (currentSyncTarget.empty()) {
@@ -86,7 +82,6 @@ Reporter::PrepareReplSetUpdatePositionCommandFn makePrepareReplSetUpdatePosition
             }
         }
 
-        auto replCoord = repl::ReplicationCoordinator::get(txn);
         if (replCoord->getMemberState().primary()) {
             // Primary has no one to send updates to.
             return Status(ErrorCodes::InvalidSyncSource,
@@ -114,7 +109,7 @@ void SyncSourceFeedback::forwardSlaveProgress() {
     }
 }
 
-Status SyncSourceFeedback::_updateUpstream(OperationContext* txn,
+Status SyncSourceFeedback::_updateUpstream(ReplicationCoordinator* replCoord,
                                            BackgroundSync* bgsync,
                                            Reporter* reporter) {
     auto syncTarget = reporter->getTarget();
@@ -139,7 +134,6 @@ Status SyncSourceFeedback::_updateUpstream(OperationContext* txn,
         } else {
             // Blacklist sync target for .5 seconds and find a new one.
             stdx::lock_guard<stdx::mutex> lock(_mtx);
-            auto replCoord = repl::ReplicationCoordinator::get(txn);
             const auto blacklistDuration = Milliseconds{500};
             const auto until = Date_t::now() + blacklistDuration;
             log() << "Blacklisting " << syncTarget << " due to error: '" << status << "' for "
@@ -161,7 +155,9 @@ void SyncSourceFeedback::shutdown() {
     _cond.notify_all();
 }
 
-void SyncSourceFeedback::run(executor::TaskExecutor* executor, BackgroundSync* bgsync) {
+void SyncSourceFeedback::run(executor::TaskExecutor* executor,
+                             BackgroundSync* bgsync,
+                             ReplicationCoordinator* replCoord) {
     Client::initThread("SyncSourceFeedback");
 
     HostAndPort syncTarget;
@@ -170,10 +166,9 @@ void SyncSourceFeedback::run(executor::TaskExecutor* executor, BackgroundSync* b
     Milliseconds keepAliveInterval(0);
 
     while (true) {  // breaks once _shutdownSignaled is true
-        auto txn = cc().makeOperationContext();
 
         if (keepAliveInterval == Milliseconds(0)) {
-            keepAliveInterval = calculateKeepAliveInterval(txn.get(), _mtx);
+            keepAliveInterval = calculateKeepAliveInterval(replCoord->getConfig());
         }
 
         {
@@ -189,7 +184,7 @@ void SyncSourceFeedback::run(executor::TaskExecutor* executor, BackgroundSync* b
                         continue;
                     }
                 }
-                MemberState state = ReplicationCoordinator::get(txn.get())->getMemberState();
+                MemberState state = replCoord->getMemberState();
                 if (!(state.primary() || state.startup())) {
                     break;
                 }
@@ -204,7 +199,7 @@ void SyncSourceFeedback::run(executor::TaskExecutor* executor, BackgroundSync* b
 
         {
             stdx::lock_guard<stdx::mutex> lock(_mtx);
-            MemberState state = ReplicationCoordinator::get(txn.get())->getMemberState();
+            MemberState state = replCoord->getMemberState();
             if (state.primary() || state.startup()) {
                 continue;
             }
@@ -226,7 +221,7 @@ void SyncSourceFeedback::run(executor::TaskExecutor* executor, BackgroundSync* b
 
             // Update keepalive value from config.
             auto oldKeepAliveInterval = keepAliveInterval;
-            keepAliveInterval = calculateKeepAliveInterval(txn.get(), _mtx);
+            keepAliveInterval = calculateKeepAliveInterval(replCoord->getConfig());
             if (oldKeepAliveInterval != keepAliveInterval) {
                 LOG(1) << "new syncSourceFeedback keep alive duration = " << keepAliveInterval
                        << " (previously " << oldKeepAliveInterval << ")";
@@ -235,7 +230,7 @@ void SyncSourceFeedback::run(executor::TaskExecutor* executor, BackgroundSync* b
 
         Reporter reporter(
             executor,
-            makePrepareReplSetUpdatePositionCommandFn(txn.get(), _mtx, syncTarget, bgsync),
+            makePrepareReplSetUpdatePositionCommandFn(replCoord, _mtx, syncTarget, bgsync),
             syncTarget,
             keepAliveInterval);
         {
@@ -250,7 +245,7 @@ void SyncSourceFeedback::run(executor::TaskExecutor* executor, BackgroundSync* b
             _reporter = nullptr;
         });
 
-        auto status = _updateUpstream(txn.get(), bgsync, &reporter);
+        auto status = _updateUpstream(replCoord, bgsync, &reporter);
         if (!status.isOK()) {
             LOG(1) << "The replication progress command (replSetUpdatePosition) failed and will be "
                       "retried: "
diff --git a/src/mongo/db/repl/sync_source_feedback.h b/src/mongo/db/repl/sync_source_feedback.h
index 40c29dc172a..35ed470fc26 100644
--- a/src/mongo/db/repl/sync_source_feedback.h
+++ b/src/mongo/db/repl/sync_source_feedback.h
@@ -31,6 +31,7 @@
 
 #include "mongo/base/disallow_copying.h"
 #include "mongo/base/status.h"
+#include "mongo/db/repl/replication_coordinator.h"
 #include "mongo/stdx/condition_variable.h"
 #include "mongo/stdx/mutex.h"
 
@@ -63,7 +64,9 @@ public:
      *
      * Task executor is used to run replSetUpdatePosition command on sync source.
      */
-    void run(executor::TaskExecutor* executor, BackgroundSync* bgsync);
+    void run(executor::TaskExecutor* executor,
+             BackgroundSync* bgsync,
+             ReplicationCoordinator* replCoord);
 
     /// Signals the run() method to terminate.
     void shutdown();
@@ -72,7 +75,9 @@ private:
     /* Inform the sync target of our current position in the oplog, as well as the positions
      * of all secondaries chained through us.
      */
-    Status _updateUpstream(OperationContext* txn, BackgroundSync* bgsync, Reporter* reporter);
+    Status _updateUpstream(ReplicationCoordinator* replCoord,
+                           BackgroundSync* bgsync,
+                           Reporter* reporter);
 
     // protects cond, _shutdownSignaled, _keepAliveInterval, and _positionChanged.
     stdx::mutex _mtx;
diff --git a/src/mongo/db/repl/sync_source_resolver.cpp b/src/mongo/db/repl/sync_source_resolver.cpp
index 948a471b245..17a706f1b61 100644
--- a/src/mongo/db/repl/sync_source_resolver.cpp
+++ b/src/mongo/db/repl/sync_source_resolver.cpp
@@ -54,6 +54,7 @@ const Seconds SyncSourceResolver::kFirstOplogEntryEmptyBlacklistDuration(10);
 const Seconds SyncSourceResolver::kFirstOplogEntryNullTimestampBlacklistDuration(10);
 const Minutes SyncSourceResolver::kTooStaleBlacklistDuration(1);
 const Seconds SyncSourceResolver::kNoRequiredOpTimeBlacklistDuration(60);
+const int SyncSourceResolver::kUninitializedRollbackId(-1);
 
 SyncSourceResolver::SyncSourceResolver(executor::TaskExecutor* taskExecutor,
                                        SyncSourceSelector* syncSourceSelector,
@@ -175,11 +176,13 @@ std::unique_ptr<Fetcher> SyncSourceResolver::_makeFirstOplogEntryFetcher(
                    candidate,
                    earliestOpTimeSeen),
         rpc::ServerSelectionMetadata(true, boost::none).toBSON(),
-        kFetcherTimeout);
+        kFetcherTimeout /* find network timeout */,
+        kFetcherTimeout /* getMore network timeout */);
 }
 
 std::unique_ptr<Fetcher> SyncSourceResolver::_makeRequiredOpTimeFetcher(HostAndPort candidate,
-                                                                        OpTime earliestOpTimeSeen) {
+                                                                        OpTime earliestOpTimeSeen,
+                                                                        int rbid) {
     // This query is structured so that it is executed on the sync source using the oplog
     // start hack (oplogReplay=true and $gt/$gte predicate over "ts").
     return stdx::make_unique<Fetcher>(
@@ -193,9 +196,11 @@ std::unique_ptr<Fetcher> SyncSourceResolver::_makeRequiredOpTimeFetcher(HostAndP
                    this,
                    stdx::placeholders::_1,
                    candidate,
-                   earliestOpTimeSeen),
+                   earliestOpTimeSeen,
+                   rbid),
         rpc::ServerSelectionMetadata(true, boost::none).toBSON(),
-        kFetcherTimeout);
+        kFetcherTimeout /* find network timeout */,
+        kFetcherTimeout /* getMore network timeout */);
 }
 
 Status SyncSourceResolver::_scheduleFetcher(std::unique_ptr<Fetcher> fetcher) {
@@ -205,6 +210,9 @@ Status SyncSourceResolver::_scheduleFetcher(std::unique_ptr<Fetcher> fetcher) {
     // executor.
     auto status = fetcher->schedule();
     if (status.isOK()) {
+        // Fetcher destruction blocks on all outstanding callbacks. If we are currently in a
+        // Fetcher-related callback, we can't destroy the Fetcher just yet, so we assign it to a
+        // temporary unique pointer to allow the destruction to run to completion.
         _shuttingDownFetcher = std::move(_fetcher);
         _fetcher = std::move(fetcher);
     } else {
@@ -311,10 +319,26 @@ void SyncSourceResolver::_firstOplogEntryFetcherCallback(
         return;
     }
 
-    _scheduleRBIDRequest(candidate, earliestOpTimeSeen);
+    auto status = _scheduleRBIDRequest(candidate, earliestOpTimeSeen);
+    if (!status.isOK()) {
+        _finishCallback(status);
+    }
 }
 
-void SyncSourceResolver::_scheduleRBIDRequest(HostAndPort candidate, OpTime earliestOpTimeSeen) {
+Status SyncSourceResolver::_scheduleRBIDRequest(HostAndPort candidate, OpTime earliestOpTimeSeen) {
+    // Once a work is scheduled, nothing prevents it finishing. We need the mutex to protect the
+    // access of member variables after scheduling, because otherwise the scheduled callback could
+    // finish and allow the destructor to fire before we access the member variables.
+    stdx::lock_guard<stdx::mutex> lk(_mutex);
+    if (_state == State::kShuttingDown) {
+        return Status(
+            ErrorCodes::CallbackCanceled,
+            str::stream()
+                << "sync source resolver shut down while checking rollbackId on candidate: "
+                << candidate);
+    }
+
+    invariant(_state == State::kRunning);
     auto handle = _taskExecutor->scheduleRemoteCommand(
         {candidate, "admin", BSON("replSetGetRBID" << 1), nullptr, kFetcherTimeout},
         stdx::bind(&SyncSourceResolver::_rbidRequestCallback,
@@ -324,15 +348,11 @@ void SyncSourceResolver::_scheduleRBIDRequest(HostAndPort candidate, OpTime earl
                    stdx::placeholders::_1));
 
     if (!handle.isOK()) {
-        _finishCallback(handle.getStatus());
-        return;
+        return handle.getStatus();
     }
 
-    stdx::lock_guard<stdx::mutex> lk(_mutex);
     _rbidCommandHandle = std::move(handle.getValue());
-    if (_state == State::kShuttingDown) {
-        _taskExecutor->cancel(_rbidCommandHandle);
-    }
+    return Status::OK();
 }
 
 void SyncSourceResolver::_rbidRequestCallback(
@@ -344,10 +364,11 @@ void SyncSourceResolver::_rbidRequestCallback(
         return;
     }
 
+    int rbid = kUninitializedRollbackId;
     try {
         uassertStatusOK(rbidReply.response.status);
         uassertStatusOK(getStatusFromCommandResult(rbidReply.response.data));
-        _rbid = rbidReply.response.data["rbid"].Int();
+        rbid = rbidReply.response.data["rbid"].Int();
     } catch (const DBException& ex) {
         const auto until = _taskExecutor->now() + kFetcherErrorBlacklistDuration;
         log() << "Blacklisting " << candidate << " due to error: '" << ex << "' for "
@@ -360,13 +381,15 @@ void SyncSourceResolver::_rbidRequestCallback(
     if (!_requiredOpTime.isNull()) {
         // Schedule fetcher to look for '_requiredOpTime' in the remote oplog.
         // Unittest requires that this kind of failure be handled specially.
-        auto status = _scheduleFetcher(_makeRequiredOpTimeFetcher(candidate, earliestOpTimeSeen));
+        auto status =
+            _scheduleFetcher(_makeRequiredOpTimeFetcher(candidate, earliestOpTimeSeen, rbid));
         if (!status.isOK()) {
             _finishCallback(status);
         }
         return;
     }
-    _finishCallback(candidate);
+
+    _finishCallback(candidate, rbid);
 }
 
 Status SyncSourceResolver::_compareRequiredOpTimeWithQueryResponse(
@@ -399,7 +422,8 @@ Status SyncSourceResolver::_compareRequiredOpTimeWithQueryResponse(
 void SyncSourceResolver::_requiredOpTimeFetcherCallback(
     const StatusWith<Fetcher::QueryResponse>& queryResult,
     HostAndPort candidate,
-    OpTime earliestOpTimeSeen) {
+    OpTime earliestOpTimeSeen,
+    int rbid) {
     if (_isShuttingDown()) {
         _finishCallback(Status(ErrorCodes::CallbackCanceled,
                                str::stream() << "sync source resolver shut down while looking for "
@@ -444,18 +468,18 @@ void SyncSourceResolver::_requiredOpTimeFetcherCallback(
         return;
     }
 
-    _finishCallback(candidate);
+    _finishCallback(candidate, rbid);
 }
 
 Status SyncSourceResolver::_chooseAndProbeNextSyncSource(OpTime earliestOpTimeSeen) {
     auto candidateResult = _chooseNewSyncSource();
     if (!candidateResult.isOK()) {
-        return _finishCallback(candidateResult);
+        return _finishCallback(candidateResult.getStatus());
     }
 
     if (candidateResult.getValue().empty()) {
         if (earliestOpTimeSeen.isNull()) {
-            return _finishCallback(candidateResult);
+            return _finishCallback(candidateResult.getValue(), kUninitializedRollbackId);
         }
 
         SyncSourceResolverResponse response;
@@ -473,16 +497,22 @@ Status SyncSourceResolver::_chooseAndProbeNextSyncSource(OpTime earliestOpTimeSe
     return Status::OK();
 }
 
-Status SyncSourceResolver::_finishCallback(StatusWith<HostAndPort> result) {
+Status SyncSourceResolver::_finishCallback(HostAndPort hostAndPort, int rbid) {
     SyncSourceResolverResponse response;
-    response.syncSourceStatus = std::move(result);
-    if (response.isOK() && !response.getSyncSource().empty()) {
-        invariant(_requiredOpTime.isNull() || _rbid);
-        response.rbid = _rbid;
+    response.syncSourceStatus = std::move(hostAndPort);
+    if (rbid != kUninitializedRollbackId) {
+        response.rbid = rbid;
     }
     return _finishCallback(response);
 }
 
+Status SyncSourceResolver::_finishCallback(Status status) {
+    invariant(!status.isOK());
+    SyncSourceResolverResponse response;
+    response.syncSourceStatus = std::move(status);
+    return _finishCallback(response);
+}
+
 Status SyncSourceResolver::_finishCallback(const SyncSourceResolverResponse& response) {
     try {
         _onCompletion(response);
diff --git a/src/mongo/db/repl/sync_source_resolver.h b/src/mongo/db/repl/sync_source_resolver.h
index 201658caacf..0f9343cfcdb 100644
--- a/src/mongo/db/repl/sync_source_resolver.h
+++ b/src/mongo/db/repl/sync_source_resolver.h
@@ -102,6 +102,7 @@ public:
     static const Seconds kFirstOplogEntryNullTimestampBlacklistDuration;
     static const Minutes kTooStaleBlacklistDuration;
     static const Seconds kNoRequiredOpTimeBlacklistDuration;
+    static const int kUninitializedRollbackId;
 
     /**
      * Callback function to report final status of resolving sync source.
@@ -154,7 +155,8 @@ private:
      * Creates fetcher to check the remote oplog for '_requiredOpTime'.
      */
     std::unique_ptr<Fetcher> _makeRequiredOpTimeFetcher(HostAndPort candidate,
-                                                        OpTime earliestOpTimeSeen);
+                                                        OpTime earliestOpTimeSeen,
+                                                        int rbid);
 
     /**
      * Schedules fetcher to read oplog on sync source.
@@ -179,7 +181,7 @@ private:
     /**
      * Schedules a replSetGetRBID command against the candidate to fetch its current rollback id.
      */
-    void _scheduleRBIDRequest(HostAndPort candidate, OpTime earliestOpTimeSeen);
+    Status _scheduleRBIDRequest(HostAndPort candidate, OpTime earliestOpTimeSeen);
     void _rbidRequestCallback(HostAndPort candidate,
                               OpTime earliestOpTimeSeen,
                               const executor::TaskExecutor::RemoteCommandCallbackArgs& rbidReply);
@@ -194,7 +196,8 @@ private:
      */
     void _requiredOpTimeFetcherCallback(const StatusWith<Fetcher::QueryResponse>& queryResult,
                                         HostAndPort candidate,
-                                        OpTime earliestOpTimeSeen);
+                                        OpTime earliestOpTimeSeen,
+                                        int rbid);
 
     /**
      * Obtains new sync source candidate and schedules remote command to fetcher first oplog entry.
@@ -207,7 +210,8 @@ private:
      * Invokes completion callback and transitions state to State::kComplete.
      * Returns result.getStatus().
      */
-    Status _finishCallback(StatusWith<HostAndPort> result);
+    Status _finishCallback(HostAndPort hostAndPort, int rbid);
+    Status _finishCallback(Status status);
     Status _finishCallback(const SyncSourceResolverResponse& response);
 
     // Executor used to send remote commands to sync source candidates.
@@ -229,9 +233,6 @@ private:
     // resolver via this callback in a SyncSourceResolverResponse struct when the resolver finishes.
     const OnCompletionFn _onCompletion;
 
-    // The rbid we will return to our caller.
-    int _rbid;
-
     // Protects members of this sync source resolver defined below.
     mutable stdx::mutex _mutex;
     mutable stdx::condition_variable _condition;
diff --git a/src/mongo/db/repl/sync_tail.cpp b/src/mongo/db/repl/sync_tail.cpp
index 30f1e28b619..3f65e3fc6b7 100644
--- a/src/mongo/db/repl/sync_tail.cpp
+++ b/src/mongo/db/repl/sync_tail.cpp
@@ -448,7 +448,6 @@ void applyOps(std::vector<MultiApplier::OperationPtrs>& writerVectors,
               const MultiApplier::ApplyOperationFn& func,
               std::vector<Status>* statusVector) {
     invariant(writerVectors.size() == statusVector->size());
-    TimerHolder timer(&applyBatchStats);
     for (size_t i = 0; i < writerVectors.size(); i++) {
         if (!writerVectors[i].empty()) {
             writerPool->schedule([&func, &writerVectors, statusVector, i] {
@@ -695,32 +694,53 @@ public:
     }
 
 private:
+    /**
+     * Calculates batch limit size (in bytes) using the maximum capped collection size of the oplog
+     * size.
+     * Batches are limited to 10% of the oplog.
+     */
+    std::size_t _calculateBatchLimitBytes() {
+        auto opCtx = cc().makeOperationContext();
+        auto storageInterface = StorageInterface::get(opCtx.get());
+        auto oplogMaxSizeResult =
+            storageInterface->getOplogMaxSize(opCtx.get(), NamespaceString(rsOplogName));
+        auto oplogMaxSize = fassertStatusOK(40301, oplogMaxSizeResult);
+        return std::min(oplogMaxSize / 10, std::size_t(replBatchLimitBytes));
+    }
+
+    /**
+     * If slaveDelay is enabled, this function calculates the most recent timestamp of any oplog
+     * entries that can be be returned in a batch.
+     */
+    boost::optional<Date_t> _calculateSlaveDelayLatestTimestamp() {
+        auto service = cc().getServiceContext();
+        auto replCoord = ReplicationCoordinator::get(service);
+        auto slaveDelay = replCoord->getSlaveDelaySecs();
+        if (slaveDelay <= Seconds(0)) {
+            return {};
+        }
+        auto fastClockSource = service->getFastClockSource();
+        return fastClockSource->now() - slaveDelay;
+    }
+
     void run() {
         Client::initThread("ReplBatcher");
-        const ServiceContext::UniqueOperationContext txnPtr = cc().makeOperationContext();
-        OperationContext& txn = *txnPtr;
-        const auto replCoord = ReplicationCoordinator::get(&txn);
-        const auto fastClockSource = txn.getServiceContext()->getFastClockSource();
-        const auto oplogMaxSize = fassertStatusOK(
-            40301,
-            StorageInterface::get(&txn)->getOplogMaxSize(&txn, NamespaceString(rsOplogName)));
 
-        // Batches are limited to 10% of the oplog.
         BatchLimits batchLimits;
-        batchLimits.bytes = std::min(oplogMaxSize / 10, size_t(replBatchLimitBytes));
+        batchLimits.bytes = _calculateBatchLimitBytes();
 
         while (true) {
-            const auto slaveDelay = replCoord->getSlaveDelaySecs();
-            batchLimits.slaveDelayLatestTimestamp = (slaveDelay > Seconds(0))
-                ? (fastClockSource->now() - slaveDelay)
-                : boost::optional<Date_t>();
+            batchLimits.slaveDelayLatestTimestamp = _calculateSlaveDelayLatestTimestamp();
 
             // Check this once per batch since users can change it at runtime.
             batchLimits.ops = replBatchLimitOperations.load();
 
             OpQueue ops;
             // tryPopAndWaitForMore adds to ops and returns true when we need to end a batch early.
-            while (!_syncTail->tryPopAndWaitForMore(&txn, &ops, batchLimits)) {
+            {
+                auto opCtx = cc().makeOperationContext();
+                while (!_syncTail->tryPopAndWaitForMore(opCtx.get(), &ops, batchLimits)) {
+                }
             }
 
             if (ops.empty() && !ops.mustShutdown()) {
@@ -755,14 +775,15 @@ private:
 void SyncTail::oplogApplication(ReplicationCoordinator* replCoord) {
     OpQueueBatcher batcher(this);
 
-    const ServiceContext::UniqueOperationContext txnPtr = cc().makeOperationContext();
-    OperationContext& txn = *txnPtr;
     std::unique_ptr<ApplyBatchFinalizer> finalizer{
         getGlobalServiceContext()->getGlobalStorageEngine()->isDurable()
             ? new ApplyBatchFinalizerForJournal(replCoord)
             : new ApplyBatchFinalizer(replCoord)};
 
     while (true) {  // Exits on message from OpQueueBatcher.
+        const ServiceContext::UniqueOperationContext txnPtr = cc().makeOperationContext();
+        OperationContext& txn = *txnPtr;
+
         // For pausing replication in tests.
         while (MONGO_FAIL_POINT(rsSyncApplyStop)) {
             // Tests should not trigger clean shutdown while that failpoint is active. If we
@@ -1301,6 +1322,9 @@ StatusWith<OpTime> multiApply(OperationContext* txn,
 
     std::vector<Status> statusVector(workerPool->getNumThreads(), Status::OK());
     {
+        // Each node records cumulative batch application stats for itself using this timer.
+        TimerHolder timer(&applyBatchStats);
+
         // We must wait for the all work we've dispatched to complete before leaving this block
         // because the spawned threads refer to objects on our stack, including writerVectors.
         std::vector<MultiApplier::OperationPtrs> writerVectors(workerPool->getNumThreads());
diff --git a/src/mongo/db/repl/topology_coordinator_impl.cpp b/src/mongo/db/repl/topology_coordinator_impl.cpp
index 88337b6a500..f5302008206 100644
--- a/src/mongo/db/repl/topology_coordinator_impl.cpp
+++ b/src/mongo/db/repl/topology_coordinator_impl.cpp
@@ -36,6 +36,7 @@
 
 #include "mongo/db/audit.h"
 #include "mongo/db/client.h"
+#include "mongo/db/mongod_options.h"
 #include "mongo/db/operation_context.h"
 #include "mongo/db/repl/heartbeat_response_action.h"
 #include "mongo/db/repl/is_master_response.h"
@@ -1411,7 +1412,7 @@ bool TopologyCoordinatorImpl::_aMajoritySeemsToBeUp() const {
     return vUp * 2 > _rsConfig.getTotalVotingMembers();
 }
 
-bool TopologyCoordinatorImpl::_canSeeHealthyPrimaryOfEqualOrGreaterPriority(
+int TopologyCoordinatorImpl::_findHealthyPrimaryOfEqualOrGreaterPriority(
     const int candidateIndex) const {
     const double candidatePriority = _rsConfig.getMemberAt(candidateIndex).getPriority();
     for (auto it = _hbdata.begin(); it != _hbdata.end(); ++it) {
@@ -1421,11 +1422,11 @@ bool TopologyCoordinatorImpl::_canSeeHealthyPrimaryOfEqualOrGreaterPriority(
         const int itIndex = indexOfIterator(_hbdata, it);
         const double priority = _rsConfig.getMemberAt(itIndex).getPriority();
         if (itIndex != candidateIndex && priority >= candidatePriority) {
-            return true;
+            return itIndex;
         }
     }
 
-    return false;
+    return -1;
 }
 
 bool TopologyCoordinatorImpl::_isOpTimeCloseEnoughToLatestToElect(
@@ -2232,7 +2233,7 @@ MemberState TopologyCoordinatorImpl::getMemberState() const {
     }
 
     if (_rsConfig.isConfigServer()) {
-        if (_options.clusterRole != ClusterRole::ConfigServer) {
+        if (_options.clusterRole != ClusterRole::ConfigServer && !skipShardingConfigurationChecks) {
             return MemberState::RS_REMOVED;
         } else {
             invariant(_storageEngineSupportsReadCommitted != ReadCommittedSupport::kUnknown);
@@ -2241,7 +2242,7 @@ MemberState TopologyCoordinatorImpl::getMemberState() const {
             }
         }
     } else {
-        if (_options.clusterRole == ClusterRole::ConfigServer) {
+        if (_options.clusterRole == ClusterRole::ConfigServer && !skipShardingConfigurationChecks) {
             return MemberState::RS_REMOVED;
         }
     }
@@ -2571,29 +2572,54 @@ void TopologyCoordinatorImpl::processReplSetRequestVotes(const ReplSetRequestVot
 
     if (args.getTerm() < _term) {
         response->setVoteGranted(false);
-        response->setReason("candidate's term is lower than mine");
+        response->setReason(str::stream() << "candidate's term (" << args.getTerm()
+                                          << ") is lower than mine ("
+                                          << _term
+                                          << ")");
     } else if (args.getConfigVersion() != _rsConfig.getConfigVersion()) {
         response->setVoteGranted(false);
-        response->setReason("candidate's config version differs from mine");
+        response->setReason(str::stream() << "candidate's config version ("
+                                          << args.getConfigVersion()
+                                          << ") differs from mine ("
+                                          << _rsConfig.getConfigVersion()
+                                          << ")");
     } else if (args.getSetName() != _rsConfig.getReplSetName()) {
         response->setVoteGranted(false);
-        response->setReason("candidate's set name differs from mine");
+        response->setReason(str::stream() << "candidate's set name (" << args.getSetName()
+                                          << ") differs from mine ("
+                                          << _rsConfig.getReplSetName()
+                                          << ")");
     } else if (args.getLastDurableOpTime() < lastAppliedOpTime) {
         response->setVoteGranted(false);
-        response->setReason("candidate's data is staler than mine");
+        response
+            ->setReason(str::stream()
+                        << "candidate's data is staler than mine. candidate's last applied OpTime: "
+                        << args.getLastDurableOpTime().toString()
+                        << ", my last applied OpTime: "
+                        << lastAppliedOpTime.toString());
     } else if (!args.isADryRun() && _lastVote.getTerm() == args.getTerm()) {
         response->setVoteGranted(false);
-        response->setReason("already voted for another candidate this term");
-    } else if (_selfConfig().isArbiter() &&
-               _canSeeHealthyPrimaryOfEqualOrGreaterPriority(args.getCandidateIndex())) {
-        response->setVoteGranted(false);
-        response->setReason("can see a healthy primary of equal or greater priority");
+        response->setReason(str::stream()
+                            << "already voted for another candidate ("
+                            << _rsConfig.getMemberAt(_lastVote.getCandidateIndex()).getHostAndPort()
+                            << ") this term ("
+                            << _lastVote.getTerm()
+                            << ")");
     } else {
-        if (!args.isADryRun()) {
-            _lastVote.setTerm(args.getTerm());
-            _lastVote.setCandidateIndex(args.getCandidateIndex());
+        int betterPrimary = _findHealthyPrimaryOfEqualOrGreaterPriority(args.getCandidateIndex());
+        if (_selfConfig().isArbiter() && betterPrimary >= 0) {
+            response->setVoteGranted(false);
+            response->setReason(str::stream()
+                                << "can see a healthy primary ("
+                                << _rsConfig.getMemberAt(betterPrimary).getHostAndPort()
+                                << ") of equal or greater priority");
+        } else {
+            if (!args.isADryRun()) {
+                _lastVote.setTerm(args.getTerm());
+                _lastVote.setCandidateIndex(args.getCandidateIndex());
+            }
+            response->setVoteGranted(true);
         }
-        response->setVoteGranted(true);
     }
 }
 
diff --git a/src/mongo/db/repl/topology_coordinator_impl.h b/src/mongo/db/repl/topology_coordinator_impl.h
index d4b0476cde1..1a03338e3c4 100644
--- a/src/mongo/db/repl/topology_coordinator_impl.h
+++ b/src/mongo/db/repl/topology_coordinator_impl.h
@@ -310,9 +310,9 @@ private:
     // Sees if a majority number of votes are held by members who are currently "up"
     bool _aMajoritySeemsToBeUp() const;
 
-    // Returns true if the node can see a healthy primary of equal or greater priority to the
-    // candidate.
-    bool _canSeeHealthyPrimaryOfEqualOrGreaterPriority(const int candidateIndex) const;
+    // Checks if the node can see a healthy primary of equal or greater priority to the
+    // candidate. If so, returns the index of that node. Otherwise returns -1.
+    int _findHealthyPrimaryOfEqualOrGreaterPriority(const int candidateIndex) const;
 
     // Is otherOpTime close enough (within 10 seconds) to the latest known optime to qualify
     // for an election
diff --git a/src/mongo/db/repl/topology_coordinator_impl_test.cpp b/src/mongo/db/repl/topology_coordinator_impl_test.cpp
index e3708a9bf54..1f8d4b56c40 100644
--- a/src/mongo/db/repl/topology_coordinator_impl_test.cpp
+++ b/src/mongo/db/repl/topology_coordinator_impl_test.cpp
@@ -5912,7 +5912,8 @@ TEST_F(TopoCoordTest, NodeDoesNotGrantVotesToTwoDifferentNodesInTheSameTerm) {
 
     // different candidate same term, should be a problem
     getTopoCoord().processReplSetRequestVotes(args2, &response2, lastAppliedOpTime);
-    ASSERT_EQUALS("already voted for another candidate this term", response2.getReason());
+    ASSERT_EQUALS("already voted for another candidate (hself:27017) this term (1)",
+                  response2.getReason());
     ASSERT_FALSE(response2.getVoteGranted());
 }
 
@@ -6028,7 +6029,8 @@ TEST_F(TopoCoordTest, VoteRequestShouldNotPreventDryRunsForThatTerm) {
     ReplSetRequestVotesResponse response2;
 
     getTopoCoord().processReplSetRequestVotes(args2, &response2, lastAppliedOpTime);
-    ASSERT_EQUALS("already voted for another candidate this term", response2.getReason());
+    ASSERT_EQUALS("already voted for another candidate (hself:27017) this term (1)",
+                  response2.getReason());
     ASSERT_FALSE(response2.getVoteGranted());
 }
 
@@ -6063,7 +6065,7 @@ TEST_F(TopoCoordTest, NodeDoesNotGrantVoteWhenReplSetNameDoesNotMatch) {
     OpTime lastAppliedOpTime;
 
     getTopoCoord().processReplSetRequestVotes(args, &response, lastAppliedOpTime);
-    ASSERT_EQUALS("candidate's set name differs from mine", response.getReason());
+    ASSERT_EQUALS("candidate's set name (wrongName) differs from mine (rs0)", response.getReason());
     ASSERT_FALSE(response.getVoteGranted());
 }
 
@@ -6098,7 +6100,7 @@ TEST_F(TopoCoordTest, NodeDoesNotGrantVoteWhenConfigVersionDoesNotMatch) {
     OpTime lastAppliedOpTime;
 
     getTopoCoord().processReplSetRequestVotes(args, &response, lastAppliedOpTime);
-    ASSERT_EQUALS("candidate's config version differs from mine", response.getReason());
+    ASSERT_EQUALS("candidate's config version (0) differs from mine (1)", response.getReason());
     ASSERT_FALSE(response.getVoteGranted());
 }
 
@@ -6145,7 +6147,8 @@ TEST_F(TopoCoordTest, ArbiterDoesNotGrantVoteWhenItCanSeeAHealthyPrimaryOfEqualO
     OpTime lastAppliedOpTime;
 
     getTopoCoord().processReplSetRequestVotes(args, &response, lastAppliedOpTime);
-    ASSERT_EQUALS("can see a healthy primary of equal or greater priority", response.getReason());
+    ASSERT_EQUALS("can see a healthy primary (h2:27017) of equal or greater priority",
+                  response.getReason());
     ASSERT_FALSE(response.getVoteGranted());
 }
 
@@ -6184,7 +6187,7 @@ TEST_F(TopoCoordTest, NodeDoesNotGrantVoteWhenTermIsStale) {
     OpTime lastAppliedOpTime;
 
     getTopoCoord().processReplSetRequestVotes(args, &response, lastAppliedOpTime);
-    ASSERT_EQUALS("candidate's term is lower than mine", response.getReason());
+    ASSERT_EQUALS("candidate's term (1) is lower than mine (2)", response.getReason());
     ASSERT_EQUALS(2, response.getTerm());
     ASSERT_FALSE(response.getVoteGranted());
 }
@@ -6221,7 +6224,12 @@ TEST_F(TopoCoordTest, NodeDoesNotGrantVoteWhenOpTimeIsStale) {
     OpTime lastAppliedOpTime2 = {Timestamp(20, 0), 0};
 
     getTopoCoord().processReplSetRequestVotes(args, &response, lastAppliedOpTime2);
-    ASSERT_EQUALS("candidate's data is staler than mine", response.getReason());
+    ASSERT_EQUALS(
+        str::stream() << "candidate's data is staler than mine. candidate's last applied OpTime: "
+                      << OpTime().toString()
+                      << ", my last applied OpTime: "
+                      << OpTime(Timestamp(20, 0), 0).toString(),
+        response.getReason());
     ASSERT_FALSE(response.getVoteGranted());
 }
 
@@ -6281,7 +6289,7 @@ TEST_F(TopoCoordTest, NodeDoesNotGrantDryRunVoteWhenReplSetNameDoesNotMatch) {
     ReplSetRequestVotesResponse response;
 
     getTopoCoord().processReplSetRequestVotes(args, &response, lastAppliedOpTime);
-    ASSERT_EQUALS("candidate's set name differs from mine", response.getReason());
+    ASSERT_EQUALS("candidate's set name (wrongName) differs from mine (rs0)", response.getReason());
     ASSERT_EQUALS(1, response.getTerm());
     ASSERT_FALSE(response.getVoteGranted());
 }
@@ -6342,7 +6350,7 @@ TEST_F(TopoCoordTest, NodeDoesNotGrantDryRunVoteWhenConfigVersionDoesNotMatch) {
     ReplSetRequestVotesResponse response;
 
     getTopoCoord().processReplSetRequestVotes(args, &response, lastAppliedOpTime);
-    ASSERT_EQUALS("candidate's config version differs from mine", response.getReason());
+    ASSERT_EQUALS("candidate's config version (0) differs from mine (1)", response.getReason());
     ASSERT_EQUALS(1, response.getTerm());
     ASSERT_FALSE(response.getVoteGranted());
 }
@@ -6402,7 +6410,7 @@ TEST_F(TopoCoordTest, NodeDoesNotGrantDryRunVoteWhenTermIsStale) {
     ReplSetRequestVotesResponse response;
 
     getTopoCoord().processReplSetRequestVotes(args, &response, lastAppliedOpTime);
-    ASSERT_EQUALS("candidate's term is lower than mine", response.getReason());
+    ASSERT_EQUALS("candidate's term (0) is lower than mine (1)", response.getReason());
     ASSERT_EQUALS(1, response.getTerm());
     ASSERT_FALSE(response.getVoteGranted());
 }
@@ -6525,7 +6533,12 @@ TEST_F(TopoCoordTest, DoNotGrantDryRunVoteWhenOpTimeIsStale) {
     OpTime lastAppliedOpTime2 = {Timestamp(20, 0), 0};
 
     getTopoCoord().processReplSetRequestVotes(args, &response, lastAppliedOpTime2);
-    ASSERT_EQUALS("candidate's data is staler than mine", response.getReason());
+    ASSERT_EQUALS(
+        str::stream() << "candidate's data is staler than mine. candidate's last applied OpTime: "
+                      << OpTime().toString()
+                      << ", my last applied OpTime: "
+                      << OpTime(Timestamp(20, 0), 0).toString(),
+        response.getReason());
     ASSERT_EQUALS(1, response.getTerm());
     ASSERT_FALSE(response.getVoteGranted());
 }
diff --git a/src/mongo/db/repl/topology_coordinator_impl_v1_test.cpp b/src/mongo/db/repl/topology_coordinator_impl_v1_test.cpp
index 8062208949b..8f60269cd71 100644
--- a/src/mongo/db/repl/topology_coordinator_impl_v1_test.cpp
+++ b/src/mongo/db/repl/topology_coordinator_impl_v1_test.cpp
@@ -2574,7 +2574,8 @@ TEST_F(TopoCoordTest, NodeDoesNotGrantVotesToTwoDifferentNodesInTheSameTerm) {
 
     // different candidate same term, should be a problem
     getTopoCoord().processReplSetRequestVotes(args2, &response2, lastAppliedOpTime);
-    ASSERT_EQUALS("already voted for another candidate this term", response2.getReason());
+    ASSERT_EQUALS("already voted for another candidate (hself:27017) this term (1)",
+                  response2.getReason());
     ASSERT_FALSE(response2.getVoteGranted());
 }
 
@@ -2674,7 +2675,8 @@ TEST_F(TopoCoordTest, DryRunVoteRequestShouldNotPreventSubsequentDryRunsForThatT
     ReplSetRequestVotesResponse response4;
 
     getTopoCoord().processReplSetRequestVotes(args4, &response4, lastAppliedOpTime);
-    ASSERT_EQUALS("already voted for another candidate this term", response4.getReason());
+    ASSERT_EQUALS("already voted for another candidate (hself:27017) this term (1)",
+                  response4.getReason());
     ASSERT_FALSE(response4.getVoteGranted());
 }
 
@@ -2732,7 +2734,8 @@ TEST_F(TopoCoordTest, VoteRequestShouldNotPreventDryRunsForThatTerm) {
     ReplSetRequestVotesResponse response2;
 
     getTopoCoord().processReplSetRequestVotes(args2, &response2, lastAppliedOpTime);
-    ASSERT_EQUALS("already voted for another candidate this term", response2.getReason());
+    ASSERT_EQUALS("already voted for another candidate (hself:27017) this term (1)",
+                  response2.getReason());
     ASSERT_FALSE(response2.getVoteGranted());
 }
 
@@ -2767,7 +2770,7 @@ TEST_F(TopoCoordTest, NodeDoesNotGrantVoteWhenReplSetNameDoesNotMatch) {
     OpTime lastAppliedOpTime;
 
     getTopoCoord().processReplSetRequestVotes(args, &response, lastAppliedOpTime);
-    ASSERT_EQUALS("candidate's set name differs from mine", response.getReason());
+    ASSERT_EQUALS("candidate's set name (wrongName) differs from mine (rs0)", response.getReason());
     ASSERT_FALSE(response.getVoteGranted());
 }
 
@@ -2802,7 +2805,7 @@ TEST_F(TopoCoordTest, NodeDoesNotGrantVoteWhenConfigVersionDoesNotMatch) {
     OpTime lastAppliedOpTime;
 
     getTopoCoord().processReplSetRequestVotes(args, &response, lastAppliedOpTime);
-    ASSERT_EQUALS("candidate's config version differs from mine", response.getReason());
+    ASSERT_EQUALS("candidate's config version (0) differs from mine (1)", response.getReason());
     ASSERT_FALSE(response.getVoteGranted());
 }
 
@@ -2841,7 +2844,7 @@ TEST_F(TopoCoordTest, NodeDoesNotGrantVoteWhenTermIsStale) {
     OpTime lastAppliedOpTime;
 
     getTopoCoord().processReplSetRequestVotes(args, &response, lastAppliedOpTime);
-    ASSERT_EQUALS("candidate's term is lower than mine", response.getReason());
+    ASSERT_EQUALS("candidate's term (1) is lower than mine (2)", response.getReason());
     ASSERT_EQUALS(2, response.getTerm());
     ASSERT_FALSE(response.getVoteGranted());
 }
@@ -2878,7 +2881,12 @@ TEST_F(TopoCoordTest, NodeDoesNotGrantVoteWhenOpTimeIsStale) {
     OpTime lastAppliedOpTime2 = {Timestamp(20, 0), 0};
 
     getTopoCoord().processReplSetRequestVotes(args, &response, lastAppliedOpTime2);
-    ASSERT_EQUALS("candidate's data is staler than mine", response.getReason());
+    ASSERT_EQUALS(
+        str::stream() << "candidate's data is staler than mine. candidate's last applied OpTime: "
+                      << OpTime().toString()
+                      << ", my last applied OpTime: "
+                      << OpTime(Timestamp(20, 0), 0).toString(),
+        response.getReason());
     ASSERT_FALSE(response.getVoteGranted());
 }
 
@@ -2938,7 +2946,7 @@ TEST_F(TopoCoordTest, NodeDoesNotGrantDryRunVoteWhenReplSetNameDoesNotMatch) {
     ReplSetRequestVotesResponse response;
 
     getTopoCoord().processReplSetRequestVotes(args, &response, lastAppliedOpTime);
-    ASSERT_EQUALS("candidate's set name differs from mine", response.getReason());
+    ASSERT_EQUALS("candidate's set name (wrongName) differs from mine (rs0)", response.getReason());
     ASSERT_EQUALS(1, response.getTerm());
     ASSERT_FALSE(response.getVoteGranted());
 }
@@ -2999,7 +3007,7 @@ TEST_F(TopoCoordTest, NodeDoesNotGrantDryRunVoteWhenConfigVersionDoesNotMatch) {
     ReplSetRequestVotesResponse response;
 
     getTopoCoord().processReplSetRequestVotes(args, &response, lastAppliedOpTime);
-    ASSERT_EQUALS("candidate's config version differs from mine", response.getReason());
+    ASSERT_EQUALS("candidate's config version (0) differs from mine (1)", response.getReason());
     ASSERT_EQUALS(1, response.getTerm());
     ASSERT_FALSE(response.getVoteGranted());
 }
@@ -3059,7 +3067,7 @@ TEST_F(TopoCoordTest, NodeDoesNotGrantDryRunVoteWhenTermIsStale) {
     ReplSetRequestVotesResponse response;
 
     getTopoCoord().processReplSetRequestVotes(args, &response, lastAppliedOpTime);
-    ASSERT_EQUALS("candidate's term is lower than mine", response.getReason());
+    ASSERT_EQUALS("candidate's term (0) is lower than mine (1)", response.getReason());
     ASSERT_EQUALS(1, response.getTerm());
     ASSERT_FALSE(response.getVoteGranted());
 }
@@ -3182,7 +3190,12 @@ TEST_F(TopoCoordTest, DoNotGrantDryRunVoteWhenOpTimeIsStale) {
     OpTime lastAppliedOpTime2 = {Timestamp(20, 0), 0};
 
     getTopoCoord().processReplSetRequestVotes(args, &response, lastAppliedOpTime2);
-    ASSERT_EQUALS("candidate's data is staler than mine", response.getReason());
+    ASSERT_EQUALS(
+        str::stream() << "candidate's data is staler than mine. candidate's last applied OpTime: "
+                      << OpTime().toString()
+                      << ", my last applied OpTime: "
+                      << OpTime(Timestamp(20, 0), 0).toString(),
+        response.getReason());
     ASSERT_EQUALS(1, response.getTerm());
     ASSERT_FALSE(response.getVoteGranted());
 }
diff --git a/src/mongo/db/s/SConscript b/src/mongo/db/s/SConscript
index b2abb86c056..b4c8b12a3d2 100644
--- a/src/mongo/db/s/SConscript
+++ b/src/mongo/db/s/SConscript
@@ -47,7 +47,6 @@ env.Library(
     source=[
         'active_migrations_registry.cpp',
         'chunk_move_write_concern_options.cpp',
-        'collection_range_deleter.cpp',
         'collection_sharding_state.cpp',
         'metadata_manager.cpp',
         'migration_chunk_cloner_source.cpp',
diff --git a/src/mongo/db/s/balancer/balancer.cpp b/src/mongo/db/s/balancer/balancer.cpp
index 36987fb81ec..7e334b8b86e 100644
--- a/src/mongo/db/s/balancer/balancer.cpp
+++ b/src/mongo/db/s/balancer/balancer.cpp
@@ -243,8 +243,9 @@ void Balancer::waitForBalancerToStop() {
 void Balancer::joinCurrentRound(OperationContext* opCtx) {
     stdx::unique_lock<stdx::mutex> scopedLock(_mutex);
     const auto numRoundsAtStart = _numBalancerRounds;
-    _condVar.wait(scopedLock,
-                  [&] { return !_inBalancerRound || _numBalancerRounds != numRoundsAtStart; });
+    opCtx->waitForConditionOrInterrupt(_condVar, scopedLock, [&] {
+        return !_inBalancerRound || _numBalancerRounds != numRoundsAtStart;
+    });
 }
 
 Status Balancer::rebalanceSingleChunk(OperationContext* opCtx, const ChunkType& chunk) {
diff --git a/src/mongo/db/s/balancer/balancer.h b/src/mongo/db/s/balancer/balancer.h
index d8847a4ee90..7f601a849e2 100644
--- a/src/mongo/db/s/balancer/balancer.h
+++ b/src/mongo/db/s/balancer/balancer.h
@@ -110,7 +110,8 @@ public:
 
     /**
      * Potentially blocking method, which will return immediately if the balancer is not running a
-     * balancer round and will block until the current round completes otherwise.
+     * balancer round and will block until the current round completes otherwise. If the operation
+     * context's deadline is exceeded, it will throw an ExceededTimeLimit exception.
      */
     void joinCurrentRound(OperationContext* txn);
 
diff --git a/src/mongo/db/s/balancer/balancer_chunk_selection_policy_impl.cpp b/src/mongo/db/s/balancer/balancer_chunk_selection_policy_impl.cpp
index 5d96be7004b..2f7c6feb7ec 100644
--- a/src/mongo/db/s/balancer/balancer_chunk_selection_policy_impl.cpp
+++ b/src/mongo/db/s/balancer/balancer_chunk_selection_policy_impl.cpp
@@ -32,7 +32,6 @@
 
 #include "mongo/db/s/balancer/balancer_chunk_selection_policy_impl.h"
 
-#include <set>
 #include <vector>
 
 #include "mongo/base/status_with.h"
@@ -59,7 +58,7 @@ namespace {
 
 /**
  * Does a linear pass over the information cached in the specified chunk manager and extracts chunk
- * distrubution and chunk placement information which is needed by the balancer policy.
+ * distribution and chunk placement information which is needed by the balancer policy.
  */
 StatusWith<DistributionStatus> createCollectionDistributionStatus(
     OperationContext* opCtx, const ShardStatisticsVector& allShards, ChunkManager* chunkMgr) {
@@ -258,6 +257,7 @@ StatusWith<MigrateInfoVector> BalancerChunkSelectionPolicyImpl::selectChunksToMo
     }
 
     MigrateInfoVector candidateChunks;
+    std::set<ShardId> usedShards;
 
     for (const auto& coll : collections) {
         if (coll.getDropped()) {
@@ -271,8 +271,8 @@ StatusWith<MigrateInfoVector> BalancerChunkSelectionPolicyImpl::selectChunksToMo
             continue;
         }
 
-        auto candidatesStatus =
-            _getMigrateCandidatesForCollection(opCtx, nss, shardStats, aggressiveBalanceHint);
+        auto candidatesStatus = _getMigrateCandidatesForCollection(
+            opCtx, nss, shardStats, aggressiveBalanceHint, &usedShards);
         if (candidatesStatus == ErrorCodes::NamespaceNotFound) {
             // Namespace got dropped before we managed to get to it, so just skip it
             continue;
@@ -416,7 +416,8 @@ StatusWith<MigrateInfoVector> BalancerChunkSelectionPolicyImpl::_getMigrateCandi
     OperationContext* opCtx,
     const NamespaceString& nss,
     const ShardStatisticsVector& shardStats,
-    bool aggressiveBalanceHint) {
+    bool aggressiveBalanceHint,
+    std::set<ShardId>* usedShards) {
     auto routingInfoStatus =
         Grid::get(opCtx)->catalogCache()->getShardedCollectionRoutingInfoWithRefresh(opCtx, nss);
     if (!routingInfoStatus.isOK()) {
@@ -475,7 +476,7 @@ StatusWith<MigrateInfoVector> BalancerChunkSelectionPolicyImpl::_getMigrateCandi
         }
     }
 
-    return BalancerPolicy::balance(shardStats, distribution, aggressiveBalanceHint);
+    return BalancerPolicy::balance(shardStats, distribution, aggressiveBalanceHint, usedShards);
 }
 
 }  // namespace mongo
diff --git a/src/mongo/db/s/balancer/balancer_chunk_selection_policy_impl.h b/src/mongo/db/s/balancer/balancer_chunk_selection_policy_impl.h
index f010d8c723b..5a1277cd5ea 100644
--- a/src/mongo/db/s/balancer/balancer_chunk_selection_policy_impl.h
+++ b/src/mongo/db/s/balancer/balancer_chunk_selection_policy_impl.h
@@ -67,7 +67,8 @@ private:
         OperationContext* txn,
         const NamespaceString& nss,
         const ShardStatisticsVector& shardStats,
-        bool aggressiveBalanceHint);
+        bool aggressiveBalanceHint,
+        std::set<ShardId>* usedShards);
 
     // Source for obtaining cluster statistics. Not owned and must not be destroyed before the
     // policy object is destroyed.
diff --git a/src/mongo/db/s/balancer/balancer_policy.cpp b/src/mongo/db/s/balancer/balancer_policy.cpp
index 348d8cec603..f5a0473d93e 100644
--- a/src/mongo/db/s/balancer/balancer_policy.cpp
+++ b/src/mongo/db/s/balancer/balancer_policy.cpp
@@ -1,30 +1,30 @@
 /**
-*    Copyright (C) 2010 10gen Inc.
-*
-*    This program is free software: you can redistribute it and/or  modify
-*    it under the terms of the GNU Affero General Public License, version 3,
-*    as published by the Free Software Foundation.
-*
-*    This program is distributed in the hope that it will be useful,
-*    but WITHOUT ANY WARRANTY; without even the implied warranty of
-*    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-*    GNU Affero General Public License for more details.
-*
-*    You should have received a copy of the GNU Affero General Public License
-*    along with this program.  If not, see <http://www.gnu.org/licenses/>.
-*
-*    As a special exception, the copyright holders give permission to link the
-*    code of portions of this program with the OpenSSL library under certain
-*    conditions as described in each individual source file and distribute
-*    linked combinations including the program with the OpenSSL library. You
-*    must comply with the GNU Affero General Public License in all respects
-*    for all of the code used other than as permitted herein. If you modify
-*    file(s) with this exception, you may extend this exception to your
-*    version of the file(s), but you are not obligated to do so. If you do not
-*    wish to do so, delete this exception statement from your version. If you
-*    delete this exception statement from all source files in the program,
-*    then also delete it in the license file.
-*/
+ *    Copyright (C) 2018 MongoDB Inc.
+ *
+ *    This program is free software: you can redistribute it and/or  modify
+ *    it under the terms of the GNU Affero General Public License, version 3,
+ *    as published by the Free Software Foundation.
+ *
+ *    This program is distributed in the hope that it will be useful,
+ *    but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *    GNU Affero General Public License for more details.
+ *
+ *    You should have received a copy of the GNU Affero General Public License
+ *    along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ *    As a special exception, the copyright holders give permission to link the
+ *    code of portions of this program with the OpenSSL library under certain
+ *    conditions as described in each individual source file and distribute
+ *    linked combinations including the program with the OpenSSL library. You
+ *    must comply with the GNU Affero General Public License in all respects for
+ *    all of the code used other than as permitted herein. If you modify file(s)
+ *    with this exception, you may extend this exception to your version of the
+ *    file(s), but you are not obligated to do so. If you do not wish to do so,
+ *    delete this exception statement from your version. If you delete this
+ *    exception statement from all source files in the program, then also delete
+ *    it in the license file.
+ */
 
 #define MONGO_LOG_DEFAULT_COMPONENT ::mongo::logger::LogComponent::kSharding
 
@@ -32,7 +32,6 @@
 
 #include "mongo/db/s/balancer/balancer_policy.h"
 
-#include "mongo/bson/simple_bsonobj_comparator.h"
 #include "mongo/s/catalog/type_shard.h"
 #include "mongo/s/catalog/type_tags.h"
 #include "mongo/util/log.h"
@@ -291,20 +290,17 @@ ShardId BalancerPolicy::_getMostOverloadedShard(const ShardStatisticsVector& sha
 
 vector<MigrateInfo> BalancerPolicy::balance(const ShardStatisticsVector& shardStats,
                                             const DistributionStatus& distribution,
-                                            bool shouldAggressivelyBalance) {
+                                            bool shouldAggressivelyBalance,
+                                            std::set<ShardId>* usedShards) {
     vector<MigrateInfo> migrations;
 
-    // Set of shards, which have already been used for migrations. Used so we don't return multiple
-    // migrations for the same shard.
-    set<ShardId> usedShards;
-
     // 1) Check for shards, which are in draining mode
     {
         for (const auto& stat : shardStats) {
             if (!stat.isDraining)
                 continue;
 
-            if (usedShards.count(stat.shardId))
+            if (usedShards->count(stat.shardId))
                 continue;
 
             const vector<ChunkType>& chunks = distribution.getChunks(stat.shardId);
@@ -326,7 +322,7 @@ vector<MigrateInfo> BalancerPolicy::balance(const ShardStatisticsVector& shardSt
                 const string tag = distribution.getTagForChunk(chunk);
 
                 const ShardId to =
-                    _getLeastLoadedReceiverShard(shardStats, distribution, tag, usedShards);
+                    _getLeastLoadedReceiverShard(shardStats, distribution, tag, *usedShards);
                 if (!to.isValid()) {
                     if (migrations.empty()) {
                         warning() << "Chunk " << redact(chunk.toString())
@@ -337,8 +333,8 @@ vector<MigrateInfo> BalancerPolicy::balance(const ShardStatisticsVector& shardSt
 
                 invariant(to != stat.shardId);
                 migrations.emplace_back(to, chunk);
-                invariant(usedShards.insert(stat.shardId).second);
-                invariant(usedShards.insert(to).second);
+                invariant(usedShards->insert(stat.shardId).second);
+                invariant(usedShards->insert(to).second);
                 break;
             }
 
@@ -352,7 +348,7 @@ vector<MigrateInfo> BalancerPolicy::balance(const ShardStatisticsVector& shardSt
     // 2) Check for chunks, which are on the wrong shard and must be moved off of it
     if (!distribution.tags().empty()) {
         for (const auto& stat : shardStats) {
-            if (usedShards.count(stat.shardId))
+            if (usedShards->count(stat.shardId))
                 continue;
 
             const vector<ChunkType>& chunks = distribution.getChunks(stat.shardId);
@@ -373,7 +369,7 @@ vector<MigrateInfo> BalancerPolicy::balance(const ShardStatisticsVector& shardSt
                 }
 
                 const ShardId to =
-                    _getLeastLoadedReceiverShard(shardStats, distribution, tag, usedShards);
+                    _getLeastLoadedReceiverShard(shardStats, distribution, tag, *usedShards);
                 if (!to.isValid()) {
                     if (migrations.empty()) {
                         warning() << "Chunk " << redact(chunk.toString()) << " violates zone "
@@ -384,8 +380,8 @@ vector<MigrateInfo> BalancerPolicy::balance(const ShardStatisticsVector& shardSt
 
                 invariant(to != stat.shardId);
                 migrations.emplace_back(to, chunk);
-                invariant(usedShards.insert(stat.shardId).second);
-                invariant(usedShards.insert(to).second);
+                invariant(usedShards->insert(stat.shardId).second);
+                invariant(usedShards->insert(to).second);
                 break;
             }
         }
@@ -434,7 +430,7 @@ vector<MigrateInfo> BalancerPolicy::balance(const ShardStatisticsVector& shardSt
                                   idealNumberOfChunksPerShardForTag,
                                   imbalanceThreshold,
                                   &migrations,
-                                  &usedShards))
+                                  usedShards))
             ;
     }
 
diff --git a/src/mongo/db/s/balancer/balancer_policy.h b/src/mongo/db/s/balancer/balancer_policy.h
index 63f44c3ca68..afba1db4915 100644
--- a/src/mongo/db/s/balancer/balancer_policy.h
+++ b/src/mongo/db/s/balancer/balancer_policy.h
@@ -1,39 +1,42 @@
 /**
-*    Copyright (C) 2010 10gen Inc.
-*
-*    This program is free software: you can redistribute it and/or  modify
-*    it under the terms of the GNU Affero General Public License, version 3,
-*    as published by the Free Software Foundation.
-*
-*    This program is distributed in the hope that it will be useful,
-*    but WITHOUT ANY WARRANTY; without even the implied warranty of
-*    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-*    GNU Affero General Public License for more details.
-*
-*    You should have received a copy of the GNU Affero General Public License
-*    along with this program.  If not, see <http://www.gnu.org/licenses/>.
-*
-*    As a special exception, the copyright holders give permission to link the
-*    code of portions of this program with the OpenSSL library under certain
-*    conditions as described in each individual source file and distribute
-*    linked combinations including the program with the OpenSSL library. You
-*    must comply with the GNU Affero General Public License in all respects
-*    for all of the code used other than as permitted herein. If you modify
-*    file(s) with this exception, you may extend this exception to your
-*    version of the file(s), but you are not obligated to do so. If you do not
-*    wish to do so, delete this exception statement from your version. If you
-*    delete this exception statement from all source files in the program,
-*    then also delete it in the license file.
-*/
+ *    Copyright (C) 2018 MongoDB Inc.
+ *
+ *    This program is free software: you can redistribute it and/or  modify
+ *    it under the terms of the GNU Affero General Public License, version 3,
+ *    as published by the Free Software Foundation.
+ *
+ *    This program is distributed in the hope that it will be useful,
+ *    but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *    GNU Affero General Public License for more details.
+ *
+ *    You should have received a copy of the GNU Affero General Public License
+ *    along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ *    As a special exception, the copyright holders give permission to link the
+ *    code of portions of this program with the OpenSSL library under certain
+ *    conditions as described in each individual source file and distribute
+ *    linked combinations including the program with the OpenSSL library. You
+ *    must comply with the GNU Affero General Public License in all respects for
+ *    all of the code used other than as permitted herein. If you modify file(s)
+ *    with this exception, you may extend this exception to your version of the
+ *    file(s), but you are not obligated to do so. If you do not wish to do so,
+ *    delete this exception statement from your version. If you delete this
+ *    exception statement from all source files in the program, then also delete
+ *    it in the license file.
+ */
 
 #pragma once
 
+#include <set>
+#include <vector>
+
 #include "mongo/base/disallow_copying.h"
 #include "mongo/bson/bsonobj.h"
 #include "mongo/bson/simple_bsonobj_comparator.h"
 #include "mongo/db/s/balancer/cluster_statistics.h"
 #include "mongo/s/catalog/type_chunk.h"
-#include "mongo/s/client/shard.h"
+#include "mongo/s/shard_id.h"
 
 namespace mongo {
 
@@ -179,10 +182,15 @@ public:
      *
      * The shouldAggressivelyBalance parameter causes the threshold for chunk could disparity
      * between shards to be lowered.
+     *
+     * The usedShards parameter is in/out and it contains the set of shards, which have already been
+     * used for migrations. Used so we don't return multiple conflicting migrations for the same
+     * shard.
      */
     static std::vector<MigrateInfo> balance(const ShardStatisticsVector& shardStats,
                                             const DistributionStatus& distribution,
-                                            bool shouldAggressivelyBalance);
+                                            bool shouldAggressivelyBalance,
+                                            std::set<ShardId>* usedShards);
 
     /**
      * Using the specified distribution information, returns a suggested better location for the
diff --git a/src/mongo/db/s/balancer/balancer_policy_test.cpp b/src/mongo/db/s/balancer/balancer_policy_test.cpp
index de5ceafaa83..f6abd7c8acf 100644
--- a/src/mongo/db/s/balancer/balancer_policy_test.cpp
+++ b/src/mongo/db/s/balancer/balancer_policy_test.cpp
@@ -110,14 +110,22 @@ std::pair<ShardStatisticsVector, ShardToChunksMap> generateCluster(
     return std::make_pair(std::move(shardStats), std::move(chunkMap));
 }
 
+std::vector<MigrateInfo> balanceChunks(const ShardStatisticsVector& shardStats,
+                                       const DistributionStatus& distribution,
+                                       bool shouldAggressivelyBalance) {
+    std::set<ShardId> usedShards;
+    return BalancerPolicy::balance(
+        shardStats, distribution, shouldAggressivelyBalance, &usedShards);
+}
+
 TEST(BalancerPolicy, Basic) {
     auto cluster = generateCluster(
         {{ShardStatistics(kShardId0, kNoMaxSize, 4, false, emptyTagSet, emptyShardVersion), 4},
          {ShardStatistics(kShardId1, kNoMaxSize, 0, false, emptyTagSet, emptyShardVersion), 0},
          {ShardStatistics(kShardId2, kNoMaxSize, 3, false, emptyTagSet, emptyShardVersion), 3}});
 
-    const auto migrations(BalancerPolicy::balance(
-        cluster.first, DistributionStatus(kNamespace, cluster.second), false));
+    const auto migrations(
+        balanceChunks(cluster.first, DistributionStatus(kNamespace, cluster.second), false));
     ASSERT_EQ(1U, migrations.size());
     ASSERT_EQ(kShardId0, migrations[0].from);
     ASSERT_EQ(kShardId1, migrations[0].to);
@@ -131,8 +139,8 @@ TEST(BalancerPolicy, SmallClusterShouldBePerfectlyBalanced) {
          {ShardStatistics(kShardId1, kNoMaxSize, 2, false, emptyTagSet, emptyShardVersion), 2},
          {ShardStatistics(kShardId2, kNoMaxSize, 0, false, emptyTagSet, emptyShardVersion), 0}});
 
-    const auto migrations(BalancerPolicy::balance(
-        cluster.first, DistributionStatus(kNamespace, cluster.second), false));
+    const auto migrations(
+        balanceChunks(cluster.first, DistributionStatus(kNamespace, cluster.second), false));
     ASSERT_EQ(1U, migrations.size());
     ASSERT_EQ(kShardId1, migrations[0].from);
     ASSERT_EQ(kShardId2, migrations[0].to);
@@ -146,10 +154,8 @@ TEST(BalancerPolicy, SingleChunkShouldNotMove) {
          {ShardStatistics(kShardId1, kNoMaxSize, 0, false, emptyTagSet, emptyShardVersion), 0}});
 
     ASSERT(
-        BalancerPolicy::balance(cluster.first, DistributionStatus(kNamespace, cluster.second), true)
-            .empty());
-    ASSERT(BalancerPolicy::balance(
-               cluster.first, DistributionStatus(kNamespace, cluster.second), false)
+        balanceChunks(cluster.first, DistributionStatus(kNamespace, cluster.second), true).empty());
+    ASSERT(balanceChunks(cluster.first, DistributionStatus(kNamespace, cluster.second), false)
                .empty());
 }
 
@@ -161,10 +167,8 @@ TEST(BalancerPolicy, BalanceThresholdObeyed) {
          {ShardStatistics(kShardId3, kNoMaxSize, 1, false, emptyTagSet, emptyShardVersion), 1}});
 
     ASSERT(
-        BalancerPolicy::balance(cluster.first, DistributionStatus(kNamespace, cluster.second), true)
-            .empty());
-    ASSERT(BalancerPolicy::balance(
-               cluster.first, DistributionStatus(kNamespace, cluster.second), false)
+        balanceChunks(cluster.first, DistributionStatus(kNamespace, cluster.second), true).empty());
+    ASSERT(balanceChunks(cluster.first, DistributionStatus(kNamespace, cluster.second), false)
                .empty());
 }
 
@@ -175,8 +179,8 @@ TEST(BalancerPolicy, ParallelBalancing) {
          {ShardStatistics(kShardId2, kNoMaxSize, 0, false, emptyTagSet, emptyShardVersion), 0},
          {ShardStatistics(kShardId3, kNoMaxSize, 0, false, emptyTagSet, emptyShardVersion), 0}});
 
-    const auto migrations(BalancerPolicy::balance(
-        cluster.first, DistributionStatus(kNamespace, cluster.second), false));
+    const auto migrations(
+        balanceChunks(cluster.first, DistributionStatus(kNamespace, cluster.second), false));
     ASSERT_EQ(2U, migrations.size());
 
     ASSERT_EQ(kShardId0, migrations[0].from);
@@ -199,8 +203,8 @@ TEST(BalancerPolicy, ParallelBalancingDoesNotPutChunksOnShardsAboveTheOptimal) {
          {ShardStatistics(kShardId4, kNoMaxSize, 0, false, emptyTagSet, emptyShardVersion), 0},
          {ShardStatistics(kShardId5, kNoMaxSize, 0, false, emptyTagSet, emptyShardVersion), 0}});
 
-    const auto migrations(BalancerPolicy::balance(
-        cluster.first, DistributionStatus(kNamespace, cluster.second), false));
+    const auto migrations(
+        balanceChunks(cluster.first, DistributionStatus(kNamespace, cluster.second), false));
     ASSERT_EQ(2U, migrations.size());
 
     ASSERT_EQ(kShardId0, migrations[0].from);
@@ -221,8 +225,60 @@ TEST(BalancerPolicy, ParallelBalancingDoesNotMoveChunksFromShardsBelowOptimal) {
          {ShardStatistics(kShardId2, kNoMaxSize, 5, false, emptyTagSet, emptyShardVersion), 5},
          {ShardStatistics(kShardId3, kNoMaxSize, 0, false, emptyTagSet, emptyShardVersion), 0}});
 
+    const auto migrations(
+        balanceChunks(cluster.first, DistributionStatus(kNamespace, cluster.second), false));
+    ASSERT_EQ(1U, migrations.size());
+
+    ASSERT_EQ(kShardId0, migrations[0].from);
+    ASSERT_EQ(kShardId3, migrations[0].to);
+    ASSERT_BSONOBJ_EQ(cluster.second[kShardId0][0].getMin(), migrations[0].minKey);
+    ASSERT_BSONOBJ_EQ(cluster.second[kShardId0][0].getMax(), migrations[0].maxKey);
+}
+
+TEST(BalancerPolicy, ParallelBalancingNotSchedulingOnInUseSourceShardsWithMoveNecessary) {
+    auto cluster = generateCluster(
+        {{ShardStatistics(kShardId0, kNoMaxSize, 8, false, emptyTagSet, emptyShardVersion), 8},
+         {ShardStatistics(kShardId1, kNoMaxSize, 4, false, emptyTagSet, emptyShardVersion), 4},
+         {ShardStatistics(kShardId2, kNoMaxSize, 0, false, emptyTagSet, emptyShardVersion), 0},
+         {ShardStatistics(kShardId3, kNoMaxSize, 0, false, emptyTagSet, emptyShardVersion), 0}});
+
+    // Here kShardId0 would have been selected as a donor
+    std::set<ShardId> usedShards{kShardId0};
+    const auto migrations(BalancerPolicy::balance(
+        cluster.first, DistributionStatus(kNamespace, cluster.second), false, &usedShards));
+    ASSERT_EQ(1U, migrations.size());
+
+    ASSERT_EQ(kShardId1, migrations[0].from);
+    ASSERT_EQ(kShardId2, migrations[0].to);
+    ASSERT_BSONOBJ_EQ(cluster.second[kShardId1][0].getMin(), migrations[0].minKey);
+    ASSERT_BSONOBJ_EQ(cluster.second[kShardId1][0].getMax(), migrations[0].maxKey);
+}
+
+TEST(BalancerPolicy, ParallelBalancingNotSchedulingOnInUseSourceShardsWithMoveNotNecessary) {
+    auto cluster = generateCluster(
+        {{ShardStatistics(kShardId0, kNoMaxSize, 12, false, emptyTagSet, emptyShardVersion), 12},
+         {ShardStatistics(kShardId1, kNoMaxSize, 4, false, emptyTagSet, emptyShardVersion), 4},
+         {ShardStatistics(kShardId2, kNoMaxSize, 0, false, emptyTagSet, emptyShardVersion), 0},
+         {ShardStatistics(kShardId3, kNoMaxSize, 0, false, emptyTagSet, emptyShardVersion), 0}});
+
+    // Here kShardId0 would have been selected as a donor
+    std::set<ShardId> usedShards{kShardId0};
+    const auto migrations(BalancerPolicy::balance(
+        cluster.first, DistributionStatus(kNamespace, cluster.second), false, &usedShards));
+    ASSERT_EQ(0U, migrations.size());
+}
+
+TEST(BalancerPolicy, ParallelBalancingNotSchedulingOnInUseDestinationShards) {
+    auto cluster = generateCluster(
+        {{ShardStatistics(kShardId0, kNoMaxSize, 4, false, emptyTagSet, emptyShardVersion), 4},
+         {ShardStatistics(kShardId1, kNoMaxSize, 4, false, emptyTagSet, emptyShardVersion), 4},
+         {ShardStatistics(kShardId2, kNoMaxSize, 0, false, emptyTagSet, emptyShardVersion), 0},
+         {ShardStatistics(kShardId3, kNoMaxSize, 1, false, emptyTagSet, emptyShardVersion), 1}});
+
+    // Here kShardId2 would have been selected as a recipient
+    std::set<ShardId> usedShards{kShardId2};
     const auto migrations(BalancerPolicy::balance(
-        cluster.first, DistributionStatus(kNamespace, cluster.second), false));
+        cluster.first, DistributionStatus(kNamespace, cluster.second), false, &usedShards));
     ASSERT_EQ(1U, migrations.size());
 
     ASSERT_EQ(kShardId0, migrations[0].from);
@@ -241,8 +297,8 @@ TEST(BalancerPolicy, JumboChunksNotMoved) {
     cluster.second[kShardId0][2].setJumbo(true);
     cluster.second[kShardId0][3].setJumbo(true);
 
-    const auto migrations(BalancerPolicy::balance(
-        cluster.first, DistributionStatus(kNamespace, cluster.second), false));
+    const auto migrations(
+        balanceChunks(cluster.first, DistributionStatus(kNamespace, cluster.second), false));
     ASSERT_EQ(1U, migrations.size());
     ASSERT_EQ(kShardId0, migrations[0].from);
     ASSERT_EQ(kShardId1, migrations[0].to);
@@ -267,8 +323,8 @@ TEST(BalancerPolicy, JumboChunksNotMovedParallel) {
     cluster.second[kShardId2][2].setJumbo(false);  // Only chunk 1 is not jumbo
     cluster.second[kShardId2][3].setJumbo(true);
 
-    const auto migrations(BalancerPolicy::balance(
-        cluster.first, DistributionStatus(kNamespace, cluster.second), false));
+    const auto migrations(
+        balanceChunks(cluster.first, DistributionStatus(kNamespace, cluster.second), false));
     ASSERT_EQ(2U, migrations.size());
 
     ASSERT_EQ(kShardId0, migrations[0].from);
@@ -288,8 +344,8 @@ TEST(BalancerPolicy, DrainingSingleChunk) {
         {{ShardStatistics(kShardId0, kNoMaxSize, 2, true, emptyTagSet, emptyShardVersion), 1},
          {ShardStatistics(kShardId1, kNoMaxSize, 0, false, emptyTagSet, emptyShardVersion), 5}});
 
-    const auto migrations(BalancerPolicy::balance(
-        cluster.first, DistributionStatus(kNamespace, cluster.second), false));
+    const auto migrations(
+        balanceChunks(cluster.first, DistributionStatus(kNamespace, cluster.second), false));
     ASSERT_EQ(1U, migrations.size());
     ASSERT_EQ(kShardId0, migrations[0].from);
     ASSERT_EQ(kShardId1, migrations[0].to);
@@ -305,8 +361,8 @@ TEST(BalancerPolicy, DrainingSingleChunkPerShard) {
          {ShardStatistics(kShardId2, kNoMaxSize, 2, true, emptyTagSet, emptyShardVersion), 1},
          {ShardStatistics(kShardId3, kNoMaxSize, 0, false, emptyTagSet, emptyShardVersion), 5}});
 
-    const auto migrations(BalancerPolicy::balance(
-        cluster.first, DistributionStatus(kNamespace, cluster.second), false));
+    const auto migrations(
+        balanceChunks(cluster.first, DistributionStatus(kNamespace, cluster.second), false));
     ASSERT_EQ(2U, migrations.size());
 
     ASSERT_EQ(kShardId0, migrations[0].from);
@@ -326,8 +382,8 @@ TEST(BalancerPolicy, DrainingWithTwoChunksFirstOneSelected) {
         {{ShardStatistics(kShardId0, kNoMaxSize, 2, true, emptyTagSet, emptyShardVersion), 2},
          {ShardStatistics(kShardId1, kNoMaxSize, 0, false, emptyTagSet, emptyShardVersion), 5}});
 
-    const auto migrations(BalancerPolicy::balance(
-        cluster.first, DistributionStatus(kNamespace, cluster.second), false));
+    const auto migrations(
+        balanceChunks(cluster.first, DistributionStatus(kNamespace, cluster.second), false));
     ASSERT_EQ(1U, migrations.size());
     ASSERT_EQ(kShardId0, migrations[0].from);
     ASSERT_EQ(kShardId1, migrations[0].to);
@@ -343,8 +399,8 @@ TEST(BalancerPolicy, DrainingMultipleShardsFirstOneSelected) {
          {ShardStatistics(kShardId1, kNoMaxSize, 5, true, emptyTagSet, emptyShardVersion), 2},
          {ShardStatistics(kShardId2, kNoMaxSize, 5, false, emptyTagSet, emptyShardVersion), 16}});
 
-    const auto migrations(BalancerPolicy::balance(
-        cluster.first, DistributionStatus(kNamespace, cluster.second), false));
+    const auto migrations(
+        balanceChunks(cluster.first, DistributionStatus(kNamespace, cluster.second), false));
     ASSERT_EQ(1U, migrations.size());
     ASSERT_EQ(kShardId0, migrations[0].from);
     ASSERT_EQ(kShardId2, migrations[0].to);
@@ -359,8 +415,8 @@ TEST(BalancerPolicy, DrainingMultipleShardsWontAcceptChunks) {
          {ShardStatistics(kShardId1, kNoMaxSize, 0, true, emptyTagSet, emptyShardVersion), 0},
          {ShardStatistics(kShardId2, kNoMaxSize, 0, true, emptyTagSet, emptyShardVersion), 0}});
 
-    const auto migrations(BalancerPolicy::balance(
-        cluster.first, DistributionStatus(kNamespace, cluster.second), false));
+    const auto migrations(
+        balanceChunks(cluster.first, DistributionStatus(kNamespace, cluster.second), false));
     ASSERT(migrations.empty());
 }
 
@@ -374,7 +430,7 @@ TEST(BalancerPolicy, DrainingSingleAppropriateShardFoundDueToTag) {
     ASSERT_OK(distribution.addRangeToZone(ZoneRange(
         cluster.second[kShardId2][0].getMin(), cluster.second[kShardId2][0].getMax(), "LAX")));
 
-    const auto migrations(BalancerPolicy::balance(cluster.first, distribution, false));
+    const auto migrations(balanceChunks(cluster.first, distribution, false));
     ASSERT_EQ(1U, migrations.size());
     ASSERT_EQ(kShardId2, migrations[0].from);
     ASSERT_EQ(kShardId1, migrations[0].to);
@@ -392,7 +448,7 @@ TEST(BalancerPolicy, DrainingNoAppropriateShardsFoundDueToTag) {
     ASSERT_OK(distribution.addRangeToZone(ZoneRange(
         cluster.second[kShardId2][0].getMin(), cluster.second[kShardId2][0].getMax(), "SEA")));
 
-    const auto migrations(BalancerPolicy::balance(cluster.first, distribution, false));
+    const auto migrations(balanceChunks(cluster.first, distribution, false));
     ASSERT(migrations.empty());
 }
 
@@ -403,8 +459,8 @@ TEST(BalancerPolicy, NoBalancingDueToAllNodesEitherDrainingOrMaxedOut) {
          {ShardStatistics(kShardId1, 1, 1, false, emptyTagSet, emptyShardVersion), 6},
          {ShardStatistics(kShardId2, kNoMaxSize, 1, true, emptyTagSet, emptyShardVersion), 1}});
 
-    const auto migrations(BalancerPolicy::balance(
-        cluster.first, DistributionStatus(kNamespace, cluster.second), false));
+    const auto migrations(
+        balanceChunks(cluster.first, DistributionStatus(kNamespace, cluster.second), false));
     ASSERT(migrations.empty());
 }
 
@@ -417,8 +473,8 @@ TEST(BalancerPolicy, BalancerRespectsMaxShardSizeOnlyBalanceToNonMaxed) {
          {ShardStatistics(kShardId1, kNoMaxSize, 5, false, emptyTagSet, emptyShardVersion), 5},
          {ShardStatistics(kShardId2, kNoMaxSize, 10, false, emptyTagSet, emptyShardVersion), 10}});
 
-    const auto migrations(BalancerPolicy::balance(
-        cluster.first, DistributionStatus(kNamespace, cluster.second), false));
+    const auto migrations(
+        balanceChunks(cluster.first, DistributionStatus(kNamespace, cluster.second), false));
     ASSERT_EQ(1U, migrations.size());
     ASSERT_EQ(kShardId2, migrations[0].from);
     ASSERT_EQ(kShardId1, migrations[0].to);
@@ -435,8 +491,8 @@ TEST(BalancerPolicy, BalancerRespectsMaxShardSizeWhenAllBalanced) {
          {ShardStatistics(kShardId1, kNoMaxSize, 4, false, emptyTagSet, emptyShardVersion), 4},
          {ShardStatistics(kShardId2, kNoMaxSize, 4, false, emptyTagSet, emptyShardVersion), 4}});
 
-    const auto migrations(BalancerPolicy::balance(
-        cluster.first, DistributionStatus(kNamespace, cluster.second), false));
+    const auto migrations(
+        balanceChunks(cluster.first, DistributionStatus(kNamespace, cluster.second), false));
     ASSERT(migrations.empty());
 }
 
@@ -451,7 +507,7 @@ TEST(BalancerPolicy, BalancerRespectsTagsWhenDraining) {
     ASSERT_OK(distribution.addRangeToZone(ZoneRange(kMinBSONKey, BSON("x" << 7), "a")));
     ASSERT_OK(distribution.addRangeToZone(ZoneRange(BSON("x" << 8), kMaxBSONKey, "b")));
 
-    const auto migrations(BalancerPolicy::balance(cluster.first, distribution, false));
+    const auto migrations(balanceChunks(cluster.first, distribution, false));
     ASSERT_EQ(1U, migrations.size());
     ASSERT_EQ(kShardId1, migrations[0].from);
     ASSERT_EQ(kShardId0, migrations[0].to);
@@ -470,7 +526,7 @@ TEST(BalancerPolicy, BalancerRespectsTagPolicyBeforeImbalance) {
     DistributionStatus distribution(kNamespace, cluster.second);
     ASSERT_OK(distribution.addRangeToZone(ZoneRange(kMinBSONKey, BSON("x" << 100), "a")));
 
-    const auto migrations(BalancerPolicy::balance(cluster.first, distribution, false));
+    const auto migrations(balanceChunks(cluster.first, distribution, false));
     ASSERT_EQ(1U, migrations.size());
     ASSERT_EQ(kShardId2, migrations[0].from);
     ASSERT_EQ(kShardId0, migrations[0].to);
@@ -490,7 +546,7 @@ TEST(BalancerPolicy, BalancerFixesIncorrectTagsWithCrossShardViolationOfTags) {
     ASSERT_OK(distribution.addRangeToZone(ZoneRange(kMinBSONKey, BSON("x" << 1), "b")));
     ASSERT_OK(distribution.addRangeToZone(ZoneRange(BSON("x" << 8), kMaxBSONKey, "a")));
 
-    const auto migrations(BalancerPolicy::balance(cluster.first, distribution, false));
+    const auto migrations(balanceChunks(cluster.first, distribution, false));
     ASSERT_EQ(1U, migrations.size());
     ASSERT_EQ(kShardId0, migrations[0].from);
     ASSERT_EQ(kShardId2, migrations[0].to);
@@ -508,7 +564,7 @@ TEST(BalancerPolicy, BalancerFixesIncorrectTagsInOtherwiseBalancedCluster) {
     DistributionStatus distribution(kNamespace, cluster.second);
     ASSERT_OK(distribution.addRangeToZone(ZoneRange(kMinBSONKey, BSON("x" << 10), "a")));
 
-    const auto migrations(BalancerPolicy::balance(cluster.first, distribution, false));
+    const auto migrations(balanceChunks(cluster.first, distribution, false));
     ASSERT_EQ(1U, migrations.size());
     ASSERT_EQ(kShardId2, migrations[0].from);
     ASSERT_EQ(kShardId0, migrations[0].to);
@@ -527,7 +583,7 @@ TEST(BalancerPolicy, BalancerFixesIncorrectTagsInOtherwiseBalancedClusterParalle
     DistributionStatus distribution(kNamespace, cluster.second);
     ASSERT_OK(distribution.addRangeToZone(ZoneRange(kMinBSONKey, BSON("x" << 20), "a")));
 
-    const auto migrations(BalancerPolicy::balance(cluster.first, distribution, false));
+    const auto migrations(balanceChunks(cluster.first, distribution, false));
     ASSERT_EQ(2U, migrations.size());
 
     ASSERT_EQ(kShardId2, migrations[0].from);
@@ -550,7 +606,7 @@ TEST(BalancerPolicy, BalancerHandlesNoShardsWithTag) {
     ASSERT_OK(
         distribution.addRangeToZone(ZoneRange(kMinBSONKey, BSON("x" << 7), "NonExistentZone")));
 
-    ASSERT(BalancerPolicy::balance(cluster.first, distribution, false).empty());
+    ASSERT(balanceChunks(cluster.first, distribution, false).empty());
 }
 
 TEST(DistributionStatus, AddTagRangeOverlap) {
diff --git a/src/mongo/db/s/collection_range_deleter.cpp b/src/mongo/db/s/collection_range_deleter.cpp
deleted file mode 100644
index 803541d16f9..00000000000
--- a/src/mongo/db/s/collection_range_deleter.cpp
+++ /dev/null
@@ -1,222 +0,0 @@
-/**
- *    Copyright (C) 2016 MongoDB Inc.
- *
- *    This program is free software: you can redistribute it and/or  modify
- *    it under the terms of the GNU Affero General Public License, version 3,
- *    as published by the Free Software Foundation.
- *
- *    This program is distributed in the hope that it will be useful,
- *    but WITHOUT ANY WARRANTY; without even the implied warranty of
- *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
- *    GNU Affero General Public License for more details.
- *
- *    You should have received a copy of the GNU Affero General Public License
- *    along with this program.  If not, see <http://www.gnu.org/licenses/>.
- *
- *    As a special exception, the copyright holders give permission to link the
- *    code of portions of this program with the OpenSSL library under certain
- *    conditions as described in each individual source file and distribute
- *    linked combinations including the program with the OpenSSL library. You
- *    must comply with the GNU Affero General Public License in all respects for
- *    all of the code used other than as permitted herein. If you modify file(s)
- *    with this exception, you may extend this exception to your version of the
- *    file(s), but you are not obligated to do so. If you do not wish to do so,
- *    delete this exception statement from your version. If you delete this
- *    exception statement from all source files in the program, then also delete
- *    it in the license file.
- */
-
-#define MONGO_LOG_DEFAULT_COMPONENT ::mongo::logger::LogComponent::kSharding
-
-#include "mongo/platform/basic.h"
-
-#include "mongo/db/s/collection_range_deleter.h"
-
-#include <algorithm>
-
-#include "mongo/db/catalog/collection.h"
-#include "mongo/db/client.h"
-#include "mongo/db/db_raii.h"
-#include "mongo/db/dbhelpers.h"
-#include "mongo/db/exec/working_set_common.h"
-#include "mongo/db/index/index_descriptor.h"
-#include "mongo/db/keypattern.h"
-#include "mongo/db/query/internal_plans.h"
-#include "mongo/db/query/query_knobs.h"
-#include "mongo/db/query/query_planner.h"
-#include "mongo/db/repl/repl_client_info.h"
-#include "mongo/db/repl/replication_coordinator_global.h"
-#include "mongo/db/s/collection_sharding_state.h"
-#include "mongo/db/s/sharding_state.h"
-#include "mongo/db/write_concern.h"
-#include "mongo/executor/task_executor.h"
-#include "mongo/util/log.h"
-#include "mongo/util/mongoutils/str.h"
-#include "mongo/util/scopeguard.h"
-
-namespace mongo {
-
-class ChunkRange;
-class OldClientWriteContext;
-
-using CallbackArgs = executor::TaskExecutor::CallbackArgs;
-using logger::LogComponent;
-
-namespace {
-
-const WriteConcernOptions kMajorityWriteConcern(WriteConcernOptions::kMajority,
-                                                WriteConcernOptions::SyncMode::UNSET,
-                                                Seconds(60));
-
-}  // unnamed namespace
-
-CollectionRangeDeleter::CollectionRangeDeleter(NamespaceString nss) : _nss(std::move(nss)) {}
-
-void CollectionRangeDeleter::run() {
-    Client::initThread(getThreadName());
-    ON_BLOCK_EXIT([&] { Client::destroy(); });
-    auto txn = cc().makeOperationContext().get();
-    bool hasNextRangeToClean = cleanupNextRange(txn);
-
-    // If there are more ranges to run, we add <this> back onto the task executor to run again.
-    if (hasNextRangeToClean) {
-        auto executor = ShardingState::get(txn)->getRangeDeleterTaskExecutor();
-        executor->scheduleWork([this](const CallbackArgs& cbArgs) { run(); });
-    } else {
-        delete this;
-    }
-}
-
-bool CollectionRangeDeleter::cleanupNextRange(OperationContext* txn) {
-    int numDocumentsDeleted;
-
-    {
-        AutoGetCollection autoColl(txn, _nss, MODE_IX);
-        Collection* collection = autoColl.getCollection();
-        if (!collection) {
-            return false;
-        }
-
-        CollectionShardingState* shardingState = CollectionShardingState::get(txn, _nss);
-        MetadataManager& metadataManager = *shardingState->_metadataManager;
-
-        if (!_rangeInProgress && !metadataManager.hasRangesToClean()) {
-            // Nothing left to do
-            return false;
-        }
-
-        if (!_rangeInProgress || !metadataManager.isInRangesToClean(_rangeInProgress.get())) {
-            // No valid chunk in progress, get a new one
-            _rangeInProgress = metadataManager.getNextRangeToClean();
-        }
-
-        auto metadata = shardingState->getMetadata();
-        if (!metadata) {
-            return false;
-        }
-
-        numDocumentsDeleted = _doDeletion(txn, collection, metadata->getKeyPattern());
-        if (numDocumentsDeleted <= 0) {
-            metadataManager.removeRangeToClean(_rangeInProgress.get());
-            _rangeInProgress = boost::none;
-            return true;
-        }
-    }
-
-    // wait for replication
-    WriteConcernResult wcResult;
-    auto currentClientOpTime = repl::ReplClientInfo::forClient(txn->getClient()).getLastOp();
-    Status status = waitForWriteConcern(txn, currentClientOpTime, kMajorityWriteConcern, &wcResult);
-    if (!status.isOK()) {
-        warning() << "Error when waiting for write concern after removing chunks in " << _nss
-                  << " : " << status.reason();
-    }
-
-    return true;
-}
-
-int CollectionRangeDeleter::_doDeletion(OperationContext* txn,
-                                        Collection* collection,
-                                        const BSONObj& keyPattern) {
-    invariant(_rangeInProgress);
-    invariant(collection);
-
-    // The IndexChunk has a keyPattern that may apply to more than one index - we need to
-    // select the index and get the full index keyPattern here.
-    const IndexDescriptor* idx =
-        collection->getIndexCatalog()->findShardKeyPrefixedIndex(txn, keyPattern, false);
-    if (idx == NULL) {
-        warning() << "Unable to find shard key index for " << keyPattern.toString() << " in "
-                  << _nss;
-        return -1;
-    }
-
-    KeyPattern indexKeyPattern(idx->keyPattern().getOwned());
-
-    // Extend bounds to match the index we found
-    const BSONObj& min =
-        Helpers::toKeyFormat(indexKeyPattern.extendRangeBound(_rangeInProgress->getMin(), false));
-    const BSONObj& max =
-        Helpers::toKeyFormat(indexKeyPattern.extendRangeBound(_rangeInProgress->getMax(), false));
-
-    LOG(1) << "begin removal of " << min << " to " << max << " in " << _nss;
-
-    auto indexName = idx->indexName();
-    IndexDescriptor* desc = collection->getIndexCatalog()->findIndexByName(txn, indexName);
-    if (!desc) {
-        warning() << "shard key index with name " << indexName << " on '" << _nss
-                  << "' was dropped";
-        return -1;
-    }
-
-    std::unique_ptr<PlanExecutor> exec(
-        InternalPlanner::indexScan(txn,
-                                   collection,
-                                   desc,
-                                   min,
-                                   max,
-                                   BoundInclusion::kIncludeStartKeyOnly,
-                                   PlanExecutor::YIELD_MANUAL,
-                                   InternalPlanner::FORWARD,
-                                   InternalPlanner::IXSCAN_FETCH));
-    int numDeleted = 0;
-    const int maxItersBeforeYield = std::max(static_cast<int>(internalQueryExecYieldIterations), 1);
-
-    while (numDeleted < maxItersBeforeYield) {
-        RecordId rloc;
-        BSONObj obj;
-        PlanExecutor::ExecState state;
-        state = exec->getNext(&obj, &rloc);
-        if (PlanExecutor::IS_EOF == state) {
-            break;
-        }
-
-        if (PlanExecutor::FAILURE == state || PlanExecutor::DEAD == state) {
-            warning(LogComponent::kSharding)
-                << PlanExecutor::statestr(state) << " - cursor error while trying to delete " << min
-                << " to " << max << " in " << _nss << ": " << WorkingSetCommon::toStatusString(obj)
-                << ", stats: " << Explain::getWinningPlanStats(exec.get());
-            break;
-        }
-
-        invariant(PlanExecutor::ADVANCED == state);
-
-        WriteUnitOfWork wuow(txn);
-
-        NamespaceString nss(_nss);
-        if (!repl::getGlobalReplicationCoordinator()->canAcceptWritesFor(nss)) {
-            warning() << "stepped down from primary while deleting chunk; "
-                      << "orphaning data in " << _nss << " in range [" << min << ", " << max << ")";
-            return numDeleted;
-        }
-
-        OpDebug* const nullOpDebug = nullptr;
-        collection->deleteDocument(txn, rloc, nullOpDebug, true);
-        wuow.commit();
-        numDeleted++;
-    }
-
-    return numDeleted;
-}
-
-}  // namespace mongo
diff --git a/src/mongo/db/s/collection_sharding_state.h b/src/mongo/db/s/collection_sharding_state.h
index 2a01bec5567..1abababcc1b 100644
--- a/src/mongo/db/s/collection_sharding_state.h
+++ b/src/mongo/db/s/collection_sharding_state.h
@@ -168,8 +168,6 @@ public:
     void onDropCollection(OperationContext* txn, const NamespaceString& collectionName);
 
 private:
-    friend class CollectionRangeDeleter;
-
     /**
      * Checks whether the shard version of the operation matches that of the collection.
      *
diff --git a/src/mongo/db/s/metadata_manager.cpp b/src/mongo/db/s/metadata_manager.cpp
index 28ca20e2f6f..0bba1ff478e 100644
--- a/src/mongo/db/s/metadata_manager.cpp
+++ b/src/mongo/db/s/metadata_manager.cpp
@@ -34,7 +34,6 @@
 
 #include "mongo/bson/simple_bsonobj_comparator.h"
 #include "mongo/db/range_arithmetic.h"
-#include "mongo/db/s/collection_range_deleter.h"
 #include "mongo/db/s/sharding_state.h"
 #include "mongo/stdx/memory.h"
 #include "mongo/util/log.h"
@@ -47,9 +46,8 @@ MetadataManager::MetadataManager(ServiceContext* sc, NamespaceString nss)
     : _nss(std::move(nss)),
       _serviceContext(sc),
       _activeMetadataTracker(stdx::make_unique<CollectionMetadataTracker>(nullptr)),
-      _receivingChunks(SimpleBSONObjComparator::kInstance.makeBSONObjIndexedMap<CachedChunkInfo>()),
-      _rangesToClean(
-          SimpleBSONObjComparator::kInstance.makeBSONObjIndexedMap<RangeToCleanDescriptor>()) {}
+      _receivingChunks(
+          SimpleBSONObjComparator::kInstance.makeBSONObjIndexedMap<CachedChunkInfo>()) {}
 
 MetadataManager::~MetadataManager() {
     stdx::lock_guard<stdx::mutex> scopedLock(_managerLock);
@@ -74,7 +72,6 @@ void MetadataManager::refreshActiveMetadata(std::unique_ptr<CollectionMetadata>
     // collection sharding information regardless of whether the node is sharded or not.
     if (!remoteMetadata && !_activeMetadataTracker->metadata) {
         invariant(_receivingChunks.empty());
-        invariant(_rangesToClean.empty());
         return;
     }
 
@@ -84,8 +81,6 @@ void MetadataManager::refreshActiveMetadata(std::unique_ptr<CollectionMetadata>
               << _activeMetadataTracker->metadata->toStringBasic() << " as no longer sharded";
 
         _receivingChunks.clear();
-        _rangesToClean.clear();
-
         _setActiveMetadata_inlock(nullptr);
         return;
     }
@@ -100,8 +95,6 @@ void MetadataManager::refreshActiveMetadata(std::unique_ptr<CollectionMetadata>
               << remoteMetadata->toStringBasic();
 
         invariant(_receivingChunks.empty());
-        invariant(_rangesToClean.empty());
-
         _setActiveMetadata_inlock(std::move(remoteMetadata));
         return;
     }
@@ -115,8 +108,6 @@ void MetadataManager::refreshActiveMetadata(std::unique_ptr<CollectionMetadata>
               << remoteMetadata->toStringBasic() << " due to epoch change";
 
         _receivingChunks.clear();
-        _rangesToClean.clear();
-
         _setActiveMetadata_inlock(std::move(remoteMetadata));
         return;
     }
@@ -168,9 +159,6 @@ void MetadataManager::refreshActiveMetadata(std::unique_ptr<CollectionMetadata>
             const ChunkRange receivingRange(itRecv->first, itRecv->second.getMaxKey());
 
             _receivingChunks.erase(itRecv);
-
-            // Make sure any potentially partially copied chunks are scheduled to be cleaned up
-            _addRangeToClean_inlock(receivingRange);
         }
 
         // Need to reset the iterator
@@ -207,14 +195,10 @@ void MetadataManager::beginReceive(const ChunkRange& range) {
         const ChunkRange receivingRange(itRecv->first, itRecv->second.getMaxKey());
 
         _receivingChunks.erase(itRecv);
-
-        // Make sure any potentially partially copied chunks are scheduled to be cleaned up
-        _addRangeToClean_inlock(receivingRange);
     }
 
     // Need to ensure that the background range deleter task won't delete the range we are about to
     // receive
-    _removeRangeToClean_inlock(range, Status::OK());
     _receivingChunks.insert(
         std::make_pair(range.getMin().getOwned(),
                        CachedChunkInfo(range.getMax().getOwned(), ChunkVersion::IGNORED())));
@@ -241,9 +225,6 @@ void MetadataManager::forgetReceive(const ChunkRange& range) {
         _receivingChunks.erase(it);
     }
 
-    // This is potentially a partially received data, which needs to be cleaned up
-    _addRangeToClean_inlock(range);
-
     // For compatibility with the current range deleter, update the pending chunks on the collection
     // metadata to exclude the chunk being received, which was added in beginReceive
     ChunkType chunk;
@@ -343,91 +324,9 @@ ScopedCollectionMetadata::operator bool() const {
     return _tracker && _tracker->metadata.get();
 }
 
-RangeMap MetadataManager::getCopyOfRangesToClean() {
-    stdx::lock_guard<stdx::mutex> scopedLock(_managerLock);
-    return _getCopyOfRangesToClean_inlock();
-}
-
-RangeMap MetadataManager::_getCopyOfRangesToClean_inlock() {
-    RangeMap ranges = SimpleBSONObjComparator::kInstance.makeBSONObjIndexedMap<CachedChunkInfo>();
-    for (auto it = _rangesToClean.begin(); it != _rangesToClean.end(); ++it) {
-        ranges.insert(std::make_pair(
-            it->first, CachedChunkInfo(it->second.getMax(), ChunkVersion::IGNORED())));
-    }
-    return ranges;
-}
-
-std::shared_ptr<Notification<Status>> MetadataManager::addRangeToClean(const ChunkRange& range) {
-    stdx::lock_guard<stdx::mutex> scopedLock(_managerLock);
-    return _addRangeToClean_inlock(range);
-}
-
-std::shared_ptr<Notification<Status>> MetadataManager::_addRangeToClean_inlock(
-    const ChunkRange& range) {
-    // This first invariant currently makes an unnecessary copy, to reuse the
-    // rangeMapOverlaps helper function.
-    invariant(!rangeMapOverlaps(_getCopyOfRangesToClean_inlock(), range.getMin(), range.getMax()));
-    invariant(!rangeMapOverlaps(_receivingChunks, range.getMin(), range.getMax()));
-
-    RangeToCleanDescriptor descriptor(range.getMax().getOwned());
-    _rangesToClean.insert(std::make_pair(range.getMin().getOwned(), descriptor));
-
-    // If _rangesToClean was previously empty, we need to start the collection range deleter
-    if (_rangesToClean.size() == 1UL) {
-        ShardingState::get(_serviceContext)->scheduleCleanup(_nss);
-    }
-
-    return descriptor.getNotification();
-}
-
-void MetadataManager::removeRangeToClean(const ChunkRange& range, Status deletionStatus) {
-    stdx::lock_guard<stdx::mutex> scopedLock(_managerLock);
-    _removeRangeToClean_inlock(range, deletionStatus);
-}
-
-void MetadataManager::_removeRangeToClean_inlock(const ChunkRange& range, Status deletionStatus) {
-    auto it = _rangesToClean.upper_bound(range.getMin());
-    // We want our iterator to point at the greatest value
-    // that is still less than or equal to range.
-    if (it != _rangesToClean.begin()) {
-        --it;
-    }
-
-    for (; it != _rangesToClean.end() &&
-         SimpleBSONObjComparator::kInstance.evaluate(it->first < range.getMax());) {
-        if (SimpleBSONObjComparator::kInstance.evaluate(it->second.getMax() <= range.getMin())) {
-            ++it;
-            continue;
-        }
-
-        // There's overlap between *it and range so we remove *it
-        // and then replace with new ranges.
-        BSONObj oldMin = it->first;
-        BSONObj oldMax = it->second.getMax();
-        it->second.complete(deletionStatus);
-        _rangesToClean.erase(it++);
-        if (SimpleBSONObjComparator::kInstance.evaluate(oldMin < range.getMin())) {
-            _addRangeToClean_inlock(ChunkRange(oldMin, range.getMin()));
-        }
-
-        if (SimpleBSONObjComparator::kInstance.evaluate(oldMax > range.getMax())) {
-            _addRangeToClean_inlock(ChunkRange(range.getMax(), oldMax));
-        }
-    }
-}
-
 void MetadataManager::append(BSONObjBuilder* builder) {
     stdx::lock_guard<stdx::mutex> scopedLock(_managerLock);
 
-    BSONArrayBuilder rtcArr(builder->subarrayStart("rangesToClean"));
-    for (const auto& entry : _rangesToClean) {
-        BSONObjBuilder obj;
-        ChunkRange r = ChunkRange(entry.first, entry.second.getMax());
-        r.append(&obj);
-        rtcArr.append(obj.done());
-    }
-    rtcArr.done();
-
     BSONArrayBuilder pcArr(builder->subarrayStart("pendingChunks"));
     for (const auto& entry : _receivingChunks) {
         BSONObjBuilder obj;
@@ -447,23 +346,4 @@ void MetadataManager::append(BSONObjBuilder* builder) {
     amrArr.done();
 }
 
-bool MetadataManager::hasRangesToClean() {
-    stdx::lock_guard<stdx::mutex> scopedLock(_managerLock);
-    return !_rangesToClean.empty();
-}
-
-bool MetadataManager::isInRangesToClean(const ChunkRange& range) {
-    stdx::lock_guard<stdx::mutex> scopedLock(_managerLock);
-    // For convenience, this line makes an unnecessary copy, to reuse the
-    // rangeMapContains helper function.
-    return rangeMapContains(_getCopyOfRangesToClean_inlock(), range.getMin(), range.getMax());
-}
-
-ChunkRange MetadataManager::getNextRangeToClean() {
-    stdx::lock_guard<stdx::mutex> scopedLock(_managerLock);
-    invariant(!_rangesToClean.empty());
-    auto it = _rangesToClean.begin();
-    return ChunkRange(it->first, it->second.getMax());
-}
-
 }  // namespace mongo
diff --git a/src/mongo/db/s/metadata_manager.h b/src/mongo/db/s/metadata_manager.h
index fa61e9ef2ea..b2e712d7f9f 100644
--- a/src/mongo/db/s/metadata_manager.h
+++ b/src/mongo/db/s/metadata_manager.h
@@ -85,52 +85,10 @@ public:
     RangeMap getCopyOfReceivingChunks();
 
     /**
-    * Adds a new range to be cleaned up.
-    * The newly introduced range must not overlap with the existing ranges.
-    */
-    std::shared_ptr<Notification<Status>> addRangeToClean(const ChunkRange& range);
-
-    /**
-     * Calls removeRangeToClean with Status::OK.
-     */
-    void removeRangeToClean(const ChunkRange& range) {
-        removeRangeToClean(range, Status::OK());
-    }
-
-    /**
-     * Removes the specified range from the ranges to be cleaned up.
-     * The specified deletionStatus will be returned to callers waiting
-     * on whether the deletion succeeded or failed.
-     */
-    void removeRangeToClean(const ChunkRange& range, Status deletionStatus);
-
-    /**
-     * Gets copy of the set of chunk ranges which are scheduled for cleanup.
-     * Converts RangeToCleanMap to RangeMap.
-     */
-    RangeMap getCopyOfRangesToClean();
-
-    /**
      * Appends information on all the chunk ranges in rangesToClean to builder.
      */
     void append(BSONObjBuilder* builder);
 
-    /**
-     * Returns true if _rangesToClean is not empty.
-     */
-    bool hasRangesToClean();
-
-    /**
-     * Returns true if the exact range is in _rangesToClean.
-     */
-    bool isInRangesToClean(const ChunkRange& range);
-
-    /**
-     * Gets and returns, but does not remove, a single ChunkRange from _rangesToClean.
-     * Should not be called if _rangesToClean is empty: it will hit an invariant.
-     */
-    ChunkRange getNextRangeToClean();
-
 private:
     friend class ScopedCollectionMetadata;
 
@@ -193,12 +151,6 @@ private:
      */
     void _removeMetadata_inlock(CollectionMetadataTracker* metadataTracker);
 
-    std::shared_ptr<Notification<Status>> _addRangeToClean_inlock(const ChunkRange& range);
-
-    void _removeRangeToClean_inlock(const ChunkRange& range, Status deletionStatus);
-
-    RangeMap _getCopyOfRangesToClean_inlock();
-
     void _setActiveMetadata_inlock(std::unique_ptr<CollectionMetadata> newMetadata);
 
     const NamespaceString _nss;
@@ -219,10 +171,6 @@ private:
     // Chunk ranges which are currently assumed to be transferred to the shard. Indexed by the min
     // key of the range.
     RangeMap _receivingChunks;
-
-    // Set of ranges to be deleted. Indexed by the min key of the range.
-    typedef BSONObjIndexedMap<RangeToCleanDescriptor> RangeToCleanMap;
-    RangeToCleanMap _rangesToClean;
 };
 
 class ScopedCollectionMetadata {
diff --git a/src/mongo/db/s/metadata_manager_test.cpp b/src/mongo/db/s/metadata_manager_test.cpp
index 19f4fd405ab..207eb4fd37b 100644
--- a/src/mongo/db/s/metadata_manager_test.cpp
+++ b/src/mongo/db/s/metadata_manager_test.cpp
@@ -50,12 +50,6 @@ using unittest::assertGet;
 
 class MetadataManagerTest : public ServiceContextMongoDTest {
 protected:
-    void setUp() override {
-        ServiceContextMongoDTest::setUp();
-        ShardingState::get(getServiceContext())
-            ->setScheduleCleanupFunctionForTest([](const NamespaceString& nss) {});
-    }
-
     static std::unique_ptr<CollectionMetadata> makeEmptyMetadata() {
         const OID epoch = OID::gen();
 
@@ -106,8 +100,7 @@ TEST_F(MetadataManagerTest, SetAndGetActiveMetadata) {
     ScopedCollectionMetadata scopedMetadata = manager.getActiveMetadata(manager_ptr);
 
     ASSERT_EQ(cmPtr, scopedMetadata.getMetadata());
-};
-
+}
 
 TEST_F(MetadataManagerTest, ResetActiveMetadata) {
     manager.refreshActiveMetadata(makeEmptyMetadata());
@@ -124,167 +117,8 @@ TEST_F(MetadataManagerTest, ResetActiveMetadata) {
     ScopedCollectionMetadata scopedMetadata2 = manager.getActiveMetadata(manager_ptr);
 
     ASSERT_EQ(cm2Ptr, scopedMetadata2.getMetadata());
-};
-
-TEST_F(MetadataManagerTest, AddAndRemoveRangesToClean) {
-    ChunkRange cr1 = ChunkRange(BSON("key" << 0), BSON("key" << 10));
-    ChunkRange cr2 = ChunkRange(BSON("key" << 10), BSON("key" << 20));
-
-    manager.addRangeToClean(cr1);
-    ASSERT_EQ(manager.getCopyOfRangesToClean().size(), 1UL);
-    manager.removeRangeToClean(cr1);
-    ASSERT_EQ(manager.getCopyOfRangesToClean().size(), 0UL);
-
-    manager.addRangeToClean(cr1);
-    manager.addRangeToClean(cr2);
-    manager.removeRangeToClean(cr1);
-    ASSERT_EQ(manager.getCopyOfRangesToClean().size(), 1UL);
-    auto ranges = manager.getCopyOfRangesToClean();
-    auto it = ranges.find(cr2.getMin());
-    ChunkRange remainingChunk = ChunkRange(it->first, it->second.getMaxKey());
-    ASSERT_EQ(remainingChunk.toString(), cr2.toString());
-    manager.removeRangeToClean(cr2);
-}
-
-// Tests that a removal in the middle of an existing ChunkRange results in
-// two correct chunk ranges.
-TEST_F(MetadataManagerTest, RemoveRangeInMiddleOfRange) {
-    ChunkRange cr1 = ChunkRange(BSON("key" << 0), BSON("key" << 10));
-
-    manager.addRangeToClean(cr1);
-    manager.removeRangeToClean(ChunkRange(BSON("key" << 4), BSON("key" << 6)));
-    ASSERT_EQ(manager.getCopyOfRangesToClean().size(), 2UL);
-
-    auto ranges = manager.getCopyOfRangesToClean();
-    auto it = ranges.find(BSON("key" << 0));
-    ChunkRange expectedChunk = ChunkRange(BSON("key" << 0), BSON("key" << 4));
-    ChunkRange remainingChunk = ChunkRange(it->first, it->second.getMaxKey());
-    ASSERT_EQ(remainingChunk.toString(), expectedChunk.toString());
-
-    it++;
-    expectedChunk = ChunkRange(BSON("key" << 6), BSON("key" << 10));
-    remainingChunk = ChunkRange(it->first, it->second.getMaxKey());
-    ASSERT_EQ(remainingChunk.toString(), expectedChunk.toString());
-
-    manager.removeRangeToClean(cr1);
-    ASSERT_EQ(manager.getCopyOfRangesToClean().size(), 0UL);
-}
-
-// Tests removals that overlap with just one ChunkRange.
-TEST_F(MetadataManagerTest, RemoveRangeWithSingleRangeOverlap) {
-    ChunkRange cr1 = ChunkRange(BSON("key" << 0), BSON("key" << 10));
-
-    manager.addRangeToClean(cr1);
-    manager.removeRangeToClean(ChunkRange(BSON("key" << 0), BSON("key" << 5)));
-    ASSERT_EQ(manager.getCopyOfRangesToClean().size(), 1UL);
-    auto ranges = manager.getCopyOfRangesToClean();
-    auto it = ranges.find(BSON("key" << 5));
-    ChunkRange remainingChunk = ChunkRange(it->first, it->second.getMaxKey());
-    ChunkRange expectedChunk = ChunkRange(BSON("key" << 5), BSON("key" << 10));
-    ASSERT_EQ(remainingChunk.toString(), expectedChunk.toString());
-
-    manager.removeRangeToClean(ChunkRange(BSON("key" << 4), BSON("key" << 6)));
-    ASSERT_EQ(manager.getCopyOfRangesToClean().size(), 1UL);
-    ranges = manager.getCopyOfRangesToClean();
-    it = ranges.find(BSON("key" << 6));
-    remainingChunk = ChunkRange(it->first, it->second.getMaxKey());
-    expectedChunk = ChunkRange(BSON("key" << 6), BSON("key" << 10));
-    ASSERT_EQ(remainingChunk.toString(), expectedChunk.toString());
-
-    manager.removeRangeToClean(ChunkRange(BSON("key" << 9), BSON("key" << 13)));
-    ASSERT_EQ(manager.getCopyOfRangesToClean().size(), 1UL);
-    ranges = manager.getCopyOfRangesToClean();
-    it = ranges.find(BSON("key" << 6));
-    remainingChunk = ChunkRange(it->first, it->second.getMaxKey());
-    expectedChunk = ChunkRange(BSON("key" << 6), BSON("key" << 9));
-    ASSERT_EQ(remainingChunk.toString(), expectedChunk.toString());
-
-    manager.removeRangeToClean(ChunkRange(BSON("key" << 0), BSON("key" << 10)));
-    ASSERT_EQ(manager.getCopyOfRangesToClean().size(), 0UL);
-}
-
-// Tests removals that overlap with more than one ChunkRange.
-TEST_F(MetadataManagerTest, RemoveRangeWithMultipleRangeOverlaps) {
-    ChunkRange cr1 = ChunkRange(BSON("key" << 0), BSON("key" << 10));
-    ChunkRange cr2 = ChunkRange(BSON("key" << 10), BSON("key" << 20));
-    ChunkRange cr3 = ChunkRange(BSON("key" << 20), BSON("key" << 30));
-
-    manager.addRangeToClean(cr1);
-    manager.addRangeToClean(cr2);
-    manager.addRangeToClean(cr3);
-    ASSERT_EQ(manager.getCopyOfRangesToClean().size(), 3UL);
-
-    manager.removeRangeToClean(ChunkRange(BSON("key" << 8), BSON("key" << 22)));
-    ASSERT_EQ(manager.getCopyOfRangesToClean().size(), 2UL);
-    auto ranges = manager.getCopyOfRangesToClean();
-    auto it = ranges.find(BSON("key" << 0));
-    ChunkRange remainingChunk = ChunkRange(it->first, it->second.getMaxKey());
-    ChunkRange expectedChunk = ChunkRange(BSON("key" << 0), BSON("key" << 8));
-    ASSERT_EQ(remainingChunk.toString(), expectedChunk.toString());
-    it++;
-    remainingChunk = ChunkRange(it->first, it->second.getMaxKey());
-    expectedChunk = ChunkRange(BSON("key" << 22), BSON("key" << 30));
-    ASSERT_EQ(remainingChunk.toString(), expectedChunk.toString());
-
-    manager.removeRangeToClean(ChunkRange(BSON("key" << 0), BSON("key" << 30)));
-    ASSERT_EQ(manager.getCopyOfRangesToClean().size(), 0UL);
-}
-
-TEST_F(MetadataManagerTest, AddAndRemoveRangeNotificationsBlockAndYield) {
-    manager.refreshActiveMetadata(makeEmptyMetadata());
-
-    ChunkRange cr1(BSON("key" << 0), BSON("key" << 10));
-    auto notification = manager.addRangeToClean(cr1);
-    manager.removeRangeToClean(cr1, Status::OK());
-    ASSERT_OK(notification->get());
-    ASSERT_EQ(manager.getCopyOfRangesToClean().size(), 0UL);
-}
-
-TEST_F(MetadataManagerTest, RemoveRangeToCleanCorrectlySetsBadStatus) {
-    manager.refreshActiveMetadata(makeEmptyMetadata());
-
-    ChunkRange cr1(BSON("key" << 0), BSON("key" << 10));
-    auto notification = manager.addRangeToClean(cr1);
-    manager.removeRangeToClean(cr1, Status(ErrorCodes::InternalError, "test error"));
-    ASSERT_NOT_OK(notification->get());
-    ASSERT_EQ(manager.getCopyOfRangesToClean().size(), 0UL);
-}
-
-TEST_F(MetadataManagerTest, RemovingSubrangeStillSetsNotificationStatus) {
-    manager.refreshActiveMetadata(makeEmptyMetadata());
-
-    ChunkRange cr1(BSON("key" << 0), BSON("key" << 10));
-    auto notification = manager.addRangeToClean(cr1);
-    manager.removeRangeToClean(ChunkRange(BSON("key" << 3), BSON("key" << 7)));
-    ASSERT_OK(notification->get());
-    ASSERT_EQ(manager.getCopyOfRangesToClean().size(), 2UL);
-    manager.removeRangeToClean(cr1);
-    ASSERT_EQ(manager.getCopyOfRangesToClean().size(), 0UL);
-
-    notification = manager.addRangeToClean(cr1);
-    manager.removeRangeToClean(ChunkRange(BSON("key" << 7), BSON("key" << 15)));
-    ASSERT_OK(notification->get());
-    ASSERT_EQ(manager.getCopyOfRangesToClean().size(), 1UL);
-    manager.removeRangeToClean(cr1);
-    ASSERT_EQ(manager.getCopyOfRangesToClean().size(), 0UL);
 }
 
-TEST_F(MetadataManagerTest, NotificationBlocksUntilDeletion) {
-    manager.refreshActiveMetadata(makeEmptyMetadata());
-
-    ChunkRange cr1(BSON("key" << 0), BSON("key" << 10));
-    auto notification = manager.addRangeToClean(cr1);
-    auto txn = cc().makeOperationContext().get();
-    // Once the new range deleter is set up, this might fail if the range deleter
-    // deleted cr1 before we got here...
-    ASSERT_FALSE(notification->waitFor(txn, Milliseconds(0)));
-
-    manager.removeRangeToClean(cr1);
-    ASSERT_TRUE(notification->waitFor(txn, Milliseconds(0)));
-    ASSERT_OK(notification->get());
-}
-
-
 TEST_F(MetadataManagerTest, RefreshAfterSuccessfulMigrationSinglePending) {
     manager.refreshActiveMetadata(makeEmptyMetadata());
 
@@ -417,29 +251,5 @@ TEST_F(MetadataManagerTest, RefreshMetadataAfterDropAndRecreate) {
     ASSERT_EQ(newVersion, chunkEntry->second.getVersion());
 }
 
-// Tests membership functions for _rangesToClean
-TEST_F(MetadataManagerTest, RangesToCleanMembership) {
-    manager.refreshActiveMetadata(makeEmptyMetadata());
-
-    ASSERT(!manager.hasRangesToClean());
-
-    ChunkRange cr1 = ChunkRange(BSON("key" << 0), BSON("key" << 10));
-    manager.addRangeToClean(cr1);
-
-    ASSERT(manager.hasRangesToClean());
-    ASSERT(manager.isInRangesToClean(cr1));
-}
-
-// Tests that getNextRangeToClean successfully pulls a stored ChunkRange
-TEST_F(MetadataManagerTest, GetNextRangeToClean) {
-    manager.refreshActiveMetadata(makeEmptyMetadata());
-
-    ChunkRange cr1 = ChunkRange(BSON("key" << 0), BSON("key" << 10));
-    manager.addRangeToClean(cr1);
-
-    ChunkRange cr2 = manager.getNextRangeToClean();
-    ASSERT_EQ(cr1.toString(), cr2.toString());
-}
-
 }  // namespace
 }  // namespace mongo
diff --git a/src/mongo/db/s/migration_chunk_cloner_source.h b/src/mongo/db/s/migration_chunk_cloner_source.h
index 04cf9e36df2..cb3b10a8ef8 100644
--- a/src/mongo/db/s/migration_chunk_cloner_source.h
+++ b/src/mongo/db/s/migration_chunk_cloner_source.h
@@ -88,9 +88,13 @@ public:
      * This must only be called once and no more methods on the cloner must be used afterwards
      * regardless of whether it succeeds or not.
      *
+     * Returns statistics about the move. These are informational only and should not be
+     * interpreted by the caller for any means other than reporting.
+     *
      * NOTE: Must be called without any locks.
      */
-    virtual Status commitClone(OperationContext* txn) = 0;
+
+    virtual StatusWith<BSONObj> commitClone(OperationContext* opCtx) = 0;
 
     /**
      * Tells the recipient to abort the clone and cleanup any unused data. This method's
diff --git a/src/mongo/db/s/migration_chunk_cloner_source_legacy.cpp b/src/mongo/db/s/migration_chunk_cloner_source_legacy.cpp
index a11d74ef587..ea83fa83d0d 100644
--- a/src/mongo/db/s/migration_chunk_cloner_source_legacy.cpp
+++ b/src/mongo/db/s/migration_chunk_cloner_source_legacy.cpp
@@ -75,9 +75,11 @@ bool isInRange(const BSONObj& obj,
 
 BSONObj createRequestWithSessionId(StringData commandName,
                                    const NamespaceString& nss,
-                                   const MigrationSessionId& sessionId) {
+                                   const MigrationSessionId& sessionId,
+                                   bool waitForSteadyOrDone = false) {
     BSONObjBuilder builder;
     builder.append(commandName, nss.ns());
+    builder.append("waitForSteadyOrDone", waitForSteadyOrDone);
     sessionId.append(&builder);
     return builder.obj();
 }
@@ -231,13 +233,8 @@ Status MigrationChunkClonerSourceLegacy::awaitUntilCriticalSectionIsAppropriate(
 
     int iteration = 0;
     while ((Date_t::now() - startTime) < maxTimeToWait) {
-        // Exponential sleep backoff, up to 1024ms. Don't sleep much on the first few iterations,
-        // since we want empty chunk migrations to be fast.
-        sleepmillis(1LL << std::min(iteration, 10));
-        iteration++;
-
         auto responseStatus = _callRecipient(
-            createRequestWithSessionId(kRecvChunkStatus, _args.getNss(), _sessionId));
+            createRequestWithSessionId(kRecvChunkStatus, _args.getNss(), _sessionId, true));
         if (!responseStatus.isOK()) {
             return {responseStatus.getStatus().code(),
                     str::stream()
@@ -247,6 +244,11 @@ Status MigrationChunkClonerSourceLegacy::awaitUntilCriticalSectionIsAppropriate(
 
         const BSONObj& res = responseStatus.getValue();
 
+        if (!res["waited"].boolean()) {
+            sleepmillis(1LL << std::min(iteration, 10));
+        }
+        iteration++;
+
         stdx::lock_guard<stdx::mutex> sl(_mutex);
 
         const std::size_t cloneLocsRemaining = _cloneLocs.size();
@@ -306,7 +308,7 @@ Status MigrationChunkClonerSourceLegacy::awaitUntilCriticalSectionIsAppropriate(
     return {ErrorCodes::ExceededTimeLimit, "Timed out waiting for the cloner to catch up"};
 }
 
-Status MigrationChunkClonerSourceLegacy::commitClone(OperationContext* txn) {
+StatusWith<BSONObj> MigrationChunkClonerSourceLegacy::commitClone(OperationContext* txn) {
     invariant(_state == kCloning);
     invariant(!txn->lockState()->isLocked());
 
@@ -314,7 +316,7 @@ Status MigrationChunkClonerSourceLegacy::commitClone(OperationContext* txn) {
         _callRecipient(createRequestWithSessionId(kRecvChunkCommit, _args.getNss(), _sessionId));
     if (responseStatus.isOK()) {
         _cleanup(txn);
-        return Status::OK();
+        return responseStatus;
     }
 
     cancelClone(txn);
@@ -576,8 +578,7 @@ Status MigrationChunkClonerSourceLegacy::_storeCurrentLocs(OperationContext* txn
     if (totalRecs > 0) {
         avgRecSize = collection->dataSize(txn) / totalRecs;
         maxRecsWhenFull = _args.getMaxChunkSizeBytes() / avgRecSize;
-        maxRecsWhenFull = std::min((unsigned long long)(kMaxObjectPerChunk + 1),
-                                   130 * maxRecsWhenFull / 100 /* slack */);
+        maxRecsWhenFull = 130 * maxRecsWhenFull / 100;  // pad some slack
     } else {
         avgRecSize = 0;
         maxRecsWhenFull = kMaxObjectPerChunk + 1;
diff --git a/src/mongo/db/s/migration_chunk_cloner_source_legacy.h b/src/mongo/db/s/migration_chunk_cloner_source_legacy.h
index c683df2be29..860282a16d0 100644
--- a/src/mongo/db/s/migration_chunk_cloner_source_legacy.h
+++ b/src/mongo/db/s/migration_chunk_cloner_source_legacy.h
@@ -66,7 +66,7 @@ public:
     Status awaitUntilCriticalSectionIsAppropriate(OperationContext* txn,
                                                   Milliseconds maxTimeToWait) override;
 
-    Status commitClone(OperationContext* txn) override;
+    StatusWith<BSONObj> commitClone(OperationContext* txn) override;
 
     void cancelClone(OperationContext* txn) override;
 
diff --git a/src/mongo/db/s/migration_destination_manager.cpp b/src/mongo/db/s/migration_destination_manager.cpp
index 2e748cf3ca9..1c02208ee27 100644
--- a/src/mongo/db/s/migration_destination_manager.cpp
+++ b/src/mongo/db/s/migration_destination_manager.cpp
@@ -148,10 +148,14 @@ bool opReplicatedEnough(OperationContext* txn,
                         const repl::OpTime& lastOpApplied,
                         const WriteConcernOptions& writeConcern) {
     WriteConcernResult writeConcernResult;
+    writeConcernResult.wTimedOut = false;
 
-    Status waitForMajorityWriteConcernStatus =
+    Status majorityStatus =
         waitForWriteConcern(txn, lastOpApplied, kMajorityWriteConcern, &writeConcernResult);
-    if (!waitForMajorityWriteConcernStatus.isOK()) {
+    if (!majorityStatus.isOK()) {
+        if (!writeConcernResult.wTimedOut) {
+            uassertStatusOK(majorityStatus);
+        }
         return false;
     }
 
@@ -159,13 +163,16 @@ bool opReplicatedEnough(OperationContext* txn,
     // write concerns in case the user's write concern is stronger than majority
     WriteConcernOptions userWriteConcern(writeConcern);
     userWriteConcern.wTimeout = -1;
+    writeConcernResult.wTimedOut = false;
 
-    Status waitForUserWriteConcernStatus =
+    Status userStatus =
         waitForWriteConcern(txn, lastOpApplied, userWriteConcern, &writeConcernResult);
-    if (!waitForUserWriteConcernStatus.isOK()) {
+    if (!userStatus.isOK()) {
+        if (!writeConcernResult.wTimedOut) {
+            uassertStatusOK(userStatus);
+        }
         return false;
     }
-
     return true;
 }
 
@@ -220,6 +227,7 @@ MigrationDestinationManager::State MigrationDestinationManager::getState() const
 void MigrationDestinationManager::setState(State newState) {
     stdx::lock_guard<stdx::mutex> sl(_mutex);
     _state = newState;
+    _stateChangedCV.notify_all();
 }
 
 bool MigrationDestinationManager::isActive() const {
@@ -231,7 +239,21 @@ bool MigrationDestinationManager::_isActive_inlock() const {
     return _sessionId.is_initialized();
 }
 
-void MigrationDestinationManager::report(BSONObjBuilder& b) {
+void MigrationDestinationManager::report(BSONObjBuilder& b,
+                                         OperationContext* opCtx,
+                                         bool waitForSteadyOrDone) {
+    if (waitForSteadyOrDone) {
+        stdx::unique_lock<stdx::mutex> lock(_mutex);
+        try {
+            opCtx->waitForConditionOrInterruptFor(_stateChangedCV, lock, Seconds(1), [&]() -> bool {
+                return _state != READY && _state != CLONE && _state != CATCHUP;
+            });
+        } catch (...) {
+            // Ignoring this error because this is an optional parameter and we catch timeout
+            // exceptions later.
+        }
+        b.append("waited", true);
+    }
     stdx::lock_guard<stdx::mutex> sl(_mutex);
 
     b.appendBool("active", _sessionId.is_initialized());
@@ -286,6 +308,7 @@ Status MigrationDestinationManager::start(const NamespaceString& nss,
     invariant(!_scopedRegisterReceiveChunk);
 
     _state = READY;
+    _stateChangedCV.notify_all();
     _errmsg = "";
 
     _nss = nss;
@@ -336,6 +359,7 @@ bool MigrationDestinationManager::abort(const MigrationSessionId& sessionId) {
     }
 
     _state = ABORT;
+    _stateChangedCV.notify_all();
     _errmsg = "aborted";
 
     return true;
@@ -344,6 +368,7 @@ bool MigrationDestinationManager::abort(const MigrationSessionId& sessionId) {
 void MigrationDestinationManager::abortWithoutSessionIdCheck() {
     stdx::lock_guard<stdx::mutex> sl(_mutex);
     _state = ABORT;
+    _stateChangedCV.notify_all();
     _errmsg = "aborted without session id check";
 }
 
@@ -368,6 +393,7 @@ bool MigrationDestinationManager::startCommit(const MigrationSessionId& sessionI
     }
 
     _state = COMMIT_START;
+    _stateChangedCV.notify_all();
 
     const auto deadline = Date_t::now() + Seconds(30);
 
@@ -375,6 +401,8 @@ bool MigrationDestinationManager::startCommit(const MigrationSessionId& sessionI
         if (stdx::cv_status::timeout ==
             _isActiveCV.wait_until(lock, deadline.toSystemTimePoint())) {
             _state = FAIL;
+            _stateChangedCV.notify_all();
+
             log() << "startCommit never finished!" << migrateLog;
             return false;
         }
diff --git a/src/mongo/db/s/migration_destination_manager.h b/src/mongo/db/s/migration_destination_manager.h
index 0b16202d55e..4bd269f37fd 100644
--- a/src/mongo/db/s/migration_destination_manager.h
+++ b/src/mongo/db/s/migration_destination_manager.h
@@ -77,7 +77,7 @@ public:
     /**
      * Reports the state of the migration manager as a BSON document.
      */
-    void report(BSONObjBuilder& b);
+    void report(BSONObjBuilder& b, OperationContext* opCtx, bool waitForSteadyOrDone);
 
     /**
      * Returns a report on the active migration, if the migration is active. Otherwise return an
@@ -222,6 +222,9 @@ private:
 
     State _state{READY};
     std::string _errmsg;
+
+    // Condition variable, which is signalled every time the state of the migration changes.
+    stdx::condition_variable _stateChangedCV;
 };
 
 }  // namespace mongo
diff --git a/src/mongo/db/s/migration_destination_manager_legacy_commands.cpp b/src/mongo/db/s/migration_destination_manager_legacy_commands.cpp
index a20db670259..ac9752fd96d 100644
--- a/src/mongo/db/s/migration_destination_manager_legacy_commands.cpp
+++ b/src/mongo/db/s/migration_destination_manager_legacy_commands.cpp
@@ -220,7 +220,9 @@ public:
              int,
              string& errmsg,
              BSONObjBuilder& result) {
-        ShardingState::get(txn)->migrationDestinationManager()->report(result);
+        bool waitForSteadyOrDone = cmdObj["waitForSteadyOrDone"].boolean();
+        ShardingState::get(txn)->migrationDestinationManager()->report(
+            result, txn, waitForSteadyOrDone);
         return true;
     }
 
@@ -261,12 +263,13 @@ public:
              int,
              string& errmsg,
              BSONObjBuilder& result) {
+
         const MigrationSessionId migrationSessionid(
             uassertStatusOK(MigrationSessionId::extractFromBSON(cmdObj)));
         const bool ok =
             ShardingState::get(txn)->migrationDestinationManager()->startCommit(migrationSessionid);
 
-        ShardingState::get(txn)->migrationDestinationManager()->report(result);
+        ShardingState::get(txn)->migrationDestinationManager()->report(result, txn, false);
         return ok;
     }
 
@@ -313,11 +316,11 @@ public:
 
         if (migrationSessionIdStatus.isOK()) {
             const bool ok = mdm->abort(migrationSessionIdStatus.getValue());
-            mdm->report(result);
+            mdm->report(result, txn, false);
             return ok;
         } else if (migrationSessionIdStatus == ErrorCodes::NoSuchKey) {
             mdm->abortWithoutSessionIdCheck();
-            mdm->report(result);
+            mdm->report(result, txn, false);
             return true;
         }
 
diff --git a/src/mongo/db/s/migration_source_manager.cpp b/src/mongo/db/s/migration_source_manager.cpp
index 0ad841ea56f..b70617deee3 100644
--- a/src/mongo/db/s/migration_source_manager.cpp
+++ b/src/mongo/db/s/migration_source_manager.cpp
@@ -268,18 +268,18 @@ Status MigrationSourceManager::commitChunkOnRecipient(OperationContext* txn) {
     auto scopedGuard = MakeGuard([&] { cleanupOnError(txn); });
 
     // Tell the recipient shard to fetch the latest changes.
-    Status commitCloneStatus = _cloneDriver->commitClone(txn);
+    auto commitCloneStatus = _cloneDriver->commitClone(txn);
 
     if (MONGO_FAIL_POINT(failMigrationCommit) && commitCloneStatus.isOK()) {
         commitCloneStatus = {ErrorCodes::InternalError,
                              "Failing _recvChunkCommit due to failpoint."};
     }
-
     if (!commitCloneStatus.isOK()) {
-        return {commitCloneStatus.code(),
-                str::stream() << "commit clone failed due to " << commitCloneStatus.toString()};
+        return commitCloneStatus.getStatus();
     }
 
+    _recipientCloneCounts = commitCloneStatus.getValue()["counts"].Obj().getOwned();
+
     _state = kCloneCompleted;
     scopedGuard.Dismiss();
     return Status::OK();
@@ -332,9 +332,13 @@ Status MigrationSourceManager::commitChunkMetadataOnConfig(OperationContext* txn
             ErrorCodes::InternalError, "Failpoint 'migrationCommitNetworkError' generated error");
     }
 
-    const Status migrationCommitStatus =
-        (commitChunkMigrationResponse.isOK() ? commitChunkMigrationResponse.getValue().commandStatus
-                                             : commitChunkMigrationResponse.getStatus());
+    Status migrationCommitStatus = commitChunkMigrationResponse.getStatus();
+    if (migrationCommitStatus.isOK()) {
+        migrationCommitStatus = commitChunkMigrationResponse.getValue().commandStatus;
+        if (migrationCommitStatus.isOK()) {
+            migrationCommitStatus = commitChunkMigrationResponse.getValue().writeConcernStatus;
+        }
+    }
 
     if (!migrationCommitStatus.isOK()) {
         // Need to get the latest optime in case the refresh request goes to a secondary --
@@ -431,7 +435,9 @@ Status MigrationSourceManager::commitChunkMetadataOnConfig(OperationContext* txn
                                                   << "from"
                                                   << _args.getFromShardId()
                                                   << "to"
-                                                  << _args.getToShardId()),
+                                                  << _args.getToShardId()
+                                                  << "counts"
+                                                  << _recipientCloneCounts),
                                        ShardingCatalogClient::kMajorityWriteConcern);
 
     return Status::OK();
diff --git a/src/mongo/db/s/migration_source_manager.h b/src/mongo/db/s/migration_source_manager.h
index cb5ce4be792..27d4480a665 100644
--- a/src/mongo/db/s/migration_source_manager.h
+++ b/src/mongo/db/s/migration_source_manager.h
@@ -233,6 +233,9 @@ private:
     // callers don't have to hold collection lock in order to wait on it. Available after the
     // critical section stage has completed.
     std::shared_ptr<Notification<void>> _critSecSignal;
+
+    // The statistics about a chunk migration to be included in moveChunk.commit
+    BSONObj _recipientCloneCounts;
 };
 
 }  // namespace mongo
diff --git a/src/mongo/db/s/sharding_state.cpp b/src/mongo/db/s/sharding_state.cpp
index e264c82b57f..02c75e287fe 100644
--- a/src/mongo/db/s/sharding_state.cpp
+++ b/src/mongo/db/s/sharding_state.cpp
@@ -117,8 +117,7 @@ const std::set<std::string> ShardingState::_commandsThatInitializeShardingAwaren
 ShardingState::ShardingState()
     : _initializationState(static_cast<uint32_t>(InitializationState::kNew)),
       _initializationStatus(Status(ErrorCodes::InternalError, "Uninitialized value")),
-      _globalInit(&initializeGlobalShardingStateForMongod),
-      _scheduleWorkFn([](NamespaceString nss) {}) {}
+      _globalInit(&initializeGlobalShardingStateForMongod) {}
 
 ShardingState::~ShardingState() = default;
 
@@ -222,14 +221,6 @@ void ShardingState::setGlobalInitMethodForTest(GlobalInitFunc func) {
     _globalInit = func;
 }
 
-void ShardingState::setScheduleCleanupFunctionForTest(RangeDeleterCleanupNotificationFunc fn) {
-    _scheduleWorkFn = fn;
-}
-
-void ShardingState::scheduleCleanup(const NamespaceString& nss) {
-    _scheduleWorkFn(nss);
-}
-
 Status ShardingState::onStaleShardVersion(OperationContext* txn,
                                           const NamespaceString& nss,
                                           const ChunkVersion& expectedVersion) {
@@ -416,8 +407,6 @@ Status ShardingState::initializeFromShardIdentity(OperationContext* txn,
             _shardName = shardIdentity.getShardName();
             _clusterId = shardIdentity.getClusterId();
 
-            _initializeRangeDeleterTaskExecutor();
-
             return status;
         } catch (const DBException& ex) {
             auto errorStatus = ex.toStatus();
@@ -446,8 +435,6 @@ void ShardingState::_initializeImpl(ConnectionString configSvr, string shardName
                 &ShardRegistry::replicaSetChangeShardRegistryUpdateHook);
             ReplicaSetMonitor::setAsynchronousConfigChangeHook(&updateShardIdentityConfigStringCB);
 
-            _initializeRangeDeleterTaskExecutor();
-
             _shardName = shardName;
         }
 
@@ -626,47 +613,76 @@ ChunkVersion ShardingState::_refreshMetadata(OperationContext* txn, const Namesp
                           << " before shard name has been set",
             shardId.isValid());
 
-    auto newCollectionMetadata = [&]() -> std::unique_ptr<CollectionMetadata> {
-        auto const catalogCache = Grid::get(txn)->catalogCache();
-        catalogCache->invalidateShardedCollection(nss);
+    auto const catalogCache = Grid::get(txn)->catalogCache();
+    catalogCache->invalidateShardedCollection(nss);
 
-        const auto routingInfo = uassertStatusOK(catalogCache->getCollectionRoutingInfo(txn, nss));
-        const auto cm = routingInfo.cm();
-        if (!cm) {
-            return nullptr;
-        }
+    auto routingInfo = uassertStatusOK(catalogCache->getCollectionRoutingInfo(txn, nss));
+    const auto cm = routingInfo.cm();
 
-        RangeMap shardChunksMap =
-            SimpleBSONObjComparator::kInstance.makeBSONObjIndexedMap<CachedChunkInfo>();
+    if (!cm) {
+        // No chunk manager, so unsharded.
 
-        for (const auto& chunkMapEntry : cm->chunkMap()) {
-            const auto& chunk = chunkMapEntry.second;
+        // Exclusive collection lock needed since we're now changing the metadata
+        ScopedTransaction transaction(txn, MODE_IX);
+        AutoGetCollection autoColl(txn, nss, MODE_IX, MODE_X);
 
-            if (chunk->getShardId() != shardId)
-                continue;
+        auto css = CollectionShardingState::get(txn, nss);
+        css->refreshMetadata(txn, nullptr);
 
-            shardChunksMap.emplace_hint(shardChunksMap.end(),
-                                        chunk->getMin(),
-                                        CachedChunkInfo(chunk->getMax(), chunk->getLastmod()));
-        }
+        return ChunkVersion::UNSHARDED();
+    }
 
-        return stdx::make_unique<CollectionMetadata>(cm->getShardKeyPattern().toBSON(),
-                                                     cm->getVersion(),
-                                                     cm->getVersion(shardId),
-                                                     std::move(shardChunksMap));
-    }();
+    {
+        AutoGetCollection autoColl(txn, nss, MODE_IS);
+        auto css = CollectionShardingState::get(txn, nss);
+
+        // We already have newer version
+        if (css->getMetadata() &&
+            css->getMetadata()->getCollVersion().epoch() == cm->getVersion().epoch() &&
+            css->getMetadata()->getCollVersion() >= cm->getVersion()) {
+            LOG(1) << "Skipping refresh of metadata for " << nss << " "
+                   << css->getMetadata()->getCollVersion() << " with an older " << cm->getVersion();
+            return css->getMetadata()->getShardVersion();
+        }
+    }
 
     // Exclusive collection lock needed since we're now changing the metadata
     ScopedTransaction transaction(txn, MODE_IX);
     AutoGetCollection autoColl(txn, nss, MODE_IX, MODE_X);
 
     auto css = CollectionShardingState::get(txn, nss);
-    css->refreshMetadata(txn, std::move(newCollectionMetadata));
 
-    if (!css->getMetadata()) {
-        return ChunkVersion::UNSHARDED();
+    // We already have newer version
+    if (css->getMetadata() &&
+        css->getMetadata()->getCollVersion().epoch() == cm->getVersion().epoch() &&
+        css->getMetadata()->getCollVersion() >= cm->getVersion()) {
+        LOG(1) << "Skipping refresh of metadata for " << nss << " "
+               << css->getMetadata()->getCollVersion() << " with an older " << cm->getVersion();
+        return css->getMetadata()->getShardVersion();
+    }
+
+    RangeMap shardChunksMap =
+        SimpleBSONObjComparator::kInstance.makeBSONObjIndexedMap<CachedChunkInfo>();
+
+    for (const auto& chunkMapEntry : cm->chunkMap()) {
+        const auto& chunk = chunkMapEntry.second;
+
+        if (chunk->getShardId() != shardId)
+            continue;
+
+        shardChunksMap.emplace_hint(shardChunksMap.end(),
+                                    chunk->getMin(),
+                                    CachedChunkInfo(chunk->getMax(), chunk->getLastmod()));
     }
 
+    std::unique_ptr<CollectionMetadata> newCollectionMetadata =
+        stdx::make_unique<CollectionMetadata>(cm->getShardKeyPattern().toBSON(),
+                                              cm->getVersion(),
+                                              cm->getVersion(shardId),
+                                              std::move(shardChunksMap));
+
+    css->refreshMetadata(txn, std::move(newCollectionMetadata));
+
     return css->getMetadata()->getShardVersion();
 }
 
@@ -756,19 +772,6 @@ Status ShardingState::updateShardIdentityConfigString(OperationContext* txn,
     return Status::OK();
 }
 
-void ShardingState::_initializeRangeDeleterTaskExecutor() {
-    invariant(!_rangeDeleterTaskExecutor);
-    auto net =
-        executor::makeNetworkInterface("NetworkInterfaceCollectionRangeDeleter-TaskExecutor");
-    auto netPtr = net.get();
-    _rangeDeleterTaskExecutor = stdx::make_unique<executor::ThreadPoolTaskExecutor>(
-        stdx::make_unique<executor::NetworkInterfaceThreadPool>(netPtr), std::move(net));
-}
-
-executor::ThreadPoolTaskExecutor* ShardingState::getRangeDeleterTaskExecutor() {
-    return _rangeDeleterTaskExecutor.get();
-}
-
 /**
  * Global free function.
  */
diff --git a/src/mongo/db/s/sharding_state.h b/src/mongo/db/s/sharding_state.h
index e53fee63515..f9e5a507446 100644
--- a/src/mongo/db/s/sharding_state.h
+++ b/src/mongo/db/s/sharding_state.h
@@ -35,7 +35,6 @@
 #include "mongo/bson/oid.h"
 #include "mongo/db/namespace_string.h"
 #include "mongo/db/s/active_migrations_registry.h"
-#include "mongo/db/s/collection_range_deleter.h"
 #include "mongo/db/s/migration_destination_manager.h"
 #include "mongo/executor/task_executor.h"
 #include "mongo/executor/thread_pool_task_executor.h"
@@ -236,24 +235,6 @@ public:
     void setGlobalInitMethodForTest(GlobalInitFunc func);
 
     /**
-     * Schedules for the range to clean of the given namespace to be deleted.
-     * Behavior can be modified through setScheduleCleanupFunctionForTest.
-     */
-    void scheduleCleanup(const NamespaceString& nss);
-
-    /**
-     * Returns a pointer to the collection range deleter task executor.
-     */
-    executor::ThreadPoolTaskExecutor* getRangeDeleterTaskExecutor();
-
-    /**
-     * Sets the function used by scheduleWorkOnRangeDeleterTaskExecutor to
-     * schedule work. Used for mocking the executor for testing. See the ShardingState
-     * for the default implementation of _scheduleWorkFn.
-     */
-    void setScheduleCleanupFunctionForTest(RangeDeleterCleanupNotificationFunc fn);
-
-    /**
      * If started with --shardsvr, initializes sharding awareness from the shardIdentity document
      * on disk, if there is one.
      * If started with --shardsvr in queryableBackupMode, initializes sharding awareness from the
@@ -348,9 +329,6 @@ private:
      */
     ChunkVersion _refreshMetadata(OperationContext* opCtx, const NamespaceString& nss);
 
-    // Initializes a TaskExecutor for cleaning up orphaned ranges
-    void _initializeRangeDeleterTaskExecutor();
-
     // Manages the state of the migration recipient shard
     MigrationDestinationManager _migrationDestManager;
 
@@ -386,13 +364,6 @@ private:
 
     // Function for initializing the external sharding state components not owned here.
     GlobalInitFunc _globalInit;
-
-    // Function for scheduling work on the _rangeDeleterTaskExecutor.
-    // Used in call to scheduleCleanup(NamespaceString).
-    RangeDeleterCleanupNotificationFunc _scheduleWorkFn;
-
-    // Task executor for the collection range deleter.
-    std::unique_ptr<executor::ThreadPoolTaskExecutor> _rangeDeleterTaskExecutor;
 };
 
 }  // namespace mongo
diff --git a/src/mongo/db/sorter/SConscript b/src/mongo/db/sorter/SConscript
index 959f6cdb919..1ceb05088af 100644
--- a/src/mongo/db/sorter/SConscript
+++ b/src/mongo/db/sorter/SConscript
@@ -5,6 +5,6 @@ sorterEnv.InjectThirdPartyIncludePaths(libraries=['snappy'])
 sorterEnv.CppUnitTest('sorter_test',
                       'sorter_test.cpp',
                        LIBDEPS=['$BUILD_DIR/mongo/db/service_context',
-                                '$BUILD_DIR/mongo/db/storage/wiredtiger/storage_wiredtiger_customization_hooks',
+                                '$BUILD_DIR/mongo/db/storage/encryption_hooks',
                                 '$BUILD_DIR/mongo/db/storage/storage_options',
                                 '$BUILD_DIR/third_party/shim_snappy'])
diff --git a/src/mongo/db/sorter/sorter.cpp b/src/mongo/db/sorter/sorter.cpp
index 7f6e8866562..0165e0adccc 100644
--- a/src/mongo/db/sorter/sorter.cpp
+++ b/src/mongo/db/sorter/sorter.cpp
@@ -55,8 +55,8 @@
 #include "mongo/config.h"
 #include "mongo/db/jsobj.h"
 #include "mongo/db/service_context.h"
+#include "mongo/db/storage/encryption_hooks.h"
 #include "mongo/db/storage/storage_options.h"
-#include "mongo/db/storage/wiredtiger/wiredtiger_customization_hooks.h"
 #include "mongo/platform/atomic_word.h"
 #include "mongo/s/mongos_options.h"
 #include "mongo/util/assert_util.h"
@@ -226,15 +226,16 @@ private:
         read(_buffer.get(), blockSize);
         massert(16816, "file too short?", !_done);
 
-        auto hooks = WiredTigerCustomizationHooks::get(getGlobalServiceContext());
-        if (hooks->enabled()) {
+        auto encryptionHooks = EncryptionHooks::get(getGlobalServiceContext());
+        if (encryptionHooks->enabled()) {
             std::unique_ptr<char[]> out(new char[blockSize]);
             size_t outLen;
-            Status status = hooks->unprotectTmpData(reinterpret_cast<uint8_t*>(_buffer.get()),
-                                                    blockSize,
-                                                    reinterpret_cast<uint8_t*>(out.get()),
-                                                    blockSize,
-                                                    &outLen);
+            Status status =
+                encryptionHooks->unprotectTmpData(reinterpret_cast<uint8_t*>(_buffer.get()),
+                                                  blockSize,
+                                                  reinterpret_cast<uint8_t*>(out.get()),
+                                                  blockSize,
+                                                  &outLen);
             massert(28841,
                     str::stream() << "Failed to unprotect data: " << status.toString(),
                     status.isOK());
@@ -884,16 +885,16 @@ void SortedFileWriter<Key, Value>::spill() {
     }
 
     std::unique_ptr<char[]> out;
-    auto hooks = WiredTigerCustomizationHooks::get(getGlobalServiceContext());
-    if (hooks->enabled()) {
-        size_t protectedSizeMax = size + hooks->additionalBytesForProtectedBuffer();
+    auto encryptionHooks = EncryptionHooks::get(getGlobalServiceContext());
+    if (encryptionHooks->enabled()) {
+        size_t protectedSizeMax = size + encryptionHooks->additionalBytesForProtectedBuffer();
         out.reset(new char[protectedSizeMax]);
         size_t resultLen;
-        Status status = hooks->protectTmpData(reinterpret_cast<const uint8_t*>(outBuffer),
-                                              size,
-                                              reinterpret_cast<uint8_t*>(out.get()),
-                                              protectedSizeMax,
-                                              &resultLen);
+        Status status = encryptionHooks->protectTmpData(reinterpret_cast<const uint8_t*>(outBuffer),
+                                                        size,
+                                                        reinterpret_cast<uint8_t*>(out.get()),
+                                                        protectedSizeMax,
+                                                        &resultLen);
         massert(28842,
                 str::stream() << "Failed to compress data: " << status.toString(),
                 status.isOK());
diff --git a/src/mongo/db/stats/range_deleter_server_status.cpp b/src/mongo/db/stats/range_deleter_server_status.cpp
index 817ffa444e4..e37143342f0 100644
--- a/src/mongo/db/stats/range_deleter_server_status.cpp
+++ b/src/mongo/db/stats/range_deleter_server_status.cpp
@@ -84,6 +84,11 @@ public:
                 entryBuilder.append("deleteStart", (*it)->deleteStartTS);
                 entryBuilder.append("deleteEnd", (*it)->deleteEndTS);
 
+                const auto waitForReplDurationMs =
+                    durationCount<Milliseconds>((*it)->waitForReplDurationMs);
+                if (waitForReplDurationMs > 0) {
+                    entryBuilder.append("waitForReplDurationMs", waitForReplDurationMs);
+                }
                 if ((*it)->waitForReplEndTS > Date_t()) {
                     entryBuilder.append("waitForReplStart", (*it)->waitForReplStartTS);
                     entryBuilder.append("waitForReplEnd", (*it)->waitForReplEndTS);
diff --git a/src/mongo/db/storage/SConscript b/src/mongo/db/storage/SConscript
index d65046ed6fa..e887882df87 100644
--- a/src/mongo/db/storage/SConscript
+++ b/src/mongo/db/storage/SConscript
@@ -11,7 +11,6 @@ env.SConscript(
     ],
 )
 
-
 env.Library(
     target='journal_listener',
     source=[
@@ -41,7 +40,6 @@ env.Library(
         ],
     )
 
-
 env.Library(
     target='bson_collection_catalog_entry',
     source=[
@@ -75,6 +73,19 @@ env.Library(
 )
 
 env.Library(
+    target='encryption_hooks',
+    source= [
+        'encryption_hooks.cpp',
+        ],
+    LIBDEPS= ['$BUILD_DIR/mongo/base',
+              '$BUILD_DIR/mongo/db/service_context'],
+    PROGDEPS_DEPENDENTS=[
+        '$BUILD_DIR/mongo/mongod',
+        '$BUILD_DIR/mongo/mongos',
+        ],
+    )
+
+env.Library(
     target='storage_options',
     source=[
         'storage_options.cpp',
diff --git a/src/mongo/db/storage/encryption_hooks.cpp b/src/mongo/db/storage/encryption_hooks.cpp
new file mode 100644
index 00000000000..9ee9a317dbe
--- /dev/null
+++ b/src/mongo/db/storage/encryption_hooks.cpp
@@ -0,0 +1,96 @@
+/**
+ *    Copyright (C) 2017 MongoDB Inc.
+ *
+ *    This program is free software: you can redistribute it and/or  modify
+ *    it under the terms of the GNU Affero General Public License, version 3,
+ *    as published by the Free Software Foundation.
+ *
+ *
+ *    This program is distributed in the hope that it will be useful,
+ *    but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *    GNU Affero General Public License for more details.
+ *
+ *    You should have received a copy of the GNU Affero General Public License
+ *    along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ *    As a special exception, the copyright holders give permission to link the
+ *    code of portions of this program with the OpenSSL library under certain
+ *    conditions as described in each individual source file and distribute
+ *    linked combinations including the program with the OpenSSL library. You
+ *    must comply with the GNU Affero General Public License in all respects for
+ *    all of the code used other than as permitted herein. If you modify file(s)
+ *    with this exception, you may extend this exception to your version of the
+ *    file(s), but you are not obligated to do so. If you do not wish to do so,
+ *    delete this exception statement from your version. If you delete this
+ *    exception statement from all source files in the program, then also delete
+ *    it in the license file.
+ */
+
+#include "mongo/platform/basic.h"
+
+#include "mongo/db/storage/encryption_hooks.h"
+
+#include <boost/filesystem/path.hpp>
+
+#include "mongo/base/init.h"
+#include "mongo/db/service_context.h"
+#include "mongo/db/storage/data_protector.h"
+#include "mongo/stdx/memory.h"
+
+namespace mongo {
+
+/* Make a EncryptionHooks pointer a decoration on the global ServiceContext */
+MONGO_INITIALIZER_WITH_PREREQUISITES(SetEncryptionHooks, ("SetGlobalEnvironment"))
+(InitializerContext* context) {
+    auto encryptionHooks = stdx::make_unique<EncryptionHooks>();
+    EncryptionHooks::set(getGlobalServiceContext(), std::move(encryptionHooks));
+
+    return Status::OK();
+}
+
+namespace {
+const auto getEncryptionHooks =
+    ServiceContext::declareDecoration<std::unique_ptr<EncryptionHooks>>();
+}  // namespace
+
+void EncryptionHooks::set(ServiceContext* service, std::unique_ptr<EncryptionHooks> custHooks) {
+    auto& hooks = getEncryptionHooks(service);
+    invariant(custHooks);
+    hooks = std::move(custHooks);
+}
+
+EncryptionHooks* EncryptionHooks::get(ServiceContext* service) {
+    return getEncryptionHooks(service).get();
+}
+
+EncryptionHooks::~EncryptionHooks() {}
+
+bool EncryptionHooks::enabled() const {
+    return false;
+}
+
+bool EncryptionHooks::restartRequired() {
+    return false;
+}
+
+std::unique_ptr<DataProtector> EncryptionHooks::getDataProtector() {
+    return std::unique_ptr<DataProtector>();
+}
+
+boost::filesystem::path EncryptionHooks::getProtectedPathSuffix() {
+    return "";
+}
+
+Status EncryptionHooks::protectTmpData(
+    const uint8_t* in, size_t inLen, uint8_t* out, size_t outLen, size_t* resultLen) {
+    return Status(ErrorCodes::InternalError,
+                  "Encryption hooks must be enabled to use preprocessTmpData.");
+}
+
+Status EncryptionHooks::unprotectTmpData(
+    const uint8_t* in, size_t inLen, uint8_t* out, size_t outLen, size_t* resultLen) {
+    return Status(ErrorCodes::InternalError,
+                  "Encryption hooks must be enabled to use postprocessTmpData.");
+}
+}  // namespace mongo
diff --git a/src/mongo/db/storage/encryption_hooks.h b/src/mongo/db/storage/encryption_hooks.h
new file mode 100644
index 00000000000..e1c9d553a10
--- /dev/null
+++ b/src/mongo/db/storage/encryption_hooks.h
@@ -0,0 +1,99 @@
+/**
+ *    Copyright (C) 2017 MongoDB Inc.
+ *
+ *    This program is free software: you can redistribute it and/or  modify
+ *    it under the terms of the GNU Affero General Public License, version 3,
+ *    as published by the Free Software Foundation.
+ *
+ *
+ *    This program is distributed in the hope that it will be useful,
+ *    but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *    GNU Affero General Public License for more details.
+ *
+ *    You should have received a copy of the GNU Affero General Public License
+ *    along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ *    As a special exception, the copyright holders give permission to link the
+ *    code of portions of this program with the OpenSSL library under certain
+ *    conditions as described in each individual source file and distribute
+ *    linked combinations including the program with the OpenSSL library. You
+ *    must comply with the GNU Affero General Public License in all respects for
+ *    all of the code used other than as permitted herein. If you modify file(s)
+ *    with this exception, you may extend this exception to your version of the
+ *    file(s), but you are not obligated to do so. If you do not wish to do so,
+ *    delete this exception statement from your version. If you delete this
+ *    exception statement from all source files in the program, then also delete
+ *    it in the license file.
+ */
+
+#pragma once
+
+#include <memory>
+#include <string>
+
+#include "mongo/base/disallow_copying.h"
+#include "mongo/db/jsobj.h"
+
+namespace boost {
+namespace filesystem {
+class path;
+}  // namespace filesystem
+}  // namespace boost
+
+namespace mongo {
+class DataProtector;
+class ServiceContext;
+
+class EncryptionHooks {
+public:
+    static void set(ServiceContext* service, std::unique_ptr<EncryptionHooks> custHooks);
+
+    static EncryptionHooks* get(ServiceContext* service);
+
+    virtual ~EncryptionHooks();
+
+    /**
+     * Returns true if the encryption hooks are enabled.
+     */
+    virtual bool enabled() const;
+
+    /**
+     * Perform any encryption engine initialization/sanity checking that needs to happen after
+     * storage engine initialization but before the server starts accepting incoming connections.
+     *
+     * Returns true if the server needs to be rebooted because of configuration changes.
+     */
+    virtual bool restartRequired();
+
+    /**
+     * Returns the maximum size addition when doing transforming temp data.
+     */
+    size_t additionalBytesForProtectedBuffer() {
+        return 33;
+    }
+
+    /**
+     * Get the data protector object
+     */
+    virtual std::unique_ptr<DataProtector> getDataProtector();
+
+    /**
+     * Get an implementation specific path suffix to tag files with
+     */
+    virtual boost::filesystem::path getProtectedPathSuffix();
+
+    /**
+     * Transform temp data to non-readable form before writing it to disk.
+     */
+    virtual Status protectTmpData(
+        const uint8_t* in, size_t inLen, uint8_t* out, size_t outLen, size_t* resultLen);
+
+    /**
+     * Tranforms temp data back to readable form, after reading from disk.
+     */
+    virtual Status unprotectTmpData(
+        const uint8_t* in, size_t inLen, uint8_t* out, size_t outLen, size_t* resultLen);
+};
+
+}  // namespace mongo
diff --git a/src/mongo/db/storage/ephemeral_for_test/ephemeral_for_test_record_store.cpp b/src/mongo/db/storage/ephemeral_for_test/ephemeral_for_test_record_store.cpp
index a537a9c63c6..8c816d39f31 100644
--- a/src/mongo/db/storage/ephemeral_for_test/ephemeral_for_test_record_store.cpp
+++ b/src/mongo/db/storage/ephemeral_for_test/ephemeral_for_test_record_store.cpp
@@ -556,8 +556,15 @@ void EphemeralForTestRecordStore::temp_cappedTruncateAfter(OperationContext* txn
     Records::iterator it =
         inclusive ? _data->records.lower_bound(end) : _data->records.upper_bound(end);
     while (it != _data->records.end()) {
-        txn->recoveryUnit()->registerChange(new RemoveChange(txn, _data, it->first, it->second));
-        _data->dataSize -= it->second.size;
+        RecordId id = it->first;
+        EphemeralForTestRecord record = it->second;
+
+        if (_cappedCallback) {
+            uassertStatusOK(_cappedCallback->aboutToDeleteCapped(txn, id, record.toRecordData()));
+        }
+
+        txn->recoveryUnit()->registerChange(new RemoveChange(txn, _data, id, record));
+        _data->dataSize -= record.size;
         _data->records.erase(it++);
     }
 }
diff --git a/src/mongo/db/storage/key_string.cpp b/src/mongo/db/storage/key_string.cpp
index 624c87a2404..fc580ca111d 100644
--- a/src/mongo/db/storage/key_string.cpp
+++ b/src/mongo/db/storage/key_string.cpp
@@ -1366,7 +1366,8 @@ void toBsonValue(uint8_t ctype,
                 Decimal128 dec(Decimal128::Value{lowbits, highbits});
                 if (isNegative)
                     dec = dec.negate();
-                dec = adjustDecimalExponent(typeBits, dec);
+                if (dec.isFinite())
+                    dec = adjustDecimalExponent(typeBits, dec);
                 *stream << dec;
                 break;
             }
@@ -1385,7 +1386,8 @@ void toBsonValue(uint8_t ctype,
                 Decimal128 dec(bin, Decimal128::kRoundTo34Digits, roundAwayFromZero);
                 if (hasDecimalContinuation)
                     dec = readDecimalContinuation(reader, inverted, dec);
-                dec = adjustDecimalExponent(typeBits, dec);
+                if (dec.isFinite())
+                    dec = adjustDecimalExponent(typeBits, dec);
                 *stream << dec;
             }
             break;
diff --git a/src/mongo/db/storage/key_string_test.cpp b/src/mongo/db/storage/key_string_test.cpp
index b29d1b01da8..46a11c2592a 100644
--- a/src/mongo/db/storage/key_string_test.cpp
+++ b/src/mongo/db/storage/key_string_test.cpp
@@ -596,6 +596,15 @@ const std::vector<BSONObj>& getInterestingElements(KeyString::Version version) {
 
     // Something that needs multiple bytes of typeBits
     elements.push_back(BSON("" << BSON_ARRAY("" << BSONSymbol("") << 0 << 0ll << 0.0 << -0.0)));
+    if (version != KeyString::Version::V0) {
+        // Something with exceptional typeBits for Decimal
+        elements.push_back(
+            BSON("" << BSON_ARRAY("" << BSONSymbol("") << Decimal128::kNegativeInfinity
+                                     << Decimal128::kPositiveInfinity
+                                     << Decimal128::kPositiveNaN
+                                     << Decimal128("0.0000000")
+                                     << Decimal128("-0E1000"))));
+    }
 
     //
     // Interesting numeric cases
@@ -605,6 +614,11 @@ const std::vector<BSONObj>& getInterestingElements(KeyString::Version version) {
     elements.push_back(BSON("" << 0ll));
     elements.push_back(BSON("" << 0.0));
     elements.push_back(BSON("" << -0.0));
+    if (version != KeyString::Version::V0) {
+        Decimal128("0.0.0000000");
+        Decimal128("-0E1000");
+    }
+
     elements.push_back(BSON("" << std::numeric_limits<double>::quiet_NaN()));
     elements.push_back(BSON("" << std::numeric_limits<double>::infinity()));
     elements.push_back(BSON("" << -std::numeric_limits<double>::infinity()));
@@ -740,6 +754,11 @@ const std::vector<BSONObj>& getInterestingElements(KeyString::Version version) {
         elements.push_back(BSON("" << Decimal128("4.940656458412465441765687928682214E-324")));
         elements.push_back(BSON("" << Decimal128("-4.940656458412465441765687928682214E-324")));
         elements.push_back(BSON("" << Decimal128("-4.940656458412465441765687928682213E-324")));
+
+        // Non-finite values. Note: can't roundtrip negative NaNs, so not testing here.
+        elements.push_back(BSON("" << Decimal128::kPositiveNaN));
+        elements.push_back(BSON("" << Decimal128::kNegativeInfinity));
+        elements.push_back(BSON("" << Decimal128::kPositiveInfinity));
     }
 
     // Tricky double precision number for binary/decimal conversion: very close to a decimal
@@ -823,8 +842,36 @@ void testPermutation(KeyString::Version version,
     }
 }
 
+namespace {
+std::random_device rd;
+std::mt19937_64 seedGen(rd());
+
+// To be used by perf test for seeding, so that the entire test is repeatable in case of error.
+unsigned newSeed() {
+    unsigned int seed = seedGen();  // Replace by the reported number to repeat test execution.
+    log() << "Initializing random number generator using seed " << seed;
+    return seed;
+};
+
+std::vector<BSONObj> thinElements(std::vector<BSONObj> elements,
+                                  unsigned seed,
+                                  size_t maxElements) {
+    std::mt19937_64 gen(seed);
+
+    if (elements.size() <= maxElements)
+        return elements;
+
+    log() << "only keeping " << maxElements << " of " << elements.size()
+          << " elements using random selection";
+    std::shuffle(elements.begin(), elements.end(), gen);
+    elements.resize(maxElements);
+    return elements;
+}
+}  // namespace
+
+
 TEST_F(KeyStringTest, AllPermCompare) {
-    const std::vector<BSONObj>& elements = getInterestingElements(version);
+    std::vector<BSONObj> elements = getInterestingElements(version);
 
     for (size_t i = 0; i < elements.size(); i++) {
         const BSONObj& o = elements[i];
@@ -839,19 +886,23 @@ TEST_F(KeyStringTest, AllPermCompare) {
 }
 
 TEST_F(KeyStringTest, AllPerm2Compare) {
-#if !defined(MONGO_CONFIG_OPTIMIZED_BUILD)
-    log() << "\t\t\tskipping permutation testing on non-optimized build";
-    return;
-#endif
+    std::vector<BSONObj> baseElements = getInterestingElements(version);
+    auto seed = newSeed();
 
-    const std::vector<BSONObj>& baseElements = getInterestingElements(version);
+    // Select only a small subset of elements, as the combination is quadratic.
+    // We want to select two subsets independently, so all combinations will get tested eventually.
+    // kMaxPermElements is the desired number of elements to pass to testPermutation.
+    const size_t kMaxPermElements = kDebugBuild ? 100000 : 500000;
+    size_t maxElements = sqrt(kMaxPermElements);
+    auto firstElements = thinElements(baseElements, seed, maxElements);
+    auto secondElements = thinElements(baseElements, seed + 1, maxElements);
 
     std::vector<BSONObj> elements;
-    for (size_t i = 0; i < baseElements.size(); i++) {
-        for (size_t j = 0; j < baseElements.size(); j++) {
+    for (size_t i = 0; i < firstElements.size(); i++) {
+        for (size_t j = 0; j < secondElements.size(); j++) {
             BSONObjBuilder b;
-            b.appendElements(baseElements[i]);
-            b.appendElements(baseElements[j]);
+            b.appendElements(firstElements[i]);
+            b.appendElements(secondElements[j]);
             BSONObj o = b.obj();
             elements.push_back(o);
         }
@@ -927,6 +978,27 @@ TEST_F(KeyStringTest, NaNs) {
     ASSERT(std::isnan(toBson(ks2a, ONE_ASCENDING)[""].Double()));
     ASSERT(std::isnan(toBson(ks1d, ONE_DESCENDING)[""].Double()));
     ASSERT(std::isnan(toBson(ks2d, ONE_DESCENDING)[""].Double()));
+
+    if (version == KeyString::Version::V0)
+        return;
+
+    const auto nan3 = Decimal128::kPositiveNaN;
+    const auto nan4 = Decimal128::kNegativeNaN;
+    // Since we only output a single NaN, we can only do ROUNDTRIP testing for nan1.
+    ROUNDTRIP(version, BSON("" << nan3));
+    const KeyString ks3a(version, BSON("" << nan3), ONE_ASCENDING);
+    const KeyString ks3d(version, BSON("" << nan3), ONE_DESCENDING);
+
+    const KeyString ks4a(version, BSON("" << nan4), ONE_ASCENDING);
+    const KeyString ks4d(version, BSON("" << nan4), ONE_DESCENDING);
+
+    ASSERT_EQ(ks1a, ks4a);
+    ASSERT_EQ(ks1d, ks4d);
+
+    ASSERT(toBson(ks3a, ONE_ASCENDING)[""].Decimal().isNaN());
+    ASSERT(toBson(ks4a, ONE_ASCENDING)[""].Decimal().isNaN());
+    ASSERT(toBson(ks3d, ONE_DESCENDING)[""].Decimal().isNaN());
+    ASSERT(toBson(ks4d, ONE_DESCENDING)[""].Decimal().isNaN());
 }
 TEST_F(KeyStringTest, NumberOrderLots) {
     std::vector<BSONObj> numbers;
@@ -1074,16 +1146,6 @@ const uint64_t kMinPerfMicros = 20 * 1000;
 const uint64_t kMinPerfSamples = 50 * 1000;
 typedef std::vector<BSONObj> Numbers;
 
-std::random_device rd;
-std::mt19937 seedGen(rd());
-
-// To be used by perf test for seeding, so that the entire test is repeatable in case of error.
-unsigned newSeed() {
-    unsigned int seed = seedGen();  // Replace by the reported number to repeat test execution.
-    log() << "Initializing random number generator using seed " << seed;
-    return seed;
-};
-
 /**
  * Evaluates ROUNDTRIP on all items in Numbers a sufficient number of times to take at least
  * kMinPerfMicros microseconds. Logs the elapsed time per ROUNDTRIP evaluation.
diff --git a/src/mongo/db/storage/mmap_v1/mmap_v1_engine.cpp b/src/mongo/db/storage/mmap_v1/mmap_v1_engine.cpp
index e185afe03c7..71b3ada3ed4 100644
--- a/src/mongo/db/storage/mmap_v1/mmap_v1_engine.cpp
+++ b/src/mongo/db/storage/mmap_v1/mmap_v1_engine.cpp
@@ -36,6 +36,10 @@
 #include <boost/filesystem/path.hpp>
 #include <fstream>
 
+#ifdef __linux__
+#include <sys/sysmacros.h>
+#endif
+
 #include "mongo/db/mongod_options.h"
 #include "mongo/db/storage/mmap_v1/data_file_sync.h"
 #include "mongo/db/storage/mmap_v1/dur.h"
diff --git a/src/mongo/db/storage/wiredtiger/SConscript b/src/mongo/db/storage/wiredtiger/SConscript
index 9d2a8fc8035..42eb3b4d41d 100644
--- a/src/mongo/db/storage/wiredtiger/SConscript
+++ b/src/mongo/db/storage/wiredtiger/SConscript
@@ -82,6 +82,7 @@ if wiredtiger:
             ],
         LIBDEPS=['storage_wiredtiger_core',
                  'storage_wiredtiger_customization_hooks',
+                 '$BUILD_DIR/mongo/db/concurrency/lock_manager',
                  '$BUILD_DIR/mongo/db/storage/kv/kv_engine',
                  '$BUILD_DIR/mongo/db/storage/storage_engine_lock_file',
                  '$BUILD_DIR/mongo/db/storage/storage_engine_metadata',
diff --git a/src/mongo/db/storage/wiredtiger/wiredtiger_customization_hooks.cpp b/src/mongo/db/storage/wiredtiger/wiredtiger_customization_hooks.cpp
index d14b33de5f7..e40d3a58edb 100644
--- a/src/mongo/db/storage/wiredtiger/wiredtiger_customization_hooks.cpp
+++ b/src/mongo/db/storage/wiredtiger/wiredtiger_customization_hooks.cpp
@@ -31,12 +31,9 @@
 
 #include "mongo/db/storage/wiredtiger/wiredtiger_customization_hooks.h"
 
-#include <boost/filesystem/path.hpp>
-
 #include "mongo/base/init.h"
 #include "mongo/base/string_data.h"
 #include "mongo/db/service_context.h"
-#include "mongo/db/storage/data_protector.h"
 #include "mongo/stdx/memory.h"
 
 namespace mongo {
@@ -44,7 +41,7 @@ namespace mongo {
 /* Make a WiredTigerCustomizationHooks pointer a decoration on the global ServiceContext */
 MONGO_INITIALIZER_WITH_PREREQUISITES(SetWiredTigerCustomizationHooks, ("SetGlobalEnvironment"))
 (InitializerContext* context) {
-    auto customizationHooks = stdx::make_unique<EmptyWiredTigerCustomizationHooks>();
+    auto customizationHooks = stdx::make_unique<WiredTigerCustomizationHooks>();
     WiredTigerCustomizationHooks::set(getGlobalServiceContext(), std::move(customizationHooks));
 
     return Status::OK();
@@ -66,37 +63,14 @@ WiredTigerCustomizationHooks* WiredTigerCustomizationHooks::get(ServiceContext*
     return getCustomizationHooks(service).get();
 }
 
-EmptyWiredTigerCustomizationHooks::~EmptyWiredTigerCustomizationHooks() {}
-
-bool EmptyWiredTigerCustomizationHooks::enabled() const {
-    return false;
-}
+WiredTigerCustomizationHooks::~WiredTigerCustomizationHooks() {}
 
-bool EmptyWiredTigerCustomizationHooks::restartRequired() {
+bool WiredTigerCustomizationHooks::enabled() const {
     return false;
 }
 
-std::string EmptyWiredTigerCustomizationHooks::getTableCreateConfig(StringData tableName) {
-    return "";
-}
-
-std::unique_ptr<DataProtector> EmptyWiredTigerCustomizationHooks::getDataProtector() {
-    return std::unique_ptr<DataProtector>();
-}
-
-boost::filesystem::path EmptyWiredTigerCustomizationHooks::getProtectedPathSuffix() {
+std::string WiredTigerCustomizationHooks::getTableCreateConfig(StringData tableName) {
     return "";
 }
 
-Status EmptyWiredTigerCustomizationHooks::protectTmpData(
-    const uint8_t* in, size_t inLen, uint8_t* out, size_t outLen, size_t* resultLen) {
-    return Status(ErrorCodes::InternalError,
-                  "Customization hooks must be enabled to use preprocessTmpData.");
-}
-
-Status EmptyWiredTigerCustomizationHooks::unprotectTmpData(
-    const uint8_t* in, size_t inLen, uint8_t* out, size_t outLen, size_t* resultLen) {
-    return Status(ErrorCodes::InternalError,
-                  "Customization hooks must be enabled to use postprocessTmpData.");
-}
 }  // namespace mongo
diff --git a/src/mongo/db/storage/wiredtiger/wiredtiger_customization_hooks.h b/src/mongo/db/storage/wiredtiger/wiredtiger_customization_hooks.h
index e78995481bd..1ff86a8799e 100644
--- a/src/mongo/db/storage/wiredtiger/wiredtiger_customization_hooks.h
+++ b/src/mongo/db/storage/wiredtiger/wiredtiger_customization_hooks.h
@@ -32,20 +32,11 @@
 #include <memory>
 #include <string>
 
-#include "mongo/base/disallow_copying.h"
-#include "mongo/db/jsobj.h"
-
-namespace boost {
-namespace filesystem {
-class path;
-}  // namespace filesystem
-}  // namespace boost
-
 namespace mongo {
-class DataProtector;
 class StringData;
 class ServiceContext;
 
+// Interface and default implementation for WiredTiger customization hooks
 class WiredTigerCustomizationHooks {
 public:
     static void set(ServiceContext* service,
@@ -53,76 +44,18 @@ public:
 
     static WiredTigerCustomizationHooks* get(ServiceContext* service);
 
-    virtual ~WiredTigerCustomizationHooks() = default;
+    virtual ~WiredTigerCustomizationHooks();
 
     /**
      * Returns true if the customization hooks are enabled.
      */
-    virtual bool enabled() const = 0;
-
-    /**
-     * Perform any encryption engine initialization/sanity checking that needs to happen after
-     * storage engine initialization but before the server starts accepting incoming connections.
-     *
-     * Returns true if the server needs to be rebooted because of configuration changes.
-     */
-    virtual bool restartRequired() = 0;
+    virtual bool enabled() const;
 
     /**
      *  Gets an additional configuration string for the provided table name on a
      *  `WT_SESSION::create` call.
      */
-    virtual std::string getTableCreateConfig(StringData tableName) = 0;
-
-    /**
-     * Returns the maximum size addition when doing transforming temp data.
-     */
-    size_t additionalBytesForProtectedBuffer() {
-        return 33;
-    }
-
-    /**
-     * Get the data protector object
-     */
-    virtual std::unique_ptr<DataProtector> getDataProtector() = 0;
-
-    /**
-     * Get an implementation specific path suffix to tag files with
-     */
-    virtual boost::filesystem::path getProtectedPathSuffix() = 0;
-
-    /**
-     * Transform temp data to non-readable form before writing it to disk.
-     */
-    virtual Status protectTmpData(
-        const uint8_t* in, size_t inLen, uint8_t* out, size_t outLen, size_t* resultLen) = 0;
-
-    /**
-     * Tranforms temp data back to readable form, after reading from disk.
-     */
-    virtual Status unprotectTmpData(
-        const uint8_t* in, size_t inLen, uint8_t* out, size_t outLen, size_t* resultLen) = 0;
+    virtual std::string getTableCreateConfig(StringData tableName);
 };
 
-// Empty default implementation of the abstract class WiredTigerCustomizationHooks
-class EmptyWiredTigerCustomizationHooks : public WiredTigerCustomizationHooks {
-public:
-    ~EmptyWiredTigerCustomizationHooks() override;
-
-    bool enabled() const override;
-
-    bool restartRequired() override;
-
-    std::string getTableCreateConfig(StringData tableName) override;
-
-    std::unique_ptr<DataProtector> getDataProtector() override;
-
-    boost::filesystem::path getProtectedPathSuffix() override;
-
-    Status protectTmpData(
-        const uint8_t* in, size_t inLen, uint8_t* out, size_t outLen, size_t* resultLen) override;
-
-    Status unprotectTmpData(
-        const uint8_t* in, size_t inLen, uint8_t* out, size_t outLen, size_t* resultLen) override;
-};
 }  // namespace mongo
diff --git a/src/mongo/db/storage/wiredtiger/wiredtiger_index.cpp b/src/mongo/db/storage/wiredtiger/wiredtiger_index.cpp
index 6331594a57c..ecf3c183996 100644
--- a/src/mongo/db/storage/wiredtiger/wiredtiger_index.cpp
+++ b/src/mongo/db/storage/wiredtiger/wiredtiger_index.cpp
@@ -1102,6 +1102,7 @@ void WiredTigerIndexUnique::_unindex(WT_CURSOR* c,
                 triggerWriteConflictAtPoint(c);
                 return;
             }
+            invariantWTOK(ret);
             WT_ITEM value;
             invariantWTOK(c->get_value(c, &value));
             BufReader br(value.data, value.size);
diff --git a/src/mongo/db/storage/wiredtiger/wiredtiger_kv_engine.cpp b/src/mongo/db/storage/wiredtiger/wiredtiger_kv_engine.cpp
index f48fff37b0b..aafb52b30c0 100644
--- a/src/mongo/db/storage/wiredtiger/wiredtiger_kv_engine.cpp
+++ b/src/mongo/db/storage/wiredtiger/wiredtiger_kv_engine.cpp
@@ -228,6 +228,7 @@ WiredTigerKVEngine::WiredTigerKVEngine(const std::string& canonicalName,
         ss << "checkpoint=(wait=" << wiredTigerGlobalOptions.checkpointDelaySecs;
         ss << ",log_size=2GB),";
         ss << "statistics_log=(wait=" << wiredTigerGlobalOptions.statisticsLogDelaySecs << "),";
+        ss << "verbose=(recovery_progress),";
     }
     ss << WiredTigerCustomizationHooks::get(getGlobalServiceContext())
               ->getTableCreateConfig("system");
@@ -253,6 +254,13 @@ WiredTigerKVEngine::WiredTigerKVEngine(const std::string& canonicalName,
                 msgassertedNoTrace(28718, s.reason());
             }
             invariantWTOK(_conn->close(_conn, NULL));
+            // After successful recovery, remove the journal directory.
+            try {
+                boost::filesystem::remove_all(journalPath);
+            } catch (std::exception& e) {
+                error() << "error removing journal dir " << journalPath.string() << ' ' << e.what();
+                throw;
+            }
         }
         // This setting overrides the earlier setting because it is later in the config string.
         ss << ",log=(enabled=false),";
diff --git a/src/mongo/db/storage/wiredtiger/wiredtiger_record_store.cpp b/src/mongo/db/storage/wiredtiger/wiredtiger_record_store.cpp
index 0d3cf93a307..7761857bf90 100644
--- a/src/mongo/db/storage/wiredtiger/wiredtiger_record_store.cpp
+++ b/src/mongo/db/storage/wiredtiger/wiredtiger_record_store.cpp
@@ -909,7 +909,7 @@ int64_t WiredTigerRecordStore::storageSize(OperationContext* txn,
     if (_isEphemeral) {
         return dataSize(txn);
     }
-    WiredTigerSession* session = WiredTigerRecoveryUnit::get(txn)->getSession(txn);
+    WiredTigerSession* session = WiredTigerRecoveryUnit::get(txn)->getSessionNoTxn(txn);
     StatusWith<int64_t> result =
         WiredTigerUtil::getStatisticsValueAs<int64_t>(session->getSession(),
                                                       "statistics:" + getURI(),
@@ -1194,7 +1194,9 @@ bool WiredTigerRecordStore::yieldAndAwaitOplogDeletionRequest(OperationContext*
 
     // The top-level locks were freed, so also release any potential low-level (storage engine)
     // locks that might be held.
-    txn->recoveryUnit()->abandonSnapshot();
+    WiredTigerRecoveryUnit* recoveryUnit = (WiredTigerRecoveryUnit*)txn->recoveryUnit();
+    recoveryUnit->abandonSnapshot();
+    recoveryUnit->beginIdle();
 
     // Wait for an oplog deletion request, or for this record store to have been destroyed.
     oplogStones->awaitHasExcessStonesOrDead();
@@ -1219,15 +1221,22 @@ void WiredTigerRecordStore::reclaimOplog(OperationContext* txn) {
         try {
             WriteUnitOfWork wuow(txn);
 
-            WiredTigerCursor startwrap(_uri, _tableId, true, txn);
-            WT_CURSOR* start = startwrap.get();
-            start->set_key(start, _makeKey(_oplogStones->firstRecord));
+            WiredTigerCursor cwrap(_uri, _tableId, true, txn);
+            WT_CURSOR* cursor = cwrap.get();
 
-            WiredTigerCursor endwrap(_uri, _tableId, true, txn);
-            WT_CURSOR* end = endwrap.get();
-            end->set_key(end, _makeKey(stone->lastRecord));
+            // The first record in the oplog should be within the truncate range.
+            int ret = WT_READ_CHECK(cursor->next(cursor));
+            invariantWTOK(ret);
+            int64_t key;
+            invariantWTOK(cursor->get_key(cursor, &key));
+            RecordId firstRecord = _fromKey(key);
+            if (firstRecord < _oplogStones->firstRecord || firstRecord > stone->lastRecord) {
+                warning() << "First oplog record " << firstRecord << " is not in truncation range ("
+                          << _oplogStones->firstRecord << ", " << stone->lastRecord << ")";
+            }
 
-            invariantWTOK(session->truncate(session, nullptr, start, end, nullptr));
+            cursor->set_key(cursor, _makeKey(stone->lastRecord));
+            invariantWTOK(session->truncate(session, nullptr, nullptr, cursor, nullptr));
             _changeNumRecords(txn, -stone->records);
             _increaseDataSize(txn, -stone->bytes);
 
@@ -1619,7 +1628,7 @@ void WiredTigerRecordStore::appendCustomStats(OperationContext* txn,
         result->appendIntOrLL("sleepCount", _cappedSleep.load());
         result->appendIntOrLL("sleepMS", _cappedSleepMS.load());
     }
-    WiredTigerSession* session = WiredTigerRecoveryUnit::get(txn)->getSession(txn);
+    WiredTigerSession* session = WiredTigerRecoveryUnit::get(txn)->getSessionNoTxn(txn);
     WT_SESSION* s = session->getSession();
     BSONObjBuilder bob(result->subobjStart(_engineName));
     {
diff --git a/src/mongo/db/storage/wiredtiger/wiredtiger_recovery_unit.cpp b/src/mongo/db/storage/wiredtiger/wiredtiger_recovery_unit.cpp
index 66fd5e0ddfa..fe35bd07651 100644
--- a/src/mongo/db/storage/wiredtiger/wiredtiger_recovery_unit.cpp
+++ b/src/mongo/db/storage/wiredtiger/wiredtiger_recovery_unit.cpp
@@ -43,6 +43,7 @@
 #include "mongo/util/concurrency/ticketholder.h"
 #include "mongo/util/log.h"
 #include "mongo/util/mongoutils/str.h"
+#include "mongo/util/scopeguard.h"
 #include "mongo/util/stacktrace.h"
 
 namespace mongo {
@@ -172,7 +173,12 @@ WiredTigerSession* WiredTigerRecoveryUnit::getSession(OperationContext* opCtx) {
 
 WiredTigerSession* WiredTigerRecoveryUnit::getSessionNoTxn(OperationContext* opCtx) {
     _ensureSession();
-    return _session.get();
+    WiredTigerSession* session = _session.get();
+
+    // Dropping the queued idents might block session, which is not desired for fastpath workflow
+    // like FTDC thread. Disable dropping of queued idents for such sessions.
+    session->dropQueuedIdentsAtSessionEndAllowed(false);
+    return session;
 }
 
 void WiredTigerRecoveryUnit::abandonSnapshot() {
@@ -249,6 +255,13 @@ void WiredTigerRecoveryUnit::_txnOpen(OperationContext* opCtx) {
     _active = true;
 }
 
+void WiredTigerRecoveryUnit::beginIdle() {
+    // Close all cursors, we don't want to keep any old cached cursors around.
+    if (_session) {
+        _session->closeAllCursors("");
+    }
+}
+
 // ---------------------
 
 WiredTigerCursor::WiredTigerCursor(const std::string& uri,
diff --git a/src/mongo/db/storage/wiredtiger/wiredtiger_recovery_unit.h b/src/mongo/db/storage/wiredtiger/wiredtiger_recovery_unit.h
index 86d9eece13d..695ad331240 100644
--- a/src/mongo/db/storage/wiredtiger/wiredtiger_recovery_unit.h
+++ b/src/mongo/db/storage/wiredtiger/wiredtiger_recovery_unit.h
@@ -84,6 +84,12 @@ public:
     WiredTigerSession* getSession(OperationContext* opCtx);
 
     /**
+     * Enter a period of wait or computation during which there are no WT calls.
+     * Any non-relevant cached handles can be closed.
+     */
+    void beginIdle();
+
+    /**
      * Returns a session without starting a new WT txn on the session. Will not close any already
      * running session.
      */
diff --git a/src/mongo/db/storage/wiredtiger/wiredtiger_server_status.cpp b/src/mongo/db/storage/wiredtiger/wiredtiger_server_status.cpp
index 4967dfc2f86..3cb87349fdb 100644
--- a/src/mongo/db/storage/wiredtiger/wiredtiger_server_status.cpp
+++ b/src/mongo/db/storage/wiredtiger/wiredtiger_server_status.cpp
@@ -35,6 +35,7 @@
 
 #include "mongo/base/checked_cast.h"
 #include "mongo/bson/bsonobjbuilder.h"
+#include "mongo/db/concurrency/d_concurrency.h"
 #include "mongo/db/storage/wiredtiger/wiredtiger_kv_engine.h"
 #include "mongo/db/storage/wiredtiger/wiredtiger_record_store.h"
 #include "mongo/db/storage/wiredtiger/wiredtiger_recovery_unit.h"
@@ -56,6 +57,8 @@ bool WiredTigerServerStatusSection::includeByDefault() const {
 
 BSONObj WiredTigerServerStatusSection::generateSection(OperationContext* txn,
                                                        const BSONElement& configElement) const {
+    Lock::GlobalLock lk(txn->lockState(), LockMode::MODE_IS, UINT_MAX);
+
     // The session does not open a transaction here as one is not needed and opening one would
     // mean that execution could become blocked when a new transaction cannot be allocated
     // immediately.
diff --git a/src/mongo/db/storage/wiredtiger/wiredtiger_session_cache.cpp b/src/mongo/db/storage/wiredtiger/wiredtiger_session_cache.cpp
index 4fccd06ed6d..9df6ba94651 100644
--- a/src/mongo/db/storage/wiredtiger/wiredtiger_session_cache.cpp
+++ b/src/mongo/db/storage/wiredtiger/wiredtiger_session_cache.cpp
@@ -36,6 +36,7 @@
 #include "mongo/db/storage/wiredtiger/wiredtiger_session_cache.h"
 
 #include "mongo/base/error_codes.h"
+#include "mongo/db/server_parameters.h"
 #include "mongo/db/storage/journal_listener.h"
 #include "mongo/db/storage/wiredtiger/wiredtiger_kv_engine.h"
 #include "mongo/db/storage/wiredtiger/wiredtiger_util.h"
@@ -46,13 +47,32 @@
 
 namespace mongo {
 
+std::atomic<std::int32_t> kWiredTigerCursorCacheSize(10000);  // NOLINT
+
+class WiredTigerCursorCacheSize
+    : public ExportedServerParameter<std::int32_t, ServerParameterType::kStartupAndRuntime> {
+public:
+    WiredTigerCursorCacheSize()
+        : ExportedServerParameter<std::int32_t, ServerParameterType::kStartupAndRuntime>(
+              ServerParameterSet::getGlobal(),
+              "wiredTigerCursorCacheSize",
+              &kWiredTigerCursorCacheSize) {}
+
+    virtual Status validate(const std::int32_t& potentialNewValue) {
+        if (potentialNewValue < 0) {
+            return Status(ErrorCodes::BadValue,
+                          str::stream()
+                              << "wiredTigerCursorCacheSize must be greater than or equal "
+                              << "to 0, but attempted to set to: "
+                              << potentialNewValue);
+        }
+
+        return Status::OK();
+    }
+} WiredTigerCursorCacheSizeSetting;
+
 WiredTigerSession::WiredTigerSession(WT_CONNECTION* conn, uint64_t epoch, uint64_t cursorEpoch)
-    : _epoch(epoch),
-      _cursorEpoch(cursorEpoch),
-      _session(NULL),
-      _cursorGen(0),
-      _cursorsCached(0),
-      _cursorsOut(0) {
+    : _epoch(epoch), _cursorEpoch(cursorEpoch), _session(NULL), _cursorGen(0), _cursorsOut(0) {
     invariantWTOK(conn->open_session(conn, NULL, "isolation=snapshot", &_session));
 }
 
@@ -65,7 +85,6 @@ WiredTigerSession::WiredTigerSession(WT_CONNECTION* conn,
       _cache(cache),
       _session(NULL),
       _cursorGen(0),
-      _cursorsCached(0),
       _cursorsOut(0) {
     invariantWTOK(conn->open_session(conn, NULL, "isolation=snapshot", &_session));
 }
@@ -83,7 +102,6 @@ WT_CURSOR* WiredTigerSession::getCursor(const std::string& uri, uint64_t id, boo
             WT_CURSOR* c = i->_cursor;
             _cursors.erase(i);
             _cursorsOut++;
-            _cursorsCached--;
             return c;
         }
     }
@@ -107,17 +125,11 @@ void WiredTigerSession::releaseCursor(uint64_t id, WT_CURSOR* cursor) {
 
     // Cursors are pushed to the front of the list and removed from the back
     _cursors.push_front(WiredTigerCachedCursor(id, _cursorGen++, cursor));
-    _cursorsCached++;
-
-    // "Old" is defined as not used in the last N**2 operations, if we have N cursors cached.
-    // The reasoning here is to imagine a workload with N tables performing operations randomly
-    // across all of them (i.e., each cursor has 1/N chance of used for each operation).  We
-    // would like to cache N cursors in that case, so any given cursor could go N**2 operations
-    // in between use.
-    while (_cursorGen - _cursors.back()._gen > 10000) {
+
+    std::uint64_t cursorCacheSize = static_cast<std::uint64_t>(kWiredTigerCursorCacheSize.load());
+    while (!_cursors.empty() && _cursorGen - _cursors.back()._gen > cursorCacheSize) {
         cursor = _cursors.back()._cursor;
         _cursors.pop_back();
-        _cursorsCached--;
         invariantWTOK(cursor->close(cursor));
     }
 }
@@ -125,9 +137,10 @@ void WiredTigerSession::releaseCursor(uint64_t id, WT_CURSOR* cursor) {
 void WiredTigerSession::closeAllCursors(const std::string& uri) {
     invariant(_session);
 
+    bool all = (uri == "");
     for (auto i = _cursors.begin(); i != _cursors.end();) {
         WT_CURSOR* cursor = i->_cursor;
-        if (cursor && uri == cursor->uri) {
+        if (cursor && (all || uri == cursor->uri)) {
             invariantWTOK(cursor->close(cursor));
             i = _cursors.erase(i);
         } else
@@ -344,6 +357,11 @@ void WiredTigerSessionCache::releaseSession(WiredTigerSession* session) {
 
     bool returnedToCache = false;
     uint64_t currentEpoch = _epoch.load();
+    bool dropQueuedIdentsAtSessionEnd = session->isDropQueuedIdentsAtSessionEndAllowed();
+
+    // Reset this session's flag for dropping queued idents to default, before returning it to
+    // session cache.
+    session->dropQueuedIdentsAtSessionEndAllowed(true);
 
     if (session->_getEpoch() == currentEpoch) {  // check outside of lock to reduce contention
         stdx::lock_guard<stdx::mutex> lock(_cacheLock);
@@ -357,7 +375,7 @@ void WiredTigerSessionCache::releaseSession(WiredTigerSession* session) {
     if (!returnedToCache)
         delete session;
 
-    if (_engine && _engine->haveDropsQueued())
+    if (dropQueuedIdentsAtSessionEnd && _engine && _engine->haveDropsQueued())
         _engine->dropSomeQueuedIdents();
 }
 
diff --git a/src/mongo/db/storage/wiredtiger/wiredtiger_session_cache.h b/src/mongo/db/storage/wiredtiger/wiredtiger_session_cache.h
index e8fa9811796..34c05a3304b 100644
--- a/src/mongo/db/storage/wiredtiger/wiredtiger_session_cache.h
+++ b/src/mongo/db/storage/wiredtiger/wiredtiger_session_cache.h
@@ -99,12 +99,24 @@ public:
 
     void closeCursorsForQueuedDrops(WiredTigerKVEngine* engine);
 
+    /**
+     * Closes all cached cursors matching the uri.  If the uri is empty,
+     * all cached cursors are closed.
+     */
     void closeAllCursors(const std::string& uri);
 
     int cursorsOut() const {
         return _cursorsOut;
     }
 
+    bool isDropQueuedIdentsAtSessionEndAllowed() const {
+        return _dropQueuedIdentsAtSessionEnd;
+    }
+
+    void dropQueuedIdentsAtSessionEndAllowed(bool dropQueuedIdentsAtSessionEnd) {
+        _dropQueuedIdentsAtSessionEnd = dropQueuedIdentsAtSessionEnd;
+    }
+
     static uint64_t genTableId();
 
     /**
@@ -134,7 +146,8 @@ private:
     WT_SESSION* _session;            // owned
     CursorCache _cursors;            // owned
     uint64_t _cursorGen;
-    int _cursorsCached, _cursorsOut;
+    int _cursorsOut;
+    bool _dropQueuedIdentsAtSessionEnd = true;
 };
 
 /**
@@ -174,8 +187,8 @@ public:
     void closeCursorsForQueuedDrops();
 
     /**
-     * Closes all cached cursors and ensures that previously opened cursors will be closed on
-     * release.
+     * Closes all cached cursors matching the uri.  If the uri is empty,
+     * all cached cursors are closed.
      */
     void closeAllCursors(const std::string& uri);
 
diff --git a/src/mongo/db/storage/wiredtiger/wiredtiger_size_storer.cpp b/src/mongo/db/storage/wiredtiger/wiredtiger_size_storer.cpp
index 7783cea4cbf..bab41079a9b 100644
--- a/src/mongo/db/storage/wiredtiger/wiredtiger_size_storer.cpp
+++ b/src/mongo/db/storage/wiredtiger/wiredtiger_size_storer.cpp
@@ -202,7 +202,7 @@ void WiredTigerSizeStorer::syncCache(bool syncToDisk) {
 
     WT_SESSION* session = _session.getSession();
     invariantWTOK(session->begin_transaction(session, syncToDisk ? "sync=true" : ""));
-    ScopeGuard rollbacker = MakeGuard(session->rollback_transaction, session, "");
+    auto rollbacker = MakeGuard(session->rollback_transaction, session, "");
 
     for (Map::iterator it = myMap.begin(); it != myMap.end(); ++it) {
         string uriKey = it->first;
diff --git a/src/mongo/db/storage/wiredtiger/wiredtiger_snapshot_manager.cpp b/src/mongo/db/storage/wiredtiger/wiredtiger_snapshot_manager.cpp
index da0f618f1bd..e6000e62604 100644
--- a/src/mongo/db/storage/wiredtiger/wiredtiger_snapshot_manager.cpp
+++ b/src/mongo/db/storage/wiredtiger/wiredtiger_snapshot_manager.cpp
@@ -38,6 +38,7 @@
 #include "mongo/db/storage/wiredtiger/wiredtiger_snapshot_manager.h"
 #include "mongo/util/log.h"
 #include "mongo/util/mongoutils/str.h"
+#include "mongo/util/scopeguard.h"
 
 namespace mongo {
 
diff --git a/src/mongo/db/storage/wiredtiger/wiredtiger_util.cpp b/src/mongo/db/storage/wiredtiger/wiredtiger_util.cpp
index f00b4a33804..c0f0907f8da 100644
--- a/src/mongo/db/storage/wiredtiger/wiredtiger_util.cpp
+++ b/src/mongo/db/storage/wiredtiger/wiredtiger_util.cpp
@@ -103,9 +103,13 @@ void WiredTigerUtil::fetchTypeAndSourceURI(OperationContext* opCtx,
 
 StatusWith<std::string> WiredTigerUtil::getMetadata(OperationContext* opCtx, StringData uri) {
     invariant(opCtx);
-    WiredTigerCursor curwrap("metadata:create", WiredTigerSession::kMetadataTableId, false, opCtx);
-    WT_CURSOR* cursor = curwrap.get();
+
+    auto session = WiredTigerRecoveryUnit::get(opCtx)->getSessionNoTxn(opCtx);
+    WT_CURSOR* cursor =
+        session->getCursor("metadata:create", WiredTigerSession::kMetadataTableId, false);
     invariant(cursor);
+    ON_BLOCK_EXIT([&] { session->releaseCursor(WiredTigerSession::kMetadataTableId, cursor); });
+
     std::string strUri = uri.toString();
     cursor->set_key(cursor, strUri.c_str());
     int ret = cursor->search(cursor);
diff --git a/src/mongo/db/views/resolved_view.cpp b/src/mongo/db/views/resolved_view.cpp
index d5f4faa7f68..f497cbbb9bd 100644
--- a/src/mongo/db/views/resolved_view.cpp
+++ b/src/mongo/db/views/resolved_view.cpp
@@ -62,7 +62,27 @@ ResolvedView ResolvedView::fromBSON(BSONObj commandResponseObj) {
         pipeline.push_back(item.Obj().getOwned());
     }
 
-    return {ResolvedView(NamespaceString(viewDef["ns"].valueStringData()), pipeline)};
+    BSONObj collationSpec;
+    if (auto collationElt = viewDef["collation"]) {
+        uassert(40639,
+                "View definition 'collation' field must be an object",
+                collationElt.type() == BSONType::Object);
+        collationSpec = collationElt.embeddedObject().getOwned();
+    }
+
+    return {NamespaceString(viewDef["ns"].valueStringData()),
+            std::move(pipeline),
+            std::move(collationSpec)};
+}
+
+BSONObj ResolvedView::toBSON() const {
+    BSONObjBuilder builder;
+    builder.append("ns", _namespace.ns());
+    builder.append("pipeline", _pipeline);
+    if (!_defaultCollation.isEmpty()) {
+        builder.append("collation", _defaultCollation);
+    }
+    return builder.obj();
 }
 
 StatusWith<BSONObj> ResolvedView::asExpandedViewAggregation(
@@ -104,6 +124,13 @@ StatusWith<BSONObj> ResolvedView::asExpandedViewAggregation(
         aggregationBuilder.append("allowDiskUse", true);
     }
 
+    // Operations on a view must always use the default collation of the view. We must have already
+    // checked that if the user's request specifies a collation, it matches the collation of the
+    // view.
+    if (!_defaultCollation.isEmpty()) {
+        aggregationBuilder.append("collation", _defaultCollation);
+    }
+
     return aggregationBuilder.obj();
 }
 
diff --git a/src/mongo/db/views/resolved_view.h b/src/mongo/db/views/resolved_view.h
index a5016ffc4b3..789576c737f 100644
--- a/src/mongo/db/views/resolved_view.h
+++ b/src/mongo/db/views/resolved_view.h
@@ -44,8 +44,12 @@ class AggregationRequest;
  */
 class ResolvedView {
 public:
-    ResolvedView(const NamespaceString& collectionNs, const std::vector<BSONObj>& pipeline)
-        : _namespace(collectionNs), _pipeline(pipeline) {}
+    ResolvedView(const NamespaceString& collectionNs,
+                 std::vector<BSONObj> pipeline,
+                 BSONObj defaultCollation)
+        : _namespace(collectionNs),
+          _pipeline(std::move(pipeline)),
+          _defaultCollation(std::move(defaultCollation)) {}
 
     /**
      * Returns whether 'commandResponseObj' contains a CommandOnShardedViewNotSupportedOnMongod
@@ -55,6 +59,8 @@ public:
 
     static ResolvedView fromBSON(BSONObj commandResponseObj);
 
+    BSONObj toBSON() const;
+
     /**
      * Convert an aggregation command on a view to the equivalent command against the views
      * underlying collection.
@@ -70,9 +76,21 @@ public:
         return _pipeline;
     }
 
+    const BSONObj& getDefaultCollation() const {
+        return _defaultCollation;
+    }
+
 private:
     NamespaceString _namespace;
     std::vector<BSONObj> _pipeline;
+
+    // The default collation associated with this view. An empty object means that the default is
+    // the simple collation.
+    //
+    // Currently all operations which run over a view must use the default collation. This means
+    // that operations on the view which do not specify a collation inherit the default. Operations
+    // on the view which specify any other collation fail with a user error.
+    BSONObj _defaultCollation;
 };
 
 }  // namespace mongo
diff --git a/src/mongo/db/views/resolved_view_test.cpp b/src/mongo/db/views/resolved_view_test.cpp
index eebcda01db5..d8b95308d45 100644
--- a/src/mongo/db/views/resolved_view_test.cpp
+++ b/src/mongo/db/views/resolved_view_test.cpp
@@ -33,6 +33,7 @@
 #include "mongo/bson/bsonmisc.h"
 #include "mongo/bson/bsonobj.h"
 #include "mongo/bson/bsonobjbuilder.h"
+#include "mongo/bson/json.h"
 #include "mongo/db/namespace_string.h"
 #include "mongo/db/pipeline/aggregation_request.h"
 #include "mongo/db/views/resolved_view.h"
@@ -50,9 +51,10 @@ namespace {
 const NamespaceString viewNss("testdb.testview");
 const NamespaceString backingNss("testdb.testcoll");
 const std::vector<BSONObj> emptyPipeline;
+const BSONObj kSimpleCollation;
 
 TEST(ResolvedViewTest, ExpandingCmdObjWithEmptyPipelineOnNoOpViewYieldsEmptyPipeline) {
-    const ResolvedView resolvedView{backingNss, emptyPipeline};
+    const ResolvedView resolvedView{backingNss, emptyPipeline, kSimpleCollation};
     BSONObj cmdObj = BSON("aggregate" << viewNss.coll() << "pipeline" << BSONArray());
 
     auto result = resolvedView.asExpandedViewAggregation(cmdObj);
@@ -65,7 +67,7 @@ TEST(ResolvedViewTest, ExpandingCmdObjWithEmptyPipelineOnNoOpViewYieldsEmptyPipe
 
 TEST(ResolvedViewTest, ExpandingCmdObjWithNonemptyPipelineAppendsToViewPipeline) {
     std::vector<BSONObj> viewPipeline{BSON("skip" << 7)};
-    const ResolvedView resolvedView{backingNss, viewPipeline};
+    const ResolvedView resolvedView{backingNss, viewPipeline, kSimpleCollation};
     BSONObj cmdObj =
         BSON("aggregate" << viewNss.coll() << "pipeline" << BSON_ARRAY(BSON("limit" << 3)));
 
@@ -80,13 +82,13 @@ TEST(ResolvedViewTest, ExpandingCmdObjWithNonemptyPipelineAppendsToViewPipeline)
 }
 
 TEST(ResolvedViewTest, ExpandingCmdObjFailsIfCmdObjIsNotAValidAggregationCommand) {
-    const ResolvedView resolvedView{backingNss, emptyPipeline};
+    const ResolvedView resolvedView{backingNss, emptyPipeline, kSimpleCollation};
     BSONObj badCmdObj = BSON("invalid" << 0);
     ASSERT_NOT_OK(resolvedView.asExpandedViewAggregation(badCmdObj).getStatus());
 }
 
 TEST(ResolvedViewTest, ExpandingAggRequestWithEmptyPipelineOnNoOpViewYieldsEmptyPipeline) {
-    const ResolvedView resolvedView{backingNss, emptyPipeline};
+    const ResolvedView resolvedView{backingNss, emptyPipeline, kSimpleCollation};
     AggregationRequest aggRequest(viewNss, {});
 
     auto result = resolvedView.asExpandedViewAggregation(aggRequest);
@@ -99,7 +101,7 @@ TEST(ResolvedViewTest, ExpandingAggRequestWithEmptyPipelineOnNoOpViewYieldsEmpty
 
 TEST(ResolvedViewTest, ExpandingAggRequestWithNonemptyPipelineAppendsToViewPipeline) {
     std::vector<BSONObj> viewPipeline{BSON("skip" << 7)};
-    const ResolvedView resolvedView{backingNss, viewPipeline};
+    const ResolvedView resolvedView{backingNss, viewPipeline, kSimpleCollation};
 
     std::vector<BSONObj> userAggregationPipeline = {BSON("limit" << 3)};
     AggregationRequest aggRequest(viewNss, userAggregationPipeline);
@@ -115,7 +117,7 @@ TEST(ResolvedViewTest, ExpandingAggRequestWithNonemptyPipelineAppendsToViewPipel
 }
 
 TEST(ResolvedViewTest, ExpandingAggRequestPreservesExplain) {
-    const ResolvedView resolvedView{backingNss, emptyPipeline};
+    const ResolvedView resolvedView{backingNss, emptyPipeline, kSimpleCollation};
     AggregationRequest aggRequest(viewNss, {});
     aggRequest.setExplain(true);
 
@@ -130,7 +132,7 @@ TEST(ResolvedViewTest, ExpandingAggRequestPreservesExplain) {
 }
 
 TEST(ResolvedViewTest, ExpandingAggRequestPreservesBypassDocumentValidation) {
-    const ResolvedView resolvedView{backingNss, emptyPipeline};
+    const ResolvedView resolvedView{backingNss, emptyPipeline, kSimpleCollation};
     AggregationRequest aggRequest(viewNss, {});
     aggRequest.setBypassDocumentValidation(true);
 
@@ -145,7 +147,7 @@ TEST(ResolvedViewTest, ExpandingAggRequestPreservesBypassDocumentValidation) {
 }
 
 TEST(ResolvedViewTest, ExpandingAggRequestPreservesAllowDiskUse) {
-    const ResolvedView resolvedView{backingNss, emptyPipeline};
+    const ResolvedView resolvedView{backingNss, emptyPipeline, kSimpleCollation};
     AggregationRequest aggRequest(viewNss, {});
     aggRequest.setAllowDiskUse(true);
 
@@ -159,6 +161,27 @@ TEST(ResolvedViewTest, ExpandingAggRequestPreservesAllowDiskUse) {
     ASSERT_BSONOBJ_EQ(result.getValue(), expected);
 }
 
+TEST(ResolvedViewTest, ExpandingAggRequestPreservesDefaultCollationOfView) {
+    const ResolvedView resolvedView{backingNss,
+                                    emptyPipeline,
+                                    BSON("locale"
+                                         << "fr_CA")};
+    ASSERT_BSONOBJ_EQ(resolvedView.getDefaultCollation(),
+                      BSON("locale"
+                           << "fr_CA"));
+    AggregationRequest aggRequest(viewNss, {});
+
+    auto result = resolvedView.asExpandedViewAggregation(aggRequest);
+    ASSERT_OK(result.getStatus());
+
+    BSONObj expected =
+        BSON("aggregate" << backingNss.coll() << "pipeline" << BSONArray() << "cursor" << BSONObj()
+                         << "collation"
+                         << BSON("locale"
+                                 << "fr_CA"));
+    ASSERT_BSONOBJ_EQ(result.getValue(), expected);
+}
+
 TEST(ResolvedViewTest, FromBSONFailsIfMissingResolvedView) {
     BSONObj badCmdResponse = BSON("x" << 1);
     ASSERT_THROWS_CODE(ResolvedView::fromBSON(badCmdResponse), UserException, 40248);
@@ -190,6 +213,13 @@ TEST(ResolvedViewTest, FromBSONFailsOnInvalidPipelineType) {
     ASSERT_THROWS_CODE(ResolvedView::fromBSON(badCmdResponse), UserException, 40251);
 }
 
+TEST(ResolvedViewTest, FromBSONFailsOnInvalidCollationType) {
+    BSONObj badCmdResponse =
+        BSON("resolvedView" << BSON(
+                 "ns" << backingNss.ns() << "pipeline" << BSONArray() << "collation" << 1));
+    ASSERT_THROWS_CODE(ResolvedView::fromBSON(badCmdResponse), AssertionException, 40639);
+}
+
 TEST(ResolvedViewTest, FromBSONSuccessfullyParsesEmptyBSONArrayIntoEmptyVector) {
     BSONObj cmdResponse =
         BSON("resolvedView" << BSON("ns" << backingNss.ns() << "pipeline" << BSONArray()));
@@ -222,6 +252,22 @@ TEST(ResolvedViewTest, FromBSONSuccessfullyParsesPopulatedBSONArrayIntoVector) {
                       SimpleBSONObjComparator::kInstance.makeEqualTo()));
 }
 
+TEST(ResolvedViewTest, FromBSONSuccessfullyParsesCollation) {
+    BSONObj cmdResponse = BSON(
+        "resolvedView" << BSON("ns" << backingNss.ns() << "pipeline" << BSONArray() << "collation"
+                                    << BSON("locale"
+                                            << "fil")));
+    const ResolvedView result = ResolvedView::fromBSON(cmdResponse);
+    ASSERT_EQ(result.getNamespace(), backingNss);
+    ASSERT(std::equal(emptyPipeline.begin(),
+                      emptyPipeline.end(),
+                      result.getPipeline().begin(),
+                      SimpleBSONObjComparator::kInstance.makeEqualTo()));
+    ASSERT_BSONOBJ_EQ(result.getDefaultCollation(),
+                      BSON("locale"
+                           << "fil"));
+}
+
 TEST(ResolvedViewTest, IsResolvedViewErrorResponseDetectsKickbackErrorCodeSuccessfully) {
     BSONObj errorResponse =
         BSON("ok" << 0 << "code" << ErrorCodes::CommandOnShardedViewNotSupportedOnMongod << "errmsg"
@@ -235,5 +281,34 @@ TEST(ResolvedViewTest, IsResolvedViewErrorResponseReportsFalseOnNonKickbackError
                   << "View nesting too deep or view cycle detected");
     ASSERT_FALSE(ResolvedView::isResolvedViewErrorResponse(errorResponse));
 }
+
+TEST(ResolvedViewTest, ToBSONSerializesCorrectly) {
+    const ResolvedView resolvedView{backingNss,
+                                    std::vector<BSONObj>{BSON("$match" << BSON("x" << 1))},
+                                    BSON("locale"
+                                         << "fr_CA")};
+    auto serialized = resolvedView.toBSON();
+    ASSERT_BSONOBJ_EQ(
+        serialized,
+        fromjson(
+            "{ns: 'testdb.testcoll', pipeline: [{$match: {x: 1}}], collation: {locale: 'fr_CA'}}"));
+}
+
+TEST(ResolvedViewTest, ToBSONOutputCanBeReparsed) {
+    const ResolvedView resolvedView{backingNss,
+                                    emptyPipeline,
+                                    BSON("locale"
+                                         << "fr_CA")};
+    auto serialized = resolvedView.toBSON();
+    auto reparsedResolvedView = ResolvedView::fromBSON(BSON("resolvedView" << serialized));
+    ASSERT_EQ(reparsedResolvedView.getNamespace(), backingNss);
+    ASSERT(std::equal(emptyPipeline.begin(),
+                      emptyPipeline.end(),
+                      reparsedResolvedView.getPipeline().begin(),
+                      SimpleBSONObjComparator::kInstance.makeEqualTo()));
+    ASSERT_BSONOBJ_EQ(reparsedResolvedView.getDefaultCollation(),
+                      BSON("locale"
+                           << "fr_CA"));
+}
 }  // namespace
 }  // namespace mongo
diff --git a/src/mongo/db/views/view_catalog.cpp b/src/mongo/db/views/view_catalog.cpp
index 30cabf24935..d590c509a7e 100644
--- a/src/mongo/db/views/view_catalog.cpp
+++ b/src/mongo/db/views/view_catalog.cpp
@@ -394,6 +394,7 @@ StatusWith<ResolvedView> ViewCatalog::resolveView(OperationContext* txn,
     stdx::lock_guard<stdx::mutex> lk(_mutex);
     const NamespaceString* resolvedNss = &nss;
     std::vector<BSONObj> resolvedPipeline;
+    BSONObj collation;
 
     for (int i = 0; i < ViewGraph::kMaxViewDepth; i++) {
         auto view = _lookup_inlock(txn, resolvedNss->ns());
@@ -408,10 +409,13 @@ StatusWith<ResolvedView> ViewCatalog::resolveView(OperationContext* txn,
                         str::stream() << "View pipeline exceeds maximum size; maximum size is "
                                       << ViewGraph::kMaxViewPipelineSizeBytes};
             }
-            return StatusWith<ResolvedView>({*resolvedNss, resolvedPipeline});
+            return StatusWith<ResolvedView>(
+                {*resolvedNss, std::move(resolvedPipeline), std::move(collation)});
         }
 
         resolvedNss = &(view->viewOn());
+        collation = view->defaultCollator() ? view->defaultCollator()->getSpec().toBSON()
+                                            : CollationSpec::kSimpleSpec;
 
         // Prepend the underlying view's pipeline to the current working pipeline.
         const std::vector<BSONObj>& toPrepend = view->pipeline();
@@ -419,7 +423,8 @@ StatusWith<ResolvedView> ViewCatalog::resolveView(OperationContext* txn,
 
         // If the first stage is a $collStats, then we return early with the viewOn namespace.
         if (toPrepend.size() > 0 && !toPrepend[0]["$collStats"].eoo()) {
-            return StatusWith<ResolvedView>({*resolvedNss, resolvedPipeline});
+            return StatusWith<ResolvedView>(
+                {*resolvedNss, std::move(resolvedPipeline), std::move(collation)});
         }
     }
 
diff --git a/src/mongo/db/views/view_catalog_test.cpp b/src/mongo/db/views/view_catalog_test.cpp
index 1ca54f809df..367b4f73b0c 100644
--- a/src/mongo/db/views/view_catalog_test.cpp
+++ b/src/mongo/db/views/view_catalog_test.cpp
@@ -38,6 +38,8 @@
 #include "mongo/bson/bsonobj.h"
 #include "mongo/bson/bsonobjbuilder.h"
 #include "mongo/db/namespace_string.h"
+#include "mongo/db/operation_context.h"
+#include "mongo/db/query/collation/collator_factory_interface.h"
 #include "mongo/db/query/query_test_service_context.h"
 #include "mongo/db/server_options.h"
 #include "mongo/db/service_context_noop.h"
@@ -417,6 +419,42 @@ TEST_F(ViewCatalogFixture, ResolveViewCorrectPipeline) {
     }
 }
 
+TEST_F(ViewCatalogFixture, ResolveViewCorrectlyExtractsDefaultCollation) {
+    const NamespaceString view1("db.view1");
+    const NamespaceString view2("db.view2");
+    const NamespaceString viewOn("db.coll");
+    BSONArrayBuilder pipeline1;
+    BSONArrayBuilder pipeline2;
+
+    pipeline1 << BSON("$match" << BSON("foo" << 1));
+    pipeline2 << BSON("$match" << BSON("foo" << 2));
+
+    BSONObj collation = BSON("locale"
+                             << "mock_reverse_string");
+
+    ASSERT_OK(viewCatalog.createView(opCtx.get(), view1, viewOn, pipeline1.arr(), collation));
+    ASSERT_OK(viewCatalog.createView(opCtx.get(), view2, view1, pipeline2.arr(), collation));
+
+    auto resolvedView = viewCatalog.resolveView(opCtx.get(), view2);
+    ASSERT(resolvedView.isOK());
+
+    ASSERT_EQ(resolvedView.getValue().getNamespace(), viewOn);
+
+    std::vector<BSONObj> expected = {BSON("$match" << BSON("foo" << 1)),
+                                     BSON("$match" << BSON("foo" << 2))};
+    std::vector<BSONObj> result = resolvedView.getValue().getPipeline();
+    ASSERT_EQ(expected.size(), result.size());
+    for (uint32_t i = 0; i < expected.size(); i++) {
+        ASSERT(SimpleBSONObjComparator::kInstance.evaluate(expected[i] == result[i]));
+    }
+
+    auto expectedCollation =
+        CollatorFactoryInterface::get(opCtx->getServiceContext())->makeFromBSON(collation);
+    ASSERT_OK(expectedCollation.getStatus());
+    ASSERT_BSONOBJ_EQ(resolvedView.getValue().getDefaultCollation(),
+                      expectedCollation.getValue()->getSpec().toBSON());
+}
+
 TEST_F(ViewCatalogFixture, InvalidateThenReload) {
     const NamespaceString viewName("db.view");
     const NamespaceString viewOn("db.coll");
diff --git a/src/mongo/db/views/view_sharding_check.cpp b/src/mongo/db/views/view_sharding_check.cpp
index cc662ff2549..90835e03a07 100644
--- a/src/mongo/db/views/view_sharding_check.cpp
+++ b/src/mongo/db/views/view_sharding_check.cpp
@@ -69,11 +69,7 @@ StatusWith<BSONObj> ViewShardingCheck::getResolvedViewIfSharded(OperationContext
         return BSONObj();
     }
 
-    BSONObjBuilder viewDetailBob;
-    viewDetailBob.append("ns", sourceNss.ns());
-    viewDetailBob.append("pipeline", resolvedView.getValue().getPipeline());
-
-    return viewDetailBob.obj();
+    return resolvedView.getValue().toBSON();
 }
 
 void ViewShardingCheck::appendShardedViewStatus(const BSONObj& resolvedView, BSONObjBuilder* out) {
diff --git a/src/mongo/dbtests/dbhelper_tests.cpp b/src/mongo/dbtests/dbhelper_tests.cpp
index 1ed948092c7..7a4d745e764 100644
--- a/src/mongo/dbtests/dbhelper_tests.cpp
+++ b/src/mongo/dbtests/dbhelper_tests.cpp
@@ -74,8 +74,12 @@ public:
 
             KeyRange range(ns, BSON("_id" << _min), BSON("_id" << _max), BSON("_id" << 1));
             mongo::WriteConcernOptions dummyWriteConcern;
-            Helpers::removeRange(
-                &txn, range, BoundInclusion::kIncludeStartKeyOnly, dummyWriteConcern);
+            Milliseconds dummyReplWaitDuration;
+            Helpers::removeRange(&txn,
+                                 range,
+                                 BoundInclusion::kIncludeStartKeyOnly,
+                                 dummyWriteConcern,
+                                 dummyReplWaitDuration);
         }
 
         // Check that the expected documents remain.
diff --git a/src/mongo/dbtests/jstests.cpp b/src/mongo/dbtests/jstests.cpp
index 13e71a0a74a..c0421910c2f 100644
--- a/src/mongo/dbtests/jstests.cpp
+++ b/src/mongo/dbtests/jstests.cpp
@@ -2388,6 +2388,49 @@ public:
     }
 };
 
+class RequiresOwnedObjects {
+public:
+    void run() {
+        char buf[] = {5, 0, 0, 0, 0};
+        BSONObj unowned(buf);
+        BSONObj owned = unowned.getOwned();
+
+        ASSERT(!unowned.isOwned());
+        ASSERT(owned.isOwned());
+
+        // Ensure that by default we can bind owned and unowned
+        {
+            unique_ptr<Scope> s(getGlobalScriptEngine()->newScope());
+            s->setObject("unowned", unowned, true);
+            s->setObject("owned", owned, true);
+        }
+
+        // After we set the flag, we should only be able to set owned
+        {
+            unique_ptr<Scope> s(getGlobalScriptEngine()->newScope());
+            s->requireOwnedObjects();
+            s->setObject("owned", owned, true);
+
+            bool threwException = false;
+            try {
+                s->setObject("unowned", unowned, true);
+            } catch (...) {
+                threwException = true;
+
+                auto status = exceptionToStatus();
+
+                ASSERT_EQUALS(status.code(), ErrorCodes::BadValue);
+            }
+
+            ASSERT(threwException);
+
+            // after resetting, we can set unowned's again
+            s->reset();
+            s->setObject("unowned", unowned, true);
+        }
+    }
+};
+
 class All : public Suite {
 public:
     All() : Suite("js") {}
@@ -2444,6 +2487,7 @@ public:
 
         add<RecursiveInvoke>();
         add<ErrorCodeFromInvoke>();
+        add<RequiresOwnedObjects>();
 
         add<RoundTripTests::DBRefTest>();
         add<RoundTripTests::DBPointerTest>();
diff --git a/src/mongo/dbtests/rollbacktests.cpp b/src/mongo/dbtests/rollbacktests.cpp
index b842eb5a9bc..300d3f46aed 100644
--- a/src/mongo/dbtests/rollbacktests.cpp
+++ b/src/mongo/dbtests/rollbacktests.cpp
@@ -216,7 +216,7 @@ public:
     }
 };
 
-template <bool rollback, bool defaultIndexes>
+template <bool rollback, bool defaultIndexes, bool capped>
 class RenameCollection {
 public:
     void run() {
@@ -236,7 +236,8 @@ public:
             WriteUnitOfWork uow(&txn);
             ASSERT(!collectionExists(&ctx, source.ns()));
             ASSERT(!collectionExists(&ctx, target.ns()));
-            ASSERT_OK(userCreateNS(&txn, ctx.db(), source.ns(), BSONObj(), defaultIndexes));
+            auto options = capped ? BSON("capped" << true << "size" << 1000) : BSONObj();
+            ASSERT_OK(userCreateNS(&txn, ctx.db(), source.ns(), options, defaultIndexes));
             uow.commit();
         }
         ASSERT(collectionExists(&ctx, source.ns()));
@@ -263,7 +264,7 @@ public:
     }
 };
 
-template <bool rollback, bool defaultIndexes>
+template <bool rollback, bool defaultIndexes, bool capped>
 class RenameDropTargetCollection {
 public:
     void run() {
@@ -288,8 +289,9 @@ public:
             WriteUnitOfWork uow(&txn);
             ASSERT(!collectionExists(&ctx, source.ns()));
             ASSERT(!collectionExists(&ctx, target.ns()));
-            ASSERT_OK(userCreateNS(&txn, ctx.db(), source.ns(), BSONObj(), defaultIndexes));
-            ASSERT_OK(userCreateNS(&txn, ctx.db(), target.ns(), BSONObj(), defaultIndexes));
+            auto options = capped ? BSON("capped" << true << "size" << 1000) : BSONObj();
+            ASSERT_OK(userCreateNS(&txn, ctx.db(), source.ns(), options, defaultIndexes));
+            ASSERT_OK(userCreateNS(&txn, ctx.db(), target.ns(), options, defaultIndexes));
 
             insertRecord(&txn, source, sourceDoc);
             insertRecord(&txn, target, targetDoc);
diff --git a/src/mongo/executor/connection_pool.cpp b/src/mongo/executor/connection_pool.cpp
index cf8aadd4d3a..9bbf528725e 100644
--- a/src/mongo/executor/connection_pool.cpp
+++ b/src/mongo/executor/connection_pool.cpp
@@ -60,6 +60,45 @@ namespace executor {
  */
 class ConnectionPool::SpecificPool {
 public:
+    /**
+     * These active client methods must be used whenever entering a specific pool outside of the
+     * shutdown background task.  The presence of an active client will bump a counter on the
+     * specific pool which will prevent the shutdown thread from deleting it.
+     *
+     * The complexity comes from the need to hold a lock when writing to the
+     * _activeClients param on the specific pool.  Because the code beneath the client needs to lock
+     * and unlock the parent mutex (and can leave unlocked), we want to start the client with the
+     * lock acquired, move it into the client, then re-acquire to decrement the counter on the way
+     * out.
+     *
+     * It's used like:
+     *
+     * pool.runWithActiveClient([](stdx::unique_lock<stdx::mutex> lk){ codeToBeProtected(); });
+     */
+    template <typename Callback>
+    void runWithActiveClient(Callback&& cb) {
+        runWithActiveClient(stdx::unique_lock<stdx::mutex>(_parent->_mutex),
+                            std::forward<Callback>(cb));
+    }
+
+    template <typename Callback>
+    void runWithActiveClient(stdx::unique_lock<stdx::mutex> lk, Callback&& cb) {
+        invariant(lk.owns_lock());
+
+        _activeClients++;
+
+        const auto guard = MakeGuard([&] {
+            invariant(!lk.owns_lock());
+            stdx::lock_guard<stdx::mutex> lk(_parent->_mutex);
+            _activeClients--;
+        });
+
+        {
+            decltype(lk) localLk(std::move(lk));
+            cb(std::move(localLk));
+        }
+    }
+
     SpecificPool(ConnectionPool* parent, const HostAndPort& hostAndPort);
     ~SpecificPool();
 
@@ -149,6 +188,7 @@ private:
 
     std::unique_ptr<TimerInterface> _requestTimer;
     Date_t _requestTimerExpiration;
+    size_t _activeClients;
     size_t _generation;
     bool _inFulfillRequests;
     bool _inSpawnConnections;
@@ -206,8 +246,11 @@ void ConnectionPool::dropConnections(const HostAndPort& hostAndPort) {
     if (iter == _pools.end())
         return;
 
-    iter->second.get()->processFailure(
-        Status(ErrorCodes::PooledConnectionsDropped, "Pooled connections dropped"), std::move(lk));
+    iter->second->runWithActiveClient(std::move(lk), [&](decltype(lk) lk) {
+        iter->second->processFailure(
+            Status(ErrorCodes::PooledConnectionsDropped, "Pooled connections dropped"),
+            std::move(lk));
+    });
 }
 
 void ConnectionPool::get(const HostAndPort& hostAndPort,
@@ -229,7 +272,9 @@ void ConnectionPool::get(const HostAndPort& hostAndPort,
 
     invariant(pool);
 
-    pool->getConnection(hostAndPort, timeout, std::move(lk), std::move(cb));
+    pool->runWithActiveClient(std::move(lk), [&](decltype(lk) lk) {
+        pool->getConnection(hostAndPort, timeout, std::move(lk), std::move(cb));
+    });
 }
 
 void ConnectionPool::appendConnectionStats(ConnectionPoolStats* stats) const {
@@ -264,13 +309,16 @@ void ConnectionPool::returnConnection(ConnectionInterface* conn) {
 
     invariant(iter != _pools.end());
 
-    iter->second.get()->returnConnection(conn, std::move(lk));
+    iter->second->runWithActiveClient(std::move(lk), [&](decltype(lk) lk) {
+        iter->second->returnConnection(conn, std::move(lk));
+    });
 }
 
 ConnectionPool::SpecificPool::SpecificPool(ConnectionPool* parent, const HostAndPort& hostAndPort)
     : _parent(parent),
       _hostAndPort(hostAndPort),
       _requestTimer(parent->_factory->makeTimer()),
+      _activeClients(0),
       _generation(0),
       _inFulfillRequests(false),
       _inSpawnConnections(false),
@@ -362,47 +410,48 @@ void ConnectionPool::SpecificPool::returnConnection(ConnectionInterface* connPtr
 
         // Unlock in case refresh can occur immediately
         lk.unlock();
-        connPtr->refresh(_parent->_options.refreshTimeout,
-                         [this](ConnectionInterface* connPtr, Status status) {
-                             connPtr->indicateUsed();
-
-                             stdx::unique_lock<stdx::mutex> lk(_parent->_mutex);
-
-                             auto conn = takeFromProcessingPool(connPtr);
-
-                             // If the host and port were dropped, let this lapse
-                             if (conn->getGeneration() != _generation) {
-                                 spawnConnections(lk);
-                                 return;
-                             }
-
-                             // If we're in shutdown, we don't need refreshed connections
-                             if (_state == State::kInShutdown)
-                                 return;
-
-                             // If the connection refreshed successfully, throw it back in the ready
-                             // pool
-                             if (status.isOK()) {
-                                 addToReady(lk, std::move(conn));
-                                 spawnConnections(lk);
-                                 return;
-                             }
-
-                             // If we've exceeded the time limit, start a new connect, rather than
-                             // failing all operations.  We do this because the various callers have
-                             // their own time limit which is unrelated to our internal one.
-                             if (status.code() == ErrorCodes::NetworkInterfaceExceededTimeLimit) {
-                                 log() << "Pending connection to host " << _hostAndPort
-                                       << " did not complete within the connection timeout,"
-                                       << " retrying with a new connection;" << openConnections(lk)
-                                       << " connections to that host remain open";
-                                 spawnConnections(lk);
-                                 return;
-                             }
-
-                             // Otherwise pass the failure on through
-                             processFailure(status, std::move(lk));
-                         });
+        connPtr->refresh(
+            _parent->_options.refreshTimeout, [this](ConnectionInterface* connPtr, Status status) {
+                connPtr->indicateUsed();
+
+                runWithActiveClient([&](stdx::unique_lock<stdx::mutex> lk) {
+                    auto conn = takeFromProcessingPool(connPtr);
+
+                    // If the host and port were dropped, let this lapse
+                    if (conn->getGeneration() != _generation) {
+                        spawnConnections(lk);
+                        return;
+                    }
+
+                    // If we're in shutdown, we don't need refreshed connections
+                    if (_state == State::kInShutdown)
+                        return;
+
+                    // If the connection refreshed successfully, throw it back in
+                    // the ready pool
+                    if (status.isOK()) {
+                        addToReady(lk, std::move(conn));
+                        spawnConnections(lk);
+                        return;
+                    }
+
+                    // If we've exceeded the time limit, start a new connect,
+                    // rather than failing all operations.  We do this because the
+                    // various callers have their own time limit which is unrelated
+                    // to our internal one.
+                    if (status.code() == ErrorCodes::NetworkInterfaceExceededTimeLimit) {
+                        log() << "Pending connection to host " << _hostAndPort
+                              << " did not complete within the connection timeout,"
+                              << " retrying with a new connection;" << openConnections(lk)
+                              << " connections to that host remain open";
+                        spawnConnections(lk);
+                        return;
+                    }
+
+                    // Otherwise pass the failure on through
+                    processFailure(status, std::move(lk));
+                });
+            });
         lk.lock();
     } else {
         // If it's fine as it is, just put it in the ready queue
@@ -425,25 +474,25 @@ void ConnectionPool::SpecificPool::addToReady(stdx::unique_lock<stdx::mutex>& lk
     connPtr->setTimeout(_parent->_options.refreshRequirement, [this, connPtr]() {
         OwnedConnection conn;
 
-        stdx::unique_lock<stdx::mutex> lk(_parent->_mutex);
-
-        if (!_readyPool.count(connPtr)) {
-            // We've already been checked out. We don't need to refresh
-            // ourselves.
-            return;
-        }
+        runWithActiveClient([&](stdx::unique_lock<stdx::mutex> lk) {
+            if (!_readyPool.count(connPtr)) {
+                // We've already been checked out. We don't need to refresh
+                // ourselves.
+                return;
+            }
 
-        conn = takeFromPool(_readyPool, connPtr);
+            conn = takeFromPool(_readyPool, connPtr);
 
-        // If we're in shutdown, we don't need to refresh connections
-        if (_state == State::kInShutdown)
-            return;
+            // If we're in shutdown, we don't need to refresh connections
+            if (_state == State::kInShutdown)
+                return;
 
-        _checkedOutPool[connPtr] = std::move(conn);
+            _checkedOutPool[connPtr] = std::move(conn);
 
-        connPtr->indicateSuccess();
+            connPtr->indicateSuccess();
 
-        returnConnection(connPtr, std::move(lk));
+            returnConnection(connPtr, std::move(lk));
+        });
     });
 
     fulfillRequests(lk);
@@ -586,26 +635,26 @@ void ConnectionPool::SpecificPool::spawnConnections(stdx::unique_lock<stdx::mute
             _parent->_options.refreshTimeout, [this](ConnectionInterface* connPtr, Status status) {
                 connPtr->indicateUsed();
 
-                stdx::unique_lock<stdx::mutex> lk(_parent->_mutex);
-
-                auto conn = takeFromProcessingPool(connPtr);
-
-                if (conn->getGeneration() != _generation) {
-                    // If the host and port was dropped, let the
-                    // connection lapse
-                    spawnConnections(lk);
-                } else if (status.isOK()) {
-                    addToReady(lk, std::move(conn));
-                    spawnConnections(lk);
-                } else if (status.code() == ErrorCodes::NetworkInterfaceExceededTimeLimit) {
-                    // If we've exceeded the time limit, restart the connect, rather than
-                    // failing all operations.  We do this because the various callers
-                    // have their own time limit which is unrelated to our internal one.
-                    spawnConnections(lk);
-                } else {
-                    // If the setup failed, cascade the failure edge
-                    processFailure(status, std::move(lk));
-                }
+                runWithActiveClient([&](stdx::unique_lock<stdx::mutex> lk) {
+                    auto conn = takeFromProcessingPool(connPtr);
+
+                    if (conn->getGeneration() != _generation) {
+                        // If the host and port was dropped, let the
+                        // connection lapse
+                        spawnConnections(lk);
+                    } else if (status.isOK()) {
+                        addToReady(lk, std::move(conn));
+                        spawnConnections(lk);
+                    } else if (status.code() == ErrorCodes::NetworkInterfaceExceededTimeLimit) {
+                        // If we've exceeded the time limit, restart the connect, rather than
+                        // failing all operations.  We do this because the various callers
+                        // have their own time limit which is unrelated to our internal one.
+                        spawnConnections(lk);
+                    } else {
+                        // If the setup failed, cascade the failure edge
+                        processFailure(status, std::move(lk));
+                    }
+                });
             });
         // Note that this assumes that the refreshTimeout is sound for the
         // setupTimeout
@@ -641,7 +690,7 @@ void ConnectionPool::SpecificPool::shutdown() {
 
     // If we have processing connections, wait for them to finish or timeout
     // before shutdown
-    if (_processingPool.size() || _droppedProcessingPool.size()) {
+    if (_processingPool.size() || _droppedProcessingPool.size() || _activeClients) {
         _requestTimer->setTimeout(Seconds(1), [this]() { shutdown(); });
 
         return;
@@ -693,27 +742,27 @@ void ConnectionPool::SpecificPool::updateStateInLock() {
         // We set a timer for the most recent request, then invoke each timed
         // out request we couldn't service
         _requestTimer->setTimeout(timeout, [this]() {
-            stdx::unique_lock<stdx::mutex> lk(_parent->_mutex);
-
-            auto now = _parent->_factory->now();
-
-            while (_requests.size()) {
-                auto& x = _requests.top();
-
-                if (x.first <= now) {
-                    auto cb = std::move(x.second);
-                    _requests.pop();
-
-                    lk.unlock();
-                    cb(Status(ErrorCodes::NetworkInterfaceExceededTimeLimit,
-                              "Couldn't get a connection within the time limit"));
-                    lk.lock();
-                } else {
-                    break;
+            runWithActiveClient([&](stdx::unique_lock<stdx::mutex> lk) {
+                auto now = _parent->_factory->now();
+
+                while (_requests.size()) {
+                    auto& x = _requests.top();
+
+                    if (x.first <= now) {
+                        auto cb = std::move(x.second);
+                        _requests.pop();
+
+                        lk.unlock();
+                        cb(Status(ErrorCodes::NetworkInterfaceExceededTimeLimit,
+                                  "Couldn't get a connection within the time limit"));
+                        lk.lock();
+                    } else {
+                        break;
+                    }
                 }
-            }
 
-            updateStateInLock();
+                updateStateInLock();
+            });
         });
     } else if (_checkedOutPool.size()) {
         // If we have no requests, but someone's using a connection, we just
diff --git a/src/mongo/executor/network_interface_asio_auth.cpp b/src/mongo/executor/network_interface_asio_auth.cpp
index 1f2039dbd1e..571a1be1f81 100644
--- a/src/mongo/executor/network_interface_asio_auth.cpp
+++ b/src/mongo/executor/network_interface_asio_auth.cpp
@@ -38,6 +38,7 @@
 #include "mongo/db/auth/authorization_manager_global.h"
 #include "mongo/db/auth/internal_user_auth.h"
 #include "mongo/db/commands.h"
+#include "mongo/db/commands/feature_compatibility_version_command_parser.h"
 #include "mongo/db/server_options.h"
 #include "mongo/db/wire_version.h"
 #include "mongo/rpc/factory.h"
diff --git a/src/mongo/executor/task_executor.h b/src/mongo/executor/task_executor.h
index 2d558512f91..ebf7c447ef0 100644
--- a/src/mongo/executor/task_executor.h
+++ b/src/mongo/executor/task_executor.h
@@ -196,6 +196,8 @@ public:
     /**
      * Schedules "work" to be run by the executor no sooner than "when".
      *
+     * If "when" is <= now(), then it schedules the "work" to be run ASAP.
+     *
      * Returns a handle for waiting on or canceling the callback, or
      * ErrorCodes::ShutdownInProgress.
      *
diff --git a/src/mongo/executor/task_executor_test_common.cpp b/src/mongo/executor/task_executor_test_common.cpp
index 57c12813250..d7a384231b8 100644
--- a/src/mongo/executor/task_executor_test_common.cpp
+++ b/src/mongo/executor/task_executor_test_common.cpp
@@ -328,14 +328,22 @@ COMMON_EXECUTOR_TEST(ScheduleWorkAt) {
     Status status1 = getDetectableErrorStatus();
     Status status2 = getDetectableErrorStatus();
     Status status3 = getDetectableErrorStatus();
+    Status status4 = getDetectableErrorStatus();
+
     const Date_t now = net->now();
     const TaskExecutor::CallbackHandle cb1 = unittest::assertGet(executor.scheduleWorkAt(
         now + Milliseconds(100), stdx::bind(setStatus, stdx::placeholders::_1, &status1)));
+    const TaskExecutor::CallbackHandle cb4 = unittest::assertGet(executor.scheduleWorkAt(
+        now - Milliseconds(50), stdx::bind(setStatus, stdx::placeholders::_1, &status4)));
     unittest::assertGet(executor.scheduleWorkAt(
         now + Milliseconds(5000), stdx::bind(setStatus, stdx::placeholders::_1, &status3)));
     const TaskExecutor::CallbackHandle cb2 = unittest::assertGet(executor.scheduleWorkAt(
         now + Milliseconds(200),
         stdx::bind(setStatusAndShutdown, stdx::placeholders::_1, &status2)));
+
+    executor.wait(cb4);
+    ASSERT_OK(status4);
+
     const Date_t startTime = net->now();
     net->enterNetwork();
     net->runUntil(startTime + Milliseconds(200));
diff --git a/src/mongo/rpc/get_status_from_command_result.cpp b/src/mongo/rpc/get_status_from_command_result.cpp
index 2b4adf94aae..13c176f44df 100644
--- a/src/mongo/rpc/get_status_from_command_result.cpp
+++ b/src/mongo/rpc/get_status_from_command_result.cpp
@@ -30,6 +30,7 @@
 
 #include "mongo/rpc/get_status_from_command_result.h"
 
+#include "mongo/base/error_codes.h"
 #include "mongo/base/status.h"
 #include "mongo/bson/util/bson_extract.h"
 #include "mongo/db/jsobj.h"
@@ -40,6 +41,7 @@ namespace mongo {
 
 namespace {
 const std::string kCmdResponseWriteConcernField = "writeConcernError";
+const std::string kCmdResponseWriteErrorsField = "writeErrors";
 }  // namespace
 
 Status getStatusFromCommandResult(const BSONObj& result) {
@@ -107,4 +109,45 @@ Status getWriteConcernStatusFromCommandResult(const BSONObj& obj) {
     return wcError.toStatus();
 }
 
+Status getFirstWriteErrorStatusFromCommandResult(const BSONObj& cmdResponse) {
+    BSONElement writeErrorElem;
+    auto status = bsonExtractTypedField(
+        cmdResponse, kCmdResponseWriteErrorsField, BSONType::Array, &writeErrorElem);
+    if (!status.isOK()) {
+        if (status == ErrorCodes::NoSuchKey) {
+            return Status::OK();
+        } else {
+            return status;
+        }
+    }
+
+    auto firstWriteErrorElem = writeErrorElem.Obj().firstElement();
+    if (!firstWriteErrorElem) {
+        return Status::OK();
+    }
+
+    if (firstWriteErrorElem.type() != BSONType::Object) {
+        return Status(ErrorCodes::UnsupportedFormat,
+                      str::stream() << "writeErrors should be an array of objects, found "
+                                    << typeName(firstWriteErrorElem.type()));
+    }
+
+    auto firstWriteErrorObj = firstWriteErrorElem.Obj();
+
+    return Status(ErrorCodes::fromInt(firstWriteErrorObj["code"].Int()),
+                  firstWriteErrorObj["errmsg"].String());
+}
+
+Status getStatusFromWriteCommandReply(const BSONObj& cmdResponse) {
+    auto status = getStatusFromCommandResult(cmdResponse);
+    if (!status.isOK()) {
+        return status;
+    }
+    status = getFirstWriteErrorStatusFromCommandResult(cmdResponse);
+    if (!status.isOK()) {
+        return status;
+    }
+    return getWriteConcernStatusFromCommandResult(cmdResponse);
+}
+
 }  // namespace mongo
diff --git a/src/mongo/rpc/get_status_from_command_result.h b/src/mongo/rpc/get_status_from_command_result.h
index 971805ee088..423b656a0ed 100644
--- a/src/mongo/rpc/get_status_from_command_result.h
+++ b/src/mongo/rpc/get_status_from_command_result.h
@@ -51,4 +51,17 @@ Status getStatusFromCommandResult(const BSONObj& result);
  */
 Status getWriteConcernStatusFromCommandResult(const BSONObj& cmdResponse);
 
+
+/**
+ * Extracts the first write error from a command response and converts it into a status. This
+ * ignores all errors after the first and does not preserve the write error index, so it should not
+ * be used with bulk writes.
+ */
+Status getFirstWriteErrorStatusFromCommandResult(const BSONObj& cmdResponse);
+
+/**
+ * Extracts any type of error from a write command response.
+ */
+Status getStatusFromWriteCommandReply(const BSONObj& cmdResponse);
+
 }  // namespace mongo
diff --git a/src/mongo/s/catalog/type_tags.cpp b/src/mongo/s/catalog/type_tags.cpp
index 59d28cc1cc4..b0c72053cca 100644
--- a/src/mongo/s/catalog/type_tags.cpp
+++ b/src/mongo/s/catalog/type_tags.cpp
@@ -34,6 +34,7 @@
 #include "mongo/bson/bsonobj.h"
 #include "mongo/bson/bsonobjbuilder.h"
 #include "mongo/bson/util/bson_extract.h"
+#include "mongo/s/catalog/type_chunk.h"
 #include "mongo/util/assert_util.h"
 #include "mongo/util/mongoutils/str.h"
 
@@ -55,8 +56,9 @@ StatusWith<TagsType> TagsType::fromBSON(const BSONObj& source) {
     {
         std::string tagsNs;
         Status status = bsonExtractStringField(source, ns.name(), &tagsNs);
-        if (!status.isOK())
+        if (!status.isOK()) {
             return status;
+        }
 
         tags._ns = tagsNs;
     }
@@ -64,28 +66,21 @@ StatusWith<TagsType> TagsType::fromBSON(const BSONObj& source) {
     {
         std::string tagsTag;
         Status status = bsonExtractStringField(source, tag.name(), &tagsTag);
-        if (!status.isOK())
+        if (!status.isOK()) {
             return status;
+        }
 
         tags._tag = tagsTag;
     }
 
     {
-        BSONElement tagsMinKey;
-        Status status = bsonExtractTypedField(source, min.name(), Object, &tagsMinKey);
-        if (!status.isOK())
-            return status;
-
-        tags._minKey = tagsMinKey.Obj().getOwned();
-    }
-
-    {
-        BSONElement tagsMaxKey;
-        Status status = bsonExtractTypedField(source, max.name(), Object, &tagsMaxKey);
-        if (!status.isOK())
-            return status;
+        auto tagRangeStatus = ChunkRange::fromBSON(source);
+        if (!tagRangeStatus.isOK())
+            return tagRangeStatus.getStatus();
 
-        tags._maxKey = tagsMaxKey.Obj().getOwned();
+        const auto tagRange = std::move(tagRangeStatus.getValue());
+        tags._minKey = tagRange.getMin().getOwned();
+        tags._maxKey = tagRange.getMax().getOwned();
     }
 
     return tags;
diff --git a/src/mongo/s/catalog/type_tags_test.cpp b/src/mongo/s/catalog/type_tags_test.cpp
index 78af4d5fe03..16a7c6dc2ee 100644
--- a/src/mongo/s/catalog/type_tags_test.cpp
+++ b/src/mongo/s/catalog/type_tags_test.cpp
@@ -117,9 +117,8 @@ TEST(TagsType, KeysNotAscending) {
         BSON(TagsType::tag("tag") << TagsType::ns("test.mycol") << TagsType::min(BSON("a" << 20))
                                   << TagsType::max(BSON("a" << 10)));
 
-    StatusWith<TagsType> status = TagsType::fromBSON(obj);
-    const TagsType& tag = status.getValue();
-    ASSERT_EQUALS(ErrorCodes::BadValue, tag.validate());
+    StatusWith<TagsType> tagStatus = TagsType::fromBSON(obj);
+    ASSERT_EQUALS(ErrorCodes::FailedToParse, tagStatus.getStatus());
 }
 
 TEST(TagsType, BadType) {
diff --git a/src/mongo/s/client/shard_remote.cpp b/src/mongo/s/client/shard_remote.cpp
index 85e0bc7678b..1e0ea51ed4b 100644
--- a/src/mongo/s/client/shard_remote.cpp
+++ b/src/mongo/s/client/shard_remote.cpp
@@ -340,7 +340,8 @@ StatusWith<Shard::QueryResponse> ShardRemote::_exhaustiveFindOnConfig(
                     findCmdBuilder.done(),
                     fetcherCallback,
                     _appendMetadataForCommand(txn, readPrefWithMinOpTime),
-                    maxTimeMS);
+                    maxTimeMS /* find network timeout */,
+                    maxTimeMS /* getMore network timeout */);
     Status scheduleStatus = fetcher.schedule();
     if (!scheduleStatus.isOK()) {
         return scheduleStatus;
diff --git a/src/mongo/s/commands/SConscript b/src/mongo/s/commands/SConscript
index 800f8fe6005..b70eb831ac4 100644
--- a/src/mongo/s/commands/SConscript
+++ b/src/mongo/s/commands/SConscript
@@ -83,6 +83,7 @@ env.Library(
     LIBDEPS=[
         '$BUILD_DIR/mongo/db/commands/apply_ops_cmd_common',
         '$BUILD_DIR/mongo/db/commands/killcursors_common',
+        '$BUILD_DIR/mongo/db/ftdc/ftdc_server',
         '$BUILD_DIR/mongo/db/pipeline/aggregation',
         '$BUILD_DIR/mongo/db/views/views',
         '$BUILD_DIR/mongo/rpc/client_metadata',
diff --git a/src/mongo/s/commands/cluster_ftdc_commands.cpp b/src/mongo/s/commands/cluster_ftdc_commands.cpp
index 23903e92984..12021e98f6d 100644
--- a/src/mongo/s/commands/cluster_ftdc_commands.cpp
+++ b/src/mongo/s/commands/cluster_ftdc_commands.cpp
@@ -67,6 +67,28 @@ public:
     Status checkAuthForCommand(Client* client,
                                const std::string& dbname,
                                const BSONObj& cmdObj) override {
+
+        if (!AuthorizationSession::get(client)->isAuthorizedForActionsOnResource(
+                ResourcePattern::forClusterResource(), ActionType::serverStatus)) {
+            return Status(ErrorCodes::Unauthorized, "Unauthorized");
+        }
+
+        if (!AuthorizationSession::get(client)->isAuthorizedForActionsOnResource(
+                ResourcePattern::forClusterResource(), ActionType::replSetGetStatus)) {
+            return Status(ErrorCodes::Unauthorized, "Unauthorized");
+        }
+
+        if (!AuthorizationSession::get(client)->isAuthorizedForActionsOnResource(
+                ResourcePattern::forClusterResource(), ActionType::connPoolStats)) {
+            return Status(ErrorCodes::Unauthorized, "Unauthorized");
+        }
+
+        if (!AuthorizationSession::get(client)->isAuthorizedForActionsOnResource(
+                ResourcePattern::forExactNamespace(NamespaceString("local", "oplog.rs")),
+                ActionType::collStats)) {
+            return Status(ErrorCodes::Unauthorized, "Unauthorized");
+        }
+
         return Status::OK();
     }
 
@@ -77,9 +99,10 @@ public:
              std::string& errmsg,
              BSONObjBuilder& result) override {
 
-        errmsg = "getDiagnosticData not allowed through mongos";
+        result.append(
+            "data", FTDCController::get(txn->getServiceContext())->getMostRecentPeriodicDocument());
 
-        return false;
+        return true;
     }
 };
 
diff --git a/src/mongo/s/commands/cluster_shard_collection_cmd.cpp b/src/mongo/s/commands/cluster_shard_collection_cmd.cpp
index f1ba15396a9..eac3e2a24e0 100644
--- a/src/mongo/s/commands/cluster_shard_collection_cmd.cpp
+++ b/src/mongo/s/commands/cluster_shard_collection_cmd.cpp
@@ -52,6 +52,7 @@
 #include "mongo/db/write_concern_options.h"
 #include "mongo/s/balancer_configuration.h"
 #include "mongo/s/catalog/sharding_catalog_client.h"
+#include "mongo/s/catalog/type_database.h"
 #include "mongo/s/catalog_cache.h"
 #include "mongo/s/client/shard_registry.h"
 #include "mongo/s/commands/cluster_write.h"
@@ -186,12 +187,17 @@ public:
         auto const shardRegistry = Grid::get(opCtx)->shardRegistry();
         auto const catalogCache = Grid::get(opCtx)->catalogCache();
 
-        auto dbInfo = uassertStatusOK(catalogCache->getDatabase(opCtx, nss.db()));
-
-        // Ensure sharding is allowed on the database
-        uassert(ErrorCodes::IllegalOperation,
-                str::stream() << "sharding not enabled for db " << nss.db(),
-                dbInfo.shardingEnabled());
+        // Ensure sharding is allowed on the database by reading directly from the config server,
+        // because reading through the cache might produce a stale entry if the "enableSharding" was
+        // called through a different mongos
+        {
+            const auto opTimeWithDbt =
+                uassertStatusOK(catalogClient->getDatabase(opCtx, nss.db().toString()));
+            const auto& dbt = opTimeWithDbt.value;
+            uassert(ErrorCodes::IllegalOperation,
+                    str::stream() << "sharding not enabled for db " << nss.db(),
+                    dbt.getSharded());
+        }
 
         auto routingInfo = uassertStatusOK(catalogCache->getCollectionRoutingInfo(opCtx, nss));
 
diff --git a/src/mongo/s/commands/cluster_write.cpp b/src/mongo/s/commands/cluster_write.cpp
index b17035993d9..565a1ddc4d7 100644
--- a/src/mongo/s/commands/cluster_write.cpp
+++ b/src/mongo/s/commands/cluster_write.cpp
@@ -66,28 +66,6 @@ void toBatchError(const Status& status, BatchedCommandResponse* response) {
 }
 
 /**
- * Given a maxChunkSize configuration and the number of chunks in a particular sharded collection,
- * returns an optimal chunk size to use in order to achieve a good ratio between number of chunks
- * and their size.
- */
-uint64_t calculateDesiredChunkSize(uint64_t maxChunkSizeBytes, uint64_t numChunks) {
-    // Splitting faster in early chunks helps spread out an initial load better
-    const uint64_t minChunkSize = 1 << 20;  // 1 MBytes
-
-    if (numChunks <= 1) {
-        return 1024;
-    } else if (numChunks < 3) {
-        return minChunkSize / 2;
-    } else if (numChunks < 10) {
-        return std::max(maxChunkSizeBytes / 4, minChunkSize);
-    } else if (numChunks < 20) {
-        return std::max(maxChunkSizeBytes / 2, minChunkSize);
-    } else {
-        return maxChunkSizeBytes;
-    }
-}
-
-/**
  * Returns the split point that will result in one of the chunk having exactly one document. Also
  * returns an empty document if the split point cannot be determined.
  *
@@ -341,8 +319,7 @@ void updateChunkWriteStatsAndSplitIfNeeded(OperationContext* opCtx,
 
     const uint64_t chunkBytesWritten = chunk->addBytesWritten(dataWritten);
 
-    const uint64_t desiredChunkSize =
-        calculateDesiredChunkSize(balancerConfig->getMaxChunkSizeBytes(), manager->numChunks());
+    const uint64_t desiredChunkSize = balancerConfig->getMaxChunkSizeBytes();
 
     // If this chunk is at either end of the range, trigger auto-split at 10% less data written in
     // order to trigger the top-chunk optimization.
diff --git a/src/mongo/s/move_chunk_request.cpp b/src/mongo/s/move_chunk_request.cpp
index 4622f26fd29..104587ef751 100644
--- a/src/mongo/s/move_chunk_request.cpp
+++ b/src/mongo/s/move_chunk_request.cpp
@@ -195,7 +195,8 @@ bool MoveChunkRequest::operator==(const MoveChunkRequest& other) const {
         return false;
     if (_range != other._range)
         return false;
-
+    if (_waitForDelete != other._waitForDelete)
+        return false;
     return true;
 }
 
diff --git a/src/mongo/s/server.cpp b/src/mongo/s/server.cpp
index 9177cccdb73..0b8a9a17651 100644
--- a/src/mongo/s/server.cpp
+++ b/src/mongo/s/server.cpp
@@ -51,6 +51,7 @@
 #include "mongo/db/auth/user_cache_invalidator_job.h"
 #include "mongo/db/client.h"
 #include "mongo/db/dbwebserver.h"
+#include "mongo/db/ftdc/ftdc_mongos.h"
 #include "mongo/db/initialize_server_global_state.h"
 #include "mongo/db/instance.h"
 #include "mongo/db/lasterror.h"
@@ -152,6 +153,9 @@ static void cleanupTask() {
         if (auto catalog = Grid::get(txn)->catalogClient(txn)) {
             catalog->shutDown(txn);
         }
+
+        // Shutdown Full-Time Data Capture
+        stopMongoSFTDC();
     }
 
     audit::logShutdown(Client::getCurrent());
@@ -293,6 +297,8 @@ static ExitCode runMongosServer() {
         web.detach();
     }
 
+    startMongoSFTDC();
+
     Status status = getGlobalAuthorizationManager()->initialize(NULL);
     if (!status.isOK()) {
         error() << "Initializing authorization data failed: " << status;
diff --git a/src/mongo/s/sharding_initialization.cpp b/src/mongo/s/sharding_initialization.cpp
index 2bf84b02d18..1c28829a696 100644
--- a/src/mongo/s/sharding_initialization.cpp
+++ b/src/mongo/s/sharding_initialization.cpp
@@ -182,6 +182,30 @@ Status initializeGlobalShardingState(OperationContext* txn,
     connPoolOptions.refreshRequirement = Milliseconds(ShardingTaskExecutorPoolRefreshRequirementMS);
     connPoolOptions.refreshTimeout = Milliseconds(ShardingTaskExecutorPoolRefreshTimeoutMS);
 
+    if (connPoolOptions.refreshRequirement <= connPoolOptions.refreshTimeout) {
+        auto newRefreshTimeout = connPoolOptions.refreshRequirement - Milliseconds(1);
+        warning() << "ShardingTaskExecutorPoolRefreshRequirementMS ("
+                  << connPoolOptions.refreshRequirement
+                  << ") set below ShardingTaskExecutorPoolRefreshTimeoutMS ("
+                  << connPoolOptions.refreshTimeout
+                  << "). Adjusting ShardingTaskExecutorPoolRefreshTimeoutMS to "
+                  << newRefreshTimeout;
+        connPoolOptions.refreshTimeout = newRefreshTimeout;
+    }
+
+    if (connPoolOptions.hostTimeout <=
+        connPoolOptions.refreshRequirement + connPoolOptions.refreshTimeout) {
+        auto newHostTimeout =
+            connPoolOptions.refreshRequirement + connPoolOptions.refreshTimeout + Milliseconds(1);
+        warning() << "ShardingTaskExecutorPoolHostTimeoutMS (" << connPoolOptions.hostTimeout
+                  << ") set below ShardingTaskExecutorPoolRefreshRequirementMS ("
+                  << connPoolOptions.refreshRequirement
+                  << ") + ShardingTaskExecutorPoolRefreshTimeoutMS ("
+                  << connPoolOptions.refreshTimeout
+                  << "). Adjusting ShardingTaskExecutorPoolHostTimeoutMS to " << newHostTimeout;
+        connPoolOptions.hostTimeout = newHostTimeout;
+    }
+
     auto network =
         executor::makeNetworkInterface("NetworkInterfaceASIO-ShardRegistry",
                                        stdx::make_unique<ShardingNetworkConnectionHook>(),
diff --git a/src/mongo/scripting/deadline_monitor.cpp b/src/mongo/scripting/deadline_monitor.cpp
index f75c34a0cc2..5eb0f52e5de 100644
--- a/src/mongo/scripting/deadline_monitor.cpp
+++ b/src/mongo/scripting/deadline_monitor.cpp
@@ -34,7 +34,7 @@
 
 namespace mongo {
 
-MONGO_EXPORT_SERVER_PARAMETER(scriptingEngineInterruptIntervalMS, int, 0);
+MONGO_EXPORT_SERVER_PARAMETER(scriptingEngineInterruptIntervalMS, int, 1000);
 
 int getScriptingEngineInterruptInterval() {
     return scriptingEngineInterruptIntervalMS.load();
diff --git a/src/mongo/scripting/deadline_monitor.h b/src/mongo/scripting/deadline_monitor.h
index b3c5007c72d..9081d7366e6 100644
--- a/src/mongo/scripting/deadline_monitor.h
+++ b/src/mongo/scripting/deadline_monitor.h
@@ -141,10 +141,10 @@ private:
             const Date_t now = Date_t::now();
             const auto interruptInterval = Milliseconds{getScriptingEngineInterruptInterval()};
 
-            if ((interruptInterval.count() > 0) && (now - lastInterruptCycle > interruptInterval)) {
+            if (now - lastInterruptCycle > interruptInterval) {
                 for (const auto& task : _tasks) {
-                    if (task.second > now)
-                        task.first->interrupt();
+                    if (task.first->isKillPending())
+                        task.first->kill();
                 }
                 lastInterruptCycle = now;
             }
diff --git a/src/mongo/scripting/deadline_monitor_test.cpp b/src/mongo/scripting/deadline_monitor_test.cpp
index 71daefbce8f..d802673bd3d 100644
--- a/src/mongo/scripting/deadline_monitor_test.cpp
+++ b/src/mongo/scripting/deadline_monitor_test.cpp
@@ -73,8 +73,12 @@ public:
             _group->noteKill();
     }
     void interrupt() {}
+    const bool isKillPending() {
+        return killPending;
+    }
     TaskGroup* _group;
     uint64_t _killed;
+    bool killPending = false;
 };
 
 // single task expires before stopping the deadline
@@ -175,4 +179,13 @@ TEST(DeadlineMonitor, MultipleTasksExpireOrComplete) {
     }
 }
 
+TEST(DeadlineMonitor, IsKillPendingKills) {
+    DeadlineMonitor<Task> dm;
+    TaskGroup group;
+    Task task(&group);
+    dm.startDeadline(&task, -1);
+    task.killPending = true;
+    group.waitForKillCount(1);
+    ASSERT(task._killed);
+}
 }  // namespace mongo
diff --git a/src/mongo/scripting/engine.cpp b/src/mongo/scripting/engine.cpp
index 3e9a5c542cf..1848bccfba2 100644
--- a/src/mongo/scripting/engine.cpp
+++ b/src/mongo/scripting/engine.cpp
@@ -42,6 +42,7 @@
 #include "mongo/db/service_context.h"
 #include "mongo/platform/unordered_set.h"
 #include "mongo/scripting/dbdirectclient_factory.h"
+#include "mongo/util/fail_point_service.h"
 #include "mongo/util/file.h"
 #include "mongo/util/log.h"
 #include "mongo/util/text.h"
@@ -55,7 +56,10 @@ using std::unique_ptr;
 
 AtomicInt64 Scope::_lastVersion(1);
 
+
 namespace {
+
+MONGO_FP_DECLARE(mr_killop_test_fp);
 // 2 GB is the largest support Javascript file size.
 const fileofs kMaxJsFileLength = fileofs(2) * 1024 * 1024 * 1024;
 
@@ -231,6 +235,15 @@ void Scope::loadStored(OperationContext* txn, bool ignoreNotConnected) {
         uassert(10209, str::stream() << "name has to be a string: " << n, n.type() == String);
         uassert(10210, "value has to be set", v.type() != EOO);
 
+        if (MONGO_FAIL_POINT(mr_killop_test_fp)) {
+
+            /* This thread sleep makes the interrupts in the test come in at a time
+            *  where the js misses the interrupt and throw an exception instead of
+            *  being interrupted
+            */
+            stdx::this_thread::sleep_for(stdx::chrono::seconds(1));
+        }
+
         try {
             setElement(n.valuestr(), v, o);
             thisTime.insert(n.valuestr());
@@ -427,6 +440,9 @@ public:
     void advanceGeneration() {
         _real->advanceGeneration();
     }
+    void requireOwnedObjects() override {
+        _real->requireOwnedObjects();
+    }
     bool isKillPending() const {
         return _real->isKillPending();
     }
diff --git a/src/mongo/scripting/engine.h b/src/mongo/scripting/engine.h
index 2b2ab0ff14d..e773dec5f60 100644
--- a/src/mongo/scripting/engine.h
+++ b/src/mongo/scripting/engine.h
@@ -44,7 +44,7 @@ class OperationContext;
 
 struct JSFile {
     const char* name;
-    const StringData& source;
+    const StringData source;
 };
 
 class Scope {
@@ -103,6 +103,8 @@ public:
 
     virtual void advanceGeneration() = 0;
 
+    virtual void requireOwnedObjects() = 0;
+
     virtual ScriptingFunction createFunction(const char* code);
 
     /**
diff --git a/src/mongo/scripting/mozjs/PosixNSPR.cpp b/src/mongo/scripting/mozjs/PosixNSPR.cpp
index ed1a3d5a49d..9054e930443 100644
--- a/src/mongo/scripting/mozjs/PosixNSPR.cpp
+++ b/src/mongo/scripting/mozjs/PosixNSPR.cpp
@@ -24,6 +24,7 @@
 
 #include "mongo/stdx/chrono.h"
 #include "mongo/stdx/condition_variable.h"
+#include "mongo/stdx/memory.h"
 #include "mongo/stdx/mutex.h"
 #include "mongo/stdx/thread.h"
 #include "mongo/util/concurrency/thread_name.h"
@@ -98,9 +99,13 @@ PRThread* PR_CreateThread(PRThreadType type,
     MOZ_ASSERT(priority == PR_PRIORITY_NORMAL);
 
     try {
-        std::unique_ptr<nspr::Thread, void (*)(nspr::Thread*)> t(
-            js_new<nspr::Thread>(start, arg, state != PR_UNJOINABLE_THREAD),
-            js_delete_nonconst<nspr::Thread>);
+        // We can't use the nspr allocator to allocate this thread, because under asan we
+        // instrument the allocator so that asan can track the pointers correctly. This
+        // instrumentation
+        // requires that pointers be deleted in the same thread that they were allocated in.
+        // The threads created in PR_CreateThread are not always freed in the same thread
+        // that they were created in. So, we use the standard allocator here.
+        auto t = mongo::stdx::make_unique<nspr::Thread>(start, arg, state != PR_UNJOINABLE_THREAD);
 
         t->thread() = mongo::stdx::thread(&nspr::Thread::ThreadRoutine, t.get());
 
@@ -118,7 +123,7 @@ PRStatus PR_JoinThread(PRThread* thread) {
     try {
         thread->thread().join();
 
-        js_delete(thread);
+        delete thread;
 
         return PR_SUCCESS;
     } catch (...) {
diff --git a/src/mongo/scripting/mozjs/bson.cpp b/src/mongo/scripting/mozjs/bson.cpp
index 5a2ebd0dfed..a2881c44664 100644
--- a/src/mongo/scripting/mozjs/bson.cpp
+++ b/src/mongo/scripting/mozjs/bson.cpp
@@ -59,13 +59,17 @@ namespace {
  * the appearance of mutable state on the read/write versions.
  */
 struct BSONHolder {
-    BSONHolder(const BSONObj& obj, const BSONObj* parent, std::size_t generation, bool ro)
+    BSONHolder(const BSONObj& obj, const BSONObj* parent, const MozJSImplScope* scope, bool ro)
         : _obj(obj),
-          _generation(generation),
+          _generation(scope->getGeneration()),
           _isOwned(obj.isOwned() || (parent && parent->isOwned())),
           _resolved(false),
           _readOnly(ro),
           _altered(false) {
+        uassert(
+            ErrorCodes::BadValue,
+            "Attempt to bind an unowned BSON Object to a JS scope marked as requiring ownership",
+            _isOwned || (!scope->requiresOwnedObjects()));
         if (parent) {
             _parent.emplace(*parent);
         }
@@ -107,7 +111,7 @@ void BSONInfo::make(
     auto scope = getScope(cx);
 
     scope->getProto<BSONInfo>().newObject(obj);
-    JS_SetPrivate(obj, scope->trackedNew<BSONHolder>(bson, parent, scope->getGeneration(), ro));
+    JS_SetPrivate(obj, scope->trackedNew<BSONHolder>(bson, parent, scope, ro));
 }
 
 void BSONInfo::finalize(JSFreeOp* fop, JSObject* obj) {
diff --git a/src/mongo/scripting/mozjs/db.cpp b/src/mongo/scripting/mozjs/db.cpp
index 7fa2f179241..7aa73337237 100644
--- a/src/mongo/scripting/mozjs/db.cpp
+++ b/src/mongo/scripting/mozjs/db.cpp
@@ -45,38 +45,17 @@ namespace mozjs {
 
 const char* const DBInfo::className = "DB";
 
-void DBInfo::getProperty(JSContext* cx,
-                         JS::HandleObject obj,
-                         JS::HandleId id,
-                         JS::MutableHandleValue vp) {
-    // 2nd look into real values, may be cached collection object
-    if (!vp.isUndefined()) {
-        auto scope = getScope(cx);
-        auto opContext = scope->getOpContext();
-
-        if (opContext && vp.isObject()) {
-            ObjectWrapper o(cx, vp);
-
-            if (o.hasOwnField(InternedString::_fullName)) {
-                // need to check every time that the collection did not get sharded
-                if (haveLocalShardingInfo(opContext, o.getString(InternedString::_fullName)))
-                    uasserted(ErrorCodes::BadValue, "can't use sharded collection from db.eval");
-            }
-        }
+void DBInfo::resolve(JSContext* cx, JS::HandleObject obj, JS::HandleId id, bool* resolvedp) {
+    *resolvedp = false;
 
-        return;
-    }
+    JS::RootedValue coll(cx);
 
     JS::RootedObject parent(cx);
     if (!JS_GetPrototype(cx, obj, &parent))
         uasserted(ErrorCodes::JSInterpreterFailure, "Couldn't get prototype");
 
     ObjectWrapper parentWrapper(cx, parent);
-
-    if (parentWrapper.hasOwnField(id)) {
-        parentWrapper.getValue(id, vp);
-        return;
-    }
+    ObjectWrapper o(cx, obj);
 
     IdWrapper idw(cx, id);
 
@@ -87,21 +66,39 @@ void DBInfo::getProperty(JSContext* cx,
         if (sname.size() == 0 || sname[0] == '_') {
             return;
         }
+
+        // SpiderMonkey will call resolve even for __proto__ so acknowledge it exists.
+        if (sname == "__proto__"_sd) {
+            *resolvedp = true;
+            return;
+        }
+    }
+
+    // Check if this exists on the parent, ie. DBCollection::resolve case
+    if (parentWrapper.hasOwnField(id)) {
+        parentWrapper.getValue(id, &coll);
+
+        o.defineProperty(id, coll, 0);
+
+        *resolvedp = true;
+        return;
     }
 
     // no hit, create new collection
     JS::RootedValue getCollection(cx);
     parentWrapper.getValue(InternedString::getCollection, &getCollection);
 
+    // Check if getCollection has been installed yet
+    // It is undefined if the user has a db name the same as one of methods/properties of the DB
+    // object.
     if (!(getCollection.isObject() && JS_ObjectIsFunction(cx, getCollection.toObjectOrNull()))) {
-        uasserted(ErrorCodes::BadValue, "getCollection is not a function");
+        return;
     }
 
     JS::AutoValueArray<1> args(cx);
 
     idw.toValue(args[0]);
 
-    JS::RootedValue coll(cx);
     ObjectWrapper(cx, obj).callMethod(getCollection, args, &coll);
 
     uassert(16861,
@@ -111,7 +108,7 @@ void DBInfo::getProperty(JSContext* cx,
     // cache collection for reuse, don't enumerate
     ObjectWrapper(cx, obj).defineProperty(id, coll, 0);
 
-    vp.set(coll);
+    *resolvedp = true;
 }
 
 void DBInfo::construct(JSContext* cx, JS::CallArgs args) {
diff --git a/src/mongo/scripting/mozjs/db.h b/src/mongo/scripting/mozjs/db.h
index c752953236f..8f080265ef0 100644
--- a/src/mongo/scripting/mozjs/db.h
+++ b/src/mongo/scripting/mozjs/db.h
@@ -44,10 +44,7 @@ namespace mozjs {
  */
 struct DBInfo : public BaseInfo {
     static void construct(JSContext* cx, JS::CallArgs args);
-    static void getProperty(JSContext* cx,
-                            JS::HandleObject obj,
-                            JS::HandleId id,
-                            JS::MutableHandleValue vp);
+    static void resolve(JSContext* cx, JS::HandleObject obj, JS::HandleId id, bool* resolvedp);
 
     static const char* const className;
 };
diff --git a/src/mongo/scripting/mozjs/dbcollection.cpp b/src/mongo/scripting/mozjs/dbcollection.cpp
index e0f903dd1fc..b1983bd367c 100644
--- a/src/mongo/scripting/mozjs/dbcollection.cpp
+++ b/src/mongo/scripting/mozjs/dbcollection.cpp
@@ -44,11 +44,11 @@ namespace mozjs {
 
 const char* const DBCollectionInfo::className = "DBCollection";
 
-void DBCollectionInfo::getProperty(JSContext* cx,
-                                   JS::HandleObject obj,
-                                   JS::HandleId id,
-                                   JS::MutableHandleValue vp) {
-    DBInfo::getProperty(cx, obj, id, vp);
+void DBCollectionInfo::resolve(JSContext* cx,
+                               JS::HandleObject obj,
+                               JS::HandleId id,
+                               bool* resolvedp) {
+    DBInfo::resolve(cx, obj, id, resolvedp);
 }
 
 void DBCollectionInfo::construct(JSContext* cx, JS::CallArgs args) {
diff --git a/src/mongo/scripting/mozjs/dbcollection.h b/src/mongo/scripting/mozjs/dbcollection.h
index 34be422dfd4..ed3c39ff0f2 100644
--- a/src/mongo/scripting/mozjs/dbcollection.h
+++ b/src/mongo/scripting/mozjs/dbcollection.h
@@ -44,10 +44,7 @@ namespace mozjs {
  */
 struct DBCollectionInfo : public BaseInfo {
     static void construct(JSContext* cx, JS::CallArgs args);
-    static void getProperty(JSContext* cx,
-                            JS::HandleObject obj,
-                            JS::HandleId id,
-                            JS::MutableHandleValue vp);
+    static void resolve(JSContext* cx, JS::HandleObject obj, JS::HandleId id, bool* resolvedp);
 
     static const char* const className;
 };
diff --git a/src/mongo/scripting/mozjs/dbquery.cpp b/src/mongo/scripting/mozjs/dbquery.cpp
index c2543814505..15bd45bde3c 100644
--- a/src/mongo/scripting/mozjs/dbquery.cpp
+++ b/src/mongo/scripting/mozjs/dbquery.cpp
@@ -105,13 +105,8 @@ void DBQueryInfo::construct(JSContext* cx, JS::CallArgs args) {
     args.rval().setObjectOrNull(thisv);
 }
 
-void DBQueryInfo::getProperty(JSContext* cx,
-                              JS::HandleObject obj,
-                              JS::HandleId id,
-                              JS::MutableHandleValue vp) {
-    if (!vp.isUndefined()) {
-        return;
-    }
+void DBQueryInfo::resolve(JSContext* cx, JS::HandleObject obj, JS::HandleId id, bool* resolvedp) {
+    *resolvedp = false;
 
     IdWrapper wid(cx, id);
 
@@ -134,9 +129,19 @@ void DBQueryInfo::getProperty(JSContext* cx,
 
         args[0].setInt32(wid.toInt32());
 
-        ObjectWrapper(cx, obj).callMethod(arrayAccess, args, vp);
-    } else {
-        uasserted(ErrorCodes::BadValue, "arrayAccess is not a function");
+        JS::RootedValue vp(cx);
+
+        ObjectWrapper(cx, obj).callMethod(arrayAccess, args, &vp);
+
+        if (!vp.isNullOrUndefined()) {
+            ObjectWrapper o(cx, obj);
+
+            // Assumes the user won't modify the contents of what DBQuery::arrayAccess returns
+            // otherwise we need to install a getter.
+            o.defineProperty(id, vp, 0);
+        }
+
+        *resolvedp = true;
     }
 }
 
diff --git a/src/mongo/scripting/mozjs/dbquery.h b/src/mongo/scripting/mozjs/dbquery.h
index dc844c3084c..590185c55ee 100644
--- a/src/mongo/scripting/mozjs/dbquery.h
+++ b/src/mongo/scripting/mozjs/dbquery.h
@@ -41,10 +41,8 @@ namespace mozjs {
  */
 struct DBQueryInfo : public BaseInfo {
     static void construct(JSContext* cx, JS::CallArgs args);
-    static void getProperty(JSContext* cx,
-                            JS::HandleObject obj,
-                            JS::HandleId id,
-                            JS::MutableHandleValue vp);
+    static void resolve(JSContext* cx, JS::HandleObject obj, JS::HandleId id, bool* resolvedp);
+
 
     static const char* const className;
 };
diff --git a/src/mongo/scripting/mozjs/implscope.cpp b/src/mongo/scripting/mozjs/implscope.cpp
index c37ebc0cb2a..c31d5b940a4 100644
--- a/src/mongo/scripting/mozjs/implscope.cpp
+++ b/src/mongo/scripting/mozjs/implscope.cpp
@@ -258,7 +258,6 @@ MozJSImplScope::ASANHandles::ASANHandles() {
 }
 
 MozJSImplScope::ASANHandles::~ASANHandles() {
-    invariant(_handles.empty());
     invariant(kCurrentASANHandles == this);
     kCurrentASANHandles = nullptr;
 }
@@ -342,6 +341,14 @@ MozJSImplScope::MozRuntime::MozRuntime(const MozJSScriptEngine* engine) {
                 .setIon(true)
                 .setAsyncStack(false)
                 .setNativeRegExp(true);
+        } else {
+            JS::RuntimeOptionsRef(_runtime.get())
+                .setAsmJS(false)
+                .setThrowOnAsmJSValidationFailure(false)
+                .setBaseline(false)
+                .setIon(false)
+                .setAsyncStack(false)
+                .setNativeRegExp(false);
         }
 
         const StackLocator locator;
@@ -403,6 +410,7 @@ MozJSImplScope::MozJSImplScope(MozJSScriptEngine* engine)
       _connectState(ConnectState::Not),
       _status(Status::OK()),
       _generation(0),
+      _requireOwnedObjects(false),
       _hasOutOfMemoryException(false),
       _binDataProto(_context),
       _bsonProto(_context),
@@ -491,82 +499,81 @@ void MozJSImplScope::init(const BSONObj* data) {
     }
 }
 
-void MozJSImplScope::setNumber(const char* field, double val) {
-    MozJSEntry entry(this);
+template <typename ImplScopeFunction>
+auto MozJSImplScope::_runSafely(ImplScopeFunction&& functionToRun) -> decltype(functionToRun()) {
+    try {
+        MozJSEntry entry(this);
+        return functionToRun();
+    } catch (...) {
+        _error = _status.reason();
 
-    ObjectWrapper(_context, _global).setNumber(field, val);
+        // Clear the status state
+        auto status = std::move(_status);
+        uassertStatusOK(status);
+        throw;
+    }
 }
 
-void MozJSImplScope::setString(const char* field, StringData val) {
-    MozJSEntry entry(this);
+void MozJSImplScope::setNumber(const char* field, double val) {
+    _runSafely([this, &field, &val] { ObjectWrapper(_context, _global).setNumber(field, val); });
+}
 
-    ObjectWrapper(_context, _global).setString(field, val);
+void MozJSImplScope::setString(const char* field, StringData val) {
+    _runSafely([this, &field, &val] { ObjectWrapper(_context, _global).setString(field, val); });
 }
 
 void MozJSImplScope::setBoolean(const char* field, bool val) {
-    MozJSEntry entry(this);
-
-    ObjectWrapper(_context, _global).setBoolean(field, val);
+    _runSafely([this, &field, &val] { ObjectWrapper(_context, _global).setBoolean(field, val); });
 }
 
 void MozJSImplScope::setElement(const char* field, const BSONElement& e, const BSONObj& parent) {
-    MozJSEntry entry(this);
+    _runSafely([this, &field, &e, &parent] {
 
-    ObjectWrapper(_context, _global).setBSONElement(field, e, parent, false);
+        ObjectWrapper(_context, _global).setBSONElement(field, e, parent, false);
+    });
 }
 
 void MozJSImplScope::setObject(const char* field, const BSONObj& obj, bool readOnly) {
-    MozJSEntry entry(this);
+    _runSafely([this, &field, &obj, &readOnly] {
 
-    ObjectWrapper(_context, _global).setBSON(field, obj, readOnly);
+        ObjectWrapper(_context, _global).setBSON(field, obj, readOnly);
+    });
 }
 
 int MozJSImplScope::type(const char* field) {
-    MozJSEntry entry(this);
-
-    return ObjectWrapper(_context, _global).type(field);
+    return _runSafely([this, &field] { return ObjectWrapper(_context, _global).type(field); });
 }
 
 double MozJSImplScope::getNumber(const char* field) {
-    MozJSEntry entry(this);
-
-    return ObjectWrapper(_context, _global).getNumber(field);
+    return _runSafely([this, &field] { return ObjectWrapper(_context, _global).getNumber(field); });
 }
 
 int MozJSImplScope::getNumberInt(const char* field) {
-    MozJSEntry entry(this);
-
-    return ObjectWrapper(_context, _global).getNumberInt(field);
+    return _runSafely(
+        [this, &field] { return ObjectWrapper(_context, _global).getNumberInt(field); });
 }
 
 long long MozJSImplScope::getNumberLongLong(const char* field) {
-    MozJSEntry entry(this);
-
-    return ObjectWrapper(_context, _global).getNumberLongLong(field);
+    return _runSafely(
+        [this, &field] { return ObjectWrapper(_context, _global).getNumberLongLong(field); });
 }
 
 Decimal128 MozJSImplScope::getNumberDecimal(const char* field) {
-    MozJSEntry entry(this);
-
-    return ObjectWrapper(_context, _global).getNumberDecimal(field);
+    return _runSafely(
+        [this, &field] { return ObjectWrapper(_context, _global).getNumberDecimal(field); });
 }
 
 std::string MozJSImplScope::getString(const char* field) {
-    MozJSEntry entry(this);
-
-    return ObjectWrapper(_context, _global).getString(field);
+    return _runSafely([this, &field] { return ObjectWrapper(_context, _global).getString(field); });
 }
 
 bool MozJSImplScope::getBoolean(const char* field) {
-    MozJSEntry entry(this);
-
-    return ObjectWrapper(_context, _global).getBoolean(field);
+    return _runSafely(
+        [this, &field] { return ObjectWrapper(_context, _global).getBoolean(field); });
 }
 
 BSONObj MozJSImplScope::getObject(const char* field) {
-    MozJSEntry entry(this);
-
-    return ObjectWrapper(_context, _global).getObject(field);
+    return _runSafely([this, &field] { return ObjectWrapper(_context, _global).getObject(field); });
 }
 
 void MozJSImplScope::newFunction(StringData raw, JS::MutableHandleValue out) {
@@ -642,17 +649,15 @@ ScriptingFunction MozJSImplScope::_createFunction(const char* raw) {
 }
 
 void MozJSImplScope::setFunction(const char* field, const char* code) {
-    MozJSEntry entry(this);
-
-    JS::RootedValue fun(_context);
-    _MozJSCreateFunction(code, &fun);
-    ObjectWrapper(_context, _global).setValue(field, fun);
+    _runSafely([this, &field, &code] {
+        JS::RootedValue fun(_context);
+        _MozJSCreateFunction(code, &fun);
+        ObjectWrapper(_context, _global).setValue(field, fun);
+    });
 }
 
 void MozJSImplScope::rename(const char* from, const char* to) {
-    MozJSEntry entry(this);
-
-    ObjectWrapper(_context, _global).rename(from, to);
+    _runSafely([this, &from, &to] { ObjectWrapper(_context, _global).rename(from, to); });
 }
 
 int MozJSImplScope::invoke(ScriptingFunction func,
@@ -764,15 +769,15 @@ bool MozJSImplScope::exec(StringData code,
 }
 
 void MozJSImplScope::injectNative(const char* field, NativeFunction func, void* data) {
-    MozJSEntry entry(this);
-
-    JS::RootedObject obj(_context);
+    _runSafely([this, &field, &func, &data] {
+        JS::RootedObject obj(_context);
 
-    NativeFunctionInfo::make(_context, &obj, func, data);
+        NativeFunctionInfo::make(_context, &obj, func, data);
 
-    JS::RootedValue value(_context);
-    value.setObjectOrNull(obj);
-    ObjectWrapper(_context, _global).setValue(field, value);
+        JS::RootedValue value(_context);
+        value.setObjectOrNull(obj);
+        ObjectWrapper(_context, _global).setValue(field, value);
+    });
 }
 
 void MozJSImplScope::gc() {
@@ -781,63 +786,65 @@ void MozJSImplScope::gc() {
 }
 
 void MozJSImplScope::localConnectForDbEval(OperationContext* txn, const char* dbName) {
-    MozJSEntry entry(this);
-
-    if (_connectState == ConnectState::External)
-        uasserted(12510, "externalSetup already called, can't call localConnect");
-    if (_connectState == ConnectState::Local) {
-        if (_localDBName == dbName)
-            return;
-        uasserted(12511,
-                  str::stream() << "localConnect previously called with name " << _localDBName);
-    }
+    _runSafely([this, &txn, &dbName] {
+        if (_connectState == ConnectState::External)
+            uasserted(12510, "externalSetup already called, can't call localConnect");
+        if (_connectState == ConnectState::Local) {
+            if (_localDBName == dbName)
+                return;
+            uasserted(12511,
+                      str::stream() << "localConnect previously called with name " << _localDBName);
+        }
 
-    // NOTE: order is important here.  the following methods must be called after
-    //       the above conditional statements.
+        // NOTE: order is important here.  the following methods must be called after
+        //       the above conditional statements.
 
-    _connectState = ConnectState::Local;
-    _localDBName = dbName;
+        _connectState = ConnectState::Local;
+        _localDBName = dbName;
 
-    loadStored(txn);
+        loadStored(txn);
 
-    // install db access functions in the global object
-    installDBAccess();
+        // install db access functions in the global object
+        installDBAccess();
 
-    // install the Mongo function object and instantiate the 'db' global
-    _mongoLocalProto.install(_global);
-    execCoreFiles();
+        // install the Mongo function object and instantiate the 'db' global
+        _mongoLocalProto.install(_global);
+        execCoreFiles();
 
-    const char* const makeMongo = "const _mongo = new Mongo()";
-    exec(makeMongo, "local connect 2", false, true, true, 0);
+        const char* const makeMongo = "const _mongo = new Mongo()";
+        exec(makeMongo, "local connect 2", false, true, true, 0);
 
-    std::string makeDB = str::stream() << "const db = _mongo.getDB(\"" << dbName << "\");";
-    exec(makeDB, "local connect 3", false, true, true, 0);
+        std::string makeDB = str::stream() << "const db = _mongo.getDB(\"" << dbName << "\");";
+        exec(makeDB, "local connect 3", false, true, true, 0);
+    });
 }
 
 void MozJSImplScope::externalSetup() {
-    MozJSEntry entry(this);
 
-    if (_connectState == ConnectState::External)
-        return;
-    if (_connectState == ConnectState::Local)
-        uasserted(12512, "localConnect already called, can't call externalSetup");
+    _runSafely([&] {
+        if (_connectState == ConnectState::External)
+            return;
+        if (_connectState == ConnectState::Local)
+            uasserted(12512, "localConnect already called, can't call externalSetup");
 
-    // install db access functions in the global object
-    installDBAccess();
+        // install db access functions in the global object
+        installDBAccess();
 
-    // install thread-related functions (e.g. _threadInject)
-    installFork();
+        // install thread-related functions (e.g. _threadInject)
+        installFork();
 
-    // install the Mongo function object
-    _mongoExternalProto.install(_global);
-    execCoreFiles();
-    _connectState = ConnectState::External;
+        // install the Mongo function object
+        _mongoExternalProto.install(_global);
+        execCoreFiles();
+        _connectState = ConnectState::External;
+    });
 }
 
 void MozJSImplScope::reset() {
     unregisterOperation();
     _pendingKill.store(false);
     _pendingGC.store(false);
+    _requireOwnedObjects = false;
     advanceGeneration();
 }
 
@@ -951,6 +958,14 @@ void MozJSImplScope::advanceGeneration() {
     _generation++;
 }
 
+void MozJSImplScope::requireOwnedObjects() {
+    _requireOwnedObjects = true;
+}
+
+bool MozJSImplScope::requiresOwnedObjects() const {
+    return _requireOwnedObjects;
+}
+
 const std::string& MozJSImplScope::getParentStack() const {
     return _parentStack;
 }
diff --git a/src/mongo/scripting/mozjs/implscope.h b/src/mongo/scripting/mozjs/implscope.h
index df37553e181..2de309e87ab 100644
--- a/src/mongo/scripting/mozjs/implscope.h
+++ b/src/mongo/scripting/mozjs/implscope.h
@@ -309,6 +309,10 @@ public:
 
     void advanceGeneration() override;
 
+    void requireOwnedObjects() override;
+
+    bool requiresOwnedObjects() const;
+
     JS::HandleId getInternedStringId(InternedString name) {
         return _internedStrings.getInternedString(name);
     }
@@ -341,6 +345,9 @@ public:
     };
 
 private:
+    template <typename ImplScopeFunction>
+    auto _runSafely(ImplScopeFunction&& functionToRun) -> decltype(functionToRun());
+
     void _MozJSCreateFunction(StringData raw, JS::MutableHandleValue fun);
 
     /**
@@ -402,6 +409,7 @@ private:
     Status _status;
     std::string _parentStack;
     std::size_t _generation;
+    bool _requireOwnedObjects;
     bool _hasOutOfMemoryException;
 
     WrapType<BinDataInfo> _binDataProto;
diff --git a/src/mongo/scripting/mozjs/mongohelpers.js b/src/mongo/scripting/mozjs/mongohelpers.js
index b0b35bb2fe8..d3f743623a3 100644
--- a/src/mongo/scripting/mozjs/mongohelpers.js
+++ b/src/mongo/scripting/mozjs/mongohelpers.js
@@ -33,6 +33,14 @@
 exportToMongoHelpers = {
     // This function accepts an expression or function body and returns a function definition
     'functionExpressionParser': function functionExpressionParser(fnSrc) {
+
+        // Ensure that a provided expression or function body is not terminated with a ';'.
+        // This ensures we interpret the input as a single expression, rather than a sequence
+        // of expressions, and can wrap it in parentheses.
+        while (fnSrc.endsWith(";") || fnSrc != fnSrc.trimRight()) {
+            fnSrc = fnSrc.slice(0, -1).trimRight();
+        }
+
         var parseTree;
         try {
             parseTree = this.Reflect.parse(fnSrc);
diff --git a/src/mongo/scripting/mozjs/proxyscope.cpp b/src/mongo/scripting/mozjs/proxyscope.cpp
index 6a6d20f4448..05697468f07 100644
--- a/src/mongo/scripting/mozjs/proxyscope.cpp
+++ b/src/mongo/scripting/mozjs/proxyscope.cpp
@@ -122,6 +122,10 @@ void MozJSProxyScope::advanceGeneration() {
     run([&] { _implScope->advanceGeneration(); });
 }
 
+void MozJSProxyScope::requireOwnedObjects() {
+    run([&] { _implScope->requireOwnedObjects(); });
+}
+
 double MozJSProxyScope::getNumber(const char* field) {
     double out;
     run([&] { out = _implScope->getNumber(field); });
diff --git a/src/mongo/scripting/mozjs/proxyscope.h b/src/mongo/scripting/mozjs/proxyscope.h
index 451981330a1..4dd69a3ebe9 100644
--- a/src/mongo/scripting/mozjs/proxyscope.h
+++ b/src/mongo/scripting/mozjs/proxyscope.h
@@ -129,6 +129,8 @@ public:
 
     void advanceGeneration() override;
 
+    void requireOwnedObjects() override;
+
     double getNumber(const char* field) override;
     int getNumberInt(const char* field) override;
     long long getNumberLongLong(const char* field) override;
diff --git a/src/mongo/shell/bench.cpp b/src/mongo/shell/bench.cpp
index 040002f5c6e..338477ebe56 100644
--- a/src/mongo/shell/bench.cpp
+++ b/src/mongo/shell/bench.cpp
@@ -674,7 +674,7 @@ void BenchRunWorker::generateLoadOnConnection(DBClientBase* conn) {
     invariant(bsonTemplateEvaluator.setId(_id) == BsonTemplateEvaluator::StatusSuccess);
 
     if (_config->username != "") {
-        string errmsg;
+        std::string errmsg;
         if (!conn->auth("admin", _config->username, _config->password, errmsg)) {
             uasserted(15931, "Authenticating to connection for _benchThread failed: " + errmsg);
         }
@@ -918,7 +918,7 @@ void BenchRunWorker::generateLoadOnConnection(DBClientBase* conn) {
 
                             if (!result["err"].eoo() && result["err"].type() == String &&
                                 (_config->throwGLE || op.throwGLE))
-                                throw DBException((string) "From benchRun GLE" +
+                                throw DBException((std::string) "From benchRun GLE" +
                                                       causedBy(result["err"].String()),
                                                   result["code"].eoo() ? 0 : result["code"].Int());
                         }
@@ -984,7 +984,7 @@ void BenchRunWorker::generateLoadOnConnection(DBClientBase* conn) {
 
                             if (!result["err"].eoo() && result["err"].type() == String &&
                                 (_config->throwGLE || op.throwGLE))
-                                throw DBException((string) "From benchRun GLE" +
+                                throw DBException((std::string) "From benchRun GLE" +
                                                       causedBy(result["err"].String()),
                                                   result["code"].eoo() ? 0 : result["code"].Int());
                         }
@@ -1031,7 +1031,7 @@ void BenchRunWorker::generateLoadOnConnection(DBClientBase* conn) {
 
                             if (!result["err"].eoo() && result["err"].type() == String &&
                                 (_config->throwGLE || op.throwGLE))
-                                throw DBException((string) "From benchRun GLE " +
+                                throw DBException((std::string) "From benchRun GLE " +
                                                       causedBy(result["err"].String()),
                                                   result["code"].eoo() ? 0 : result["code"].Int());
                         }
@@ -1133,7 +1133,7 @@ void BenchRunWorker::run() {
     try {
         std::unique_ptr<DBClientBase> conn(_config->createConnection());
         if (!_config->username.empty()) {
-            string errmsg;
+            std::string errmsg;
             if (!conn->auth("admin", _config->username, _config->password, errmsg)) {
                 uasserted(15932, "Authenticating to connection for benchThread failed: " + errmsg);
             }
@@ -1165,7 +1165,7 @@ void BenchRunner::start() {
         std::unique_ptr<DBClientBase> conn(_config->createConnection());
         // Must authenticate to admin db in order to run serverStatus command
         if (_config->username != "") {
-            string errmsg;
+            std::string errmsg;
             if (!conn->auth("admin", _config->username, _config->password, errmsg)) {
                 uasserted(
                     16704,
@@ -1201,7 +1201,7 @@ void BenchRunner::stop() {
     {
         std::unique_ptr<DBClientBase> conn(_config->createConnection());
         if (_config->username != "") {
-            string errmsg;
+            std::string errmsg;
             // this can only fail if admin access was revoked since start of run
             if (!conn->auth("admin", _config->username, _config->password, errmsg)) {
                 uasserted(
diff --git a/src/mongo/shell/dbshell.cpp b/src/mongo/shell/dbshell.cpp
index 801130d07c5..90c59f0ee55 100644
--- a/src/mongo/shell/dbshell.cpp
+++ b/src/mongo/shell/dbshell.cpp
@@ -43,6 +43,7 @@
 #include "mongo/base/initializer.h"
 #include "mongo/base/status.h"
 #include "mongo/client/dbclientinterface.h"
+#include "mongo/client/mongo_uri.h"
 #include "mongo/client/sasl_client_authenticate.h"
 #include "mongo/db/client.h"
 #include "mongo/db/log_process_details.h"
@@ -215,57 +216,151 @@ void setupSignals() {
     signal(SIGINT, quitNicely);
 }
 
-string getURIFromArgs(const std::string& url, const std::string& host, const std::string& port) {
-    if (host.size() == 0 && port.size() == 0) {
-        return url.size() == 0 ? kDefaultMongoURL.toString() : url;
+string getURIFromArgs(const std::string& arg, const std::string& host, const std::string& port) {
+    if (host.empty() && arg.empty() && port.empty()) {
+        // Nothing provided, just play the default.
+        return kDefaultMongoURL.toString();
     }
 
-    // The name URL is misleading; really it's just a positional argument that wasn't a file. The
-    // check for "/" means "this 'URL' is probably a real URL and not the db name (e.g.)".
-    if (url.find("/") != string::npos) {
-        cerr << "if a full URI is provided, you cannot also specify host or port" << endl;
+    if (str::startsWith(arg, "mongodb://") && host.empty() && port.empty()) {
+        // mongo mongodb://blah
+        return arg;
+    }
+    if (str::startsWith(host, "mongodb://") && arg.empty() && port.empty()) {
+        // mongo --host mongodb://blah
+        return host;
+    }
+
+    // We expect a positional arg to be a plain dbname or plain hostname at this point
+    // since we have separate host/port args.
+    if ((arg.find('/') != string::npos) && (host.size() || port.size())) {
+        cerr << "If a full URI is provided, you cannot also specify --host or --port" << endl;
         quickExit(-1);
     }
 
-    bool hostEndsInSock = str::endsWith(host, ".sock");
+    const auto parseDbHost = [port](const std::string& db, const std::string& host) -> std::string {
+        // Parse --host as a connection string.
+        // e.g. rs0/host0:27000,host1:27001
+        const auto slashPos = host.find('/');
+        const auto hasReplSet = (slashPos > 0) && (slashPos != std::string::npos);
+
+        std::ostringstream ss;
+        ss << "mongodb://";
+
+        // Handle each sub-element of the connection string individually.
+        // Comma separated list of host elements.
+        // Each host element may be:
+        // * /unix/domain.sock
+        // * hostname
+        // * hostname:port
+        // If --port is specified and port is included in connection string,
+        // then they must match exactly.
+        auto start = hasReplSet ? slashPos + 1 : 0;
+        while (start < host.size()) {
+            // Encode each host component.
+            auto end = host.find(',', start);
+            if (end == std::string::npos) {
+                end = host.size();
+            }
+            if ((end - start) == 0) {
+                // Ignore empty components.
+                start = end + 1;
+                continue;
+            }
+
+            const auto hostElem = host.substr(start, end - start);
+            if ((hostElem.find('/') != std::string::npos) && str::endsWith(hostElem, ".sock")) {
+                // Unix domain socket, ignore --port.
+                ss << uriEncode(hostElem);
+
+            } else {
+                auto colon = hostElem.find(':');
+                if ((colon != std::string::npos) &&
+                    (hostElem.find(':', colon + 1) != std::string::npos)) {
+                    // Looks like an IPv6 numeric address.
+                    const auto close = hostElem.find(']');
+                    if ((hostElem[0] == '[') && (close != std::string::npos)) {
+                        // Encapsulated already.
+                        ss << '[' << uriEncode(hostElem.substr(1, close - 1), ":") << ']';
+                        colon = hostElem.find(':', close + 1);
+                    } else {
+                        // Not encapsulated yet.
+                        ss << '[' << uriEncode(hostElem, ":") << ']';
+                        colon = std::string::npos;
+                    }
+                } else if (colon != std::string::npos) {
+                    // Not IPv6 numeric, but does have a port.
+                    ss << uriEncode(hostElem.substr(0, colon));
+                } else {
+                    // Raw hostname/IPv4 without port.
+                    ss << uriEncode(hostElem);
+                }
+
+                if (colon != std::string::npos) {
+                    // Have a port in our host element, verify it.
+                    const auto myport = hostElem.substr(colon + 1);
+                    if (port.size() && (port != myport)) {
+                        cerr << "connection string bears different port than provided by --port"
+                             << endl;
+                        quickExit(-1);
+                    }
+                    ss << ':' << uriEncode(myport);
+                } else if (port.size()) {
+                    ss << ':' << uriEncode(port);
+                } else {
+                    ss << ":27017";
+                }
+            }
+            start = end + 1;
+            if (start < host.size()) {
+                ss << ',';
+            }
+        }
+
+        ss << '/' << uriEncode(db);
 
-    // If host looks like a full URI (i.e. has a slash and isn't a unix socket) and the other fields
-    // are empty, then just return host.
-    std::string::size_type slashPos;
-    if (url.size() == 0 && port.size() == 0 &&
-        (!hostEndsInSock && ((slashPos = host.find("/")) != string::npos))) {
-        if (str::startsWith(host, "mongodb://")) {
-            return host;
+        if (hasReplSet) {
+            // Remap included replica set name to URI option
+            ss << "?replicaSet=" << uriEncode(host.substr(0, slashPos));
         }
-        // If there's a slash in the host field, then it's the replica set name, not a database name
-        stringstream ss;
-        ss << "mongodb://" << host.substr(slashPos + 1)
-           << "/?replicaSet=" << host.substr(0, slashPos);
+
         return ss.str();
-    }
+    };
+
+    if (host.size()) {
+        // --host provided, treat it as the connect string and get db from positional arg.
+        return parseDbHost(arg, host);
+    } else if (arg.size()) {
+        // --host missing, but we have a potential host/db positional arg.
+        const auto slashPos = arg.find('/');
+        if (slashPos != std::string::npos) {
+            // host/db pair.
+            return parseDbHost(arg.substr(slashPos + 1), arg.substr(0, slashPos));
+        }
 
-    stringstream ss;
-    if (host.size() == 0) {
-        ss << "mongodb://127.0.0.1";
-    } else {
-        if (!str::startsWith(host, "mongodb://")) {
-            ss << "mongodb://";
+        // Compatability formats.
+        // * Any arg with a dot is assumed to be a hostname or IPv4 numeric address.
+        // * Any arg with a colon followed by a digit assumed to be host or IP followed by port.
+        // * Anything else is assumed to be a db.
+
+        if (arg.find('.') != std::string::npos) {
+            // Assume IPv4 or hostnameish.
+            return parseDbHost("test", arg);
         }
-        ss << host;
-    }
 
-    if (!hostEndsInSock) {
-        if (port.size() > 0) {
-            ss << ":" << port;
-        } else if (host.find(':') == string::npos || str::endsWith(host, "]")) {
-            // Default the port to 27017 if the host did not provide one (i.e. the host has no
-            // colons or ends in ']' like an IPv6 address).
-            ss << ":27017";
+        const auto colonPos = arg.find(':');
+        if ((colonPos != std::string::npos) && ((colonPos + 1) < arg.size()) &&
+            isdigit(arg[colonPos + 1])) {
+            // Assume IPv4 or hostname with port.
+            return parseDbHost("test", arg);
         }
+
+        // db, assume localhost.
+        return parseDbHost(arg, "127.0.0.1");
     }
 
-    ss << "/" << url;
-    return ss.str();
+    // --host empty, position arg empty, fallback on localhost without a dbname.
+    return parseDbHost("", "127.0.0.1");
 }
 
 static string OpSymbols = "~!%^&*-+=|:,<>/?.";
diff --git a/src/mongo/shell/mongo.js b/src/mongo/shell/mongo.js
index 69971609bb8..92fc83b38b9 100644
--- a/src/mongo/shell/mongo.js
+++ b/src/mongo/shell/mongo.js
@@ -223,6 +223,9 @@ connect = function(url, user, pass) {
     if (!url.startsWith("mongodb://")) {
         const colon = url.lastIndexOf(":");
         const slash = url.lastIndexOf("/");
+        if (url.split("/").length > 1) {
+            url = url.substring(0, slash).replace(/\//g, "%2F") + url.substring(slash);
+        }
         if (slash == 0) {
             throw Error("Failed to parse mongodb:// URL: " + url);
         }
@@ -235,7 +238,7 @@ connect = function(url, user, pass) {
 
     chatty("connecting to: " + url);
     var m = new Mongo(url);
-    db = m.getDB(m.defaultDB);
+    var db = m.getDB(m.defaultDB);
 
     if (user && pass) {
         if (!db.auth(user, pass)) {
diff --git a/src/mongo/shell/replsettest.js b/src/mongo/shell/replsettest.js
index 0e05a1bfa4a..9d6a4378009 100644
--- a/src/mongo/shell/replsettest.js
+++ b/src/mongo/shell/replsettest.js
@@ -427,8 +427,10 @@ var ReplSetTest = function(opts) {
             var member = {};
             member._id = i;
 
-            var port = this.ports[i];
-            member.host = this.host + ":" + port;
+            member.host = this.host;
+            if (!member.host.contains('/')) {
+                member.host += ":" + this.ports[i];
+            }
 
             var nodeOpts = this.nodeOptions["n" + i];
             if (nodeOpts) {
@@ -470,7 +472,7 @@ var ReplSetTest = function(opts) {
      *
      * @param options - The options passed to {@link MongoRunner.runMongod}
      */
-    this.startSet = function(options) {
+    this.startSet = function(options, restart) {
         print("ReplSetTest starting set");
 
         if (options && options.keyFile) {
@@ -482,7 +484,7 @@ var ReplSetTest = function(opts) {
 
         var nodes = [];
         for (var n = 0; n < this.ports.length; n++) {
-            nodes.push(this.start(n, options));
+            nodes.push(this.start(n, options, restart));
         }
 
         this.nodes = nodes;
@@ -1883,7 +1885,11 @@ var ReplSetTest = function(opts) {
      * Constructor, which instantiates the ReplSetTest object from an existing set.
      */
     function _constructFromExistingSeedNode(seedNode) {
-        var conf = _replSetGetConfig(new Mongo(seedNode));
+        const conn = new Mongo(seedNode);
+        if (jsTest.options().keyFile) {
+            self.keyFile = jsTest.options().keyFile;
+        }
+        var conf = asCluster(conn, () => _replSetGetConfig(conn));
         print('Recreating replica set from config ' + tojson(conf));
 
         var existingNodes = conf.members.map(member => member.host);
diff --git a/src/mongo/shell/servers_misc.js b/src/mongo/shell/servers_misc.js
index fdb49deb76a..f419428e2d0 100644
--- a/src/mongo/shell/servers_misc.js
+++ b/src/mongo/shell/servers_misc.js
@@ -214,11 +214,21 @@ function startParallelShell(jsCode, port, noConnect) {
     var args = ["mongo"];
 
     if (typeof db == "object") {
-        var hostAndPort = db.getMongo().host.split(':');
-        var host = hostAndPort[0];
-        args.push("--host", host);
-        if (!port && hostAndPort.length >= 2) {
-            var port = hostAndPort[1];
+        if (!port) {
+            // If no port override specified, just passthrough connect string.
+            args.push("--host", db.getMongo().host);
+        } else {
+            // Strip port numbers from connect string.
+            const uri = new MongoURI(db.getMongo().host);
+            var connString = uri.servers
+                                 .map(function(server) {
+                                     return server.host;
+                                 })
+                                 .join(',');
+            if (uri.setName.length > 0) {
+                connString = uri.setName + '/' + connString;
+            }
+            args.push("--host", connString);
         }
     }
     if (port) {
diff --git a/src/mongo/shell/shardingtest.js b/src/mongo/shell/shardingtest.js
index 9b06dcd89ce..7a0fdae70f3 100644
--- a/src/mongo/shell/shardingtest.js
+++ b/src/mongo/shell/shardingtest.js
@@ -370,10 +370,14 @@ var ShardingTest = function(params) {
         }
     };
 
-    this.stop = function(opts) {
+    this.stopAllMongos = function(opts) {
         for (var i = 0; i < this._mongos.length; i++) {
             this.stopMongos(i, opts);
         }
+    };
+
+    this.stop = function(opts) {
+        this.stopAllMongos(opts);
 
         for (var i = 0; i < this._connections.length; i++) {
             if (this._rs[i]) {
diff --git a/src/mongo/shell/utils.js b/src/mongo/shell/utils.js
index 9794a60df4e..148eacc1c89 100644
--- a/src/mongo/shell/utils.js
+++ b/src/mongo/shell/utils.js
@@ -158,6 +158,18 @@ print.captureAllOutput = function(fn, args) {
     return res;
 };
 
+var indentStr = function(indent, s) {
+    if (typeof(s) === "undefined") {
+        s = indent;
+        indent = 0;
+    }
+    if (indent > 0) {
+        indent = (new Array(indent + 1)).join(" ");
+        s = indent + s.replace(/\n/g, "\n" + indent);
+    }
+    return s;
+};
+
 if (typeof TestData == "undefined") {
     TestData = undefined;
 }
@@ -230,7 +242,8 @@ jsTestOptions = function() {
             networkMessageCompressors: TestData.networkMessageCompressors,
             skipValidationOnInvalidViewDefinitions: TestData.skipValidationOnInvalidViewDefinitions,
             forceValidationWithFeatureCompatibilityVersion:
-                TestData.forceValidationWithFeatureCompatibilityVersion
+                TestData.forceValidationWithFeatureCompatibilityVersion,
+            skipValidationNamespaces: TestData.skipValidationNamespaces || [],
         });
     }
     return _jsTestOptions;
diff --git a/src/mongo/shell/utils_sh.js b/src/mongo/shell/utils_sh.js
index ef8ae2dfa36..6e9c61a32e5 100644
--- a/src/mongo/shell/utils_sh.js
+++ b/src/mongo/shell/utils_sh.js
@@ -632,6 +632,18 @@ sh.getRecentMigrations = function(configDB) {
     return result;
 };
 
+sh._shardingStatusStr = function(indent, s) {
+    // convert from logical indentation to actual num of chars
+    if (indent == 0) {
+        indent = 0;
+    } else if (indent == 1) {
+        indent = 2;
+    } else {
+        indent = (indent - 1) * 8;
+    }
+    return indentStr(indent, s) + "\n";
+};
+
 function printShardingStatus(configDB, verbose) {
     // configDB is a DB object that contains the sharding metadata of interest.
     // Defaults to the db named "config" on the current connection.
@@ -646,15 +658,15 @@ function printShardingStatus(configDB, verbose) {
     }
 
     var raw = "";
-    var output = function(s) {
-        raw += s + "\n";
+    var output = function(indent, s) {
+        raw += sh._shardingStatusStr(indent, s);
     };
-    output("--- Sharding Status --- ");
-    output("  sharding version: " + tojson(configDB.getCollection("version").findOne()));
+    output(0, "--- Sharding Status --- ");
+    output(1, "sharding version: " + tojson(configDB.getCollection("version").findOne()));
 
-    output("  shards:");
+    output(1, "shards:");
     configDB.shards.find().sort({_id: 1}).forEach(function(z) {
-        output("\t" + tojsononeline(z));
+        output(2, tojsononeline(z));
     });
 
     // (most recently) active mongoses
@@ -672,9 +684,9 @@ function printShardingStatus(configDB, verbose) {
         }
     }
 
-    output("  " + mongosAdjective + " mongoses:");
+    output(1, mongosAdjective + " mongoses:");
     if (mostRecentMongosTime === null) {
-        output("\tnone");
+        output(2, "none");
     } else {
         var recentMongosQuery = {
             ping: {
@@ -688,7 +700,7 @@ function printShardingStatus(configDB, verbose) {
 
         if (verbose) {
             configDB.mongos.find(recentMongosQuery).sort({ping: -1}).forEach(function(z) {
-                output("\t" + tojsononeline(z));
+                output(2, tojsononeline(z));
             });
         } else {
             configDB.mongos
@@ -698,23 +710,23 @@ function printShardingStatus(configDB, verbose) {
                     {$sort: {num: -1}}
                 ])
                 .forEach(function(z) {
-                    output("\t" + tojson(z._id) + " : " + z.num);
+                    output(2, tojson(z._id) + " : " + z.num);
                 });
         }
     }
 
-    output(" autosplit:");
+    output(1, "autosplit:");
 
     // Is autosplit currently enabled
-    output("\tCurrently enabled: " + (sh.getShouldAutoSplit(configDB) ? "yes" : "no"));
+    output(2, "Currently enabled: " + (sh.getShouldAutoSplit(configDB) ? "yes" : "no"));
 
-    output("  balancer:");
+    output(1, "balancer:");
 
     // Is the balancer currently enabled
-    output("\tCurrently enabled:  " + (sh.getBalancerState(configDB) ? "yes" : "no"));
+    output(2, "Currently enabled:  " + (sh.getBalancerState(configDB) ? "yes" : "no"));
 
     // Is the balancer currently active
-    output("\tCurrently running:  " + (sh.isBalancerRunning(configDB) ? "yes" : "no"));
+    output(2, "Currently running:  " + (sh.isBalancerRunning(configDB) ? "yes" : "no"));
 
     // Output details of the current balancer round
     var balLock = sh.getBalancerLockDetails(configDB);
@@ -725,16 +737,17 @@ function printShardingStatus(configDB, verbose) {
     // Output the balancer window
     var balSettings = sh.getBalancerWindow(configDB);
     if (balSettings) {
-        output("\t\tBalancer active window is set between " + balSettings.start + " and " +
-               balSettings.stop + " server local time");
+        output(3,
+               "Balancer active window is set between " + balSettings.start + " and " +
+                   balSettings.stop + " server local time");
     }
 
     // Output the list of active migrations
     var activeMigrations = sh.getActiveMigrations(configDB);
     if (activeMigrations.length > 0) {
-        output("\tCollections with active migrations: ");
+        output(2, "Collections with active migrations: ");
         activeMigrations.forEach(function(migration) {
-            output("\t\t" + migration._id + " started at " + migration.when);
+            output(3, migration._id + " started at " + migration.when);
         });
     }
 
@@ -755,31 +768,32 @@ function printShardingStatus(configDB, verbose) {
         // Review config.actionlog for errors
         var actionReport = sh.getRecentFailedRounds(configDB);
         // Always print the number of failed rounds
-        output("\tFailed balancer rounds in last 5 attempts:  " + actionReport.count);
+        output(2, "Failed balancer rounds in last 5 attempts:  " + actionReport.count);
 
         // Only print the errors if there are any
         if (actionReport.count > 0) {
-            output("\tLast reported error:  " + actionReport.lastErr);
-            output("\tTime of Reported error:  " + actionReport.lastTime);
+            output(2, "Last reported error:  " + actionReport.lastErr);
+            output(2, "Time of Reported error:  " + actionReport.lastTime);
         }
 
-        output("\tMigration Results for the last 24 hours: ");
+        output(2, "Migration Results for the last 24 hours: ");
         var migrations = sh.getRecentMigrations(configDB);
         if (migrations.length > 0) {
             migrations.forEach(function(x) {
                 if (x._id === "Success") {
-                    output("\t\t" + x.count + " : " + x._id);
+                    output(3, x.count + " : " + x._id);
                 } else {
-                    output("\t\t" + x.count + " : Failed with error '" + x._id + "', from " +
-                           x.from + " to " + x.to);
+                    output(3,
+                           x.count + " : Failed with error '" + x._id + "', from " + x.from +
+                               " to " + x.to);
                 }
             });
         } else {
-            output("\t\tNo recent migrations");
+            output(3, "No recent migrations");
         }
     }
 
-    output("  databases:");
+    output(1, "databases:");
     configDB.databases.find().sort({name: 1}).forEach(function(db) {
         var truthy = function(value) {
             return !!value;
@@ -795,20 +809,22 @@ function printShardingStatus(configDB, verbose) {
             return s;
         };
 
-        output("\t" + tojsononeline(db, "", true));
+        output(2, tojsononeline(db, "", true));
 
         if (db.partitioned) {
             configDB.collections.find({_id: new RegExp("^" + RegExp.escape(db._id) + "\\.")})
                 .sort({_id: 1})
                 .forEach(function(coll) {
                     if (!coll.dropped) {
-                        output("\t\t" + coll._id);
-                        output("\t\t\tshard key: " + tojson(coll.key));
-                        output("\t\t\tunique: " + truthy(coll.unique) +
-                               nonBooleanNote("unique", coll.unique));
-                        output("\t\t\tbalancing: " + !truthy(coll.noBalance) +
-                               nonBooleanNote("noBalance", coll.noBalance));
-                        output("\t\t\tchunks:");
+                        output(3, coll._id);
+                        output(4, "shard key: " + tojson(coll.key));
+                        output(4,
+                               "unique: " + truthy(coll.unique) +
+                                   nonBooleanNote("unique", coll.unique));
+                        output(4,
+                               "balancing: " + !truthy(coll.noBalance) +
+                                   nonBooleanNote("noBalance", coll.noBalance));
+                        output(4, "chunks:");
 
                         res = configDB.chunks
                                   .aggregate({$match: {ns: coll._id}},
@@ -819,26 +835,29 @@ function printShardingStatus(configDB, verbose) {
                         var totalChunks = 0;
                         res.forEach(function(z) {
                             totalChunks += z.nChunks;
-                            output("\t\t\t\t" + z.shard + "\t" + z.nChunks);
+                            output(5, z.shard + "\t" + z.nChunks);
                         });
 
                         if (totalChunks < 20 || verbose) {
                             configDB.chunks.find({"ns": coll._id})
                                 .sort({min: 1})
                                 .forEach(function(chunk) {
-                                    output("\t\t\t" + tojson(chunk.min) + " -->> " +
-                                           tojson(chunk.max) + " on : " + chunk.shard + " " +
-                                           tojson(chunk.lastmod) + " " +
-                                           (chunk.jumbo ? "jumbo " : ""));
+                                    output(4,
+                                           tojson(chunk.min) + " -->> " + tojson(chunk.max) +
+                                               " on : " + chunk.shard + " " +
+                                               tojson(chunk.lastmod) + " " +
+                                               (chunk.jumbo ? "jumbo " : ""));
                                 });
                         } else {
                             output(
-                                "\t\t\ttoo many chunks to print, use verbose if you want to force print");
+                                4,
+                                "too many chunks to print, use verbose if you want to force print");
                         }
 
                         configDB.tags.find({ns: coll._id}).sort({min: 1}).forEach(function(tag) {
-                            output("\t\t\t tag: " + tag.tag + "  " + tojson(tag.min) + " -->> " +
-                                   tojson(tag.max));
+                            output(4,
+                                   " tag: " + tag.tag + "  " + tojson(tag.min) + " -->> " +
+                                       tojson(tag.max));
                         });
                     }
                 });
@@ -861,29 +880,29 @@ function printShardingSizes(configDB) {
     }
 
     var raw = "";
-    var output = function(s) {
-        raw += s + "\n";
+    var output = function(indent, s) {
+        raw += sh._shardingStatusStr(indent, s);
     };
-    output("--- Sharding Status --- ");
-    output("  sharding version: " + tojson(configDB.getCollection("version").findOne()));
+    output(0, "--- Sharding Sizes --- ");
+    output(1, "sharding version: " + tojson(configDB.getCollection("version").findOne()));
 
-    output("  shards:");
+    output(1, "shards:");
     var shards = {};
     configDB.shards.find().forEach(function(z) {
         shards[z._id] = new Mongo(z.host);
-        output("      " + tojson(z));
+        output(2, tojson(z));
     });
 
     var saveDB = db;
-    output("  databases:");
+    output(1, "databases:");
     configDB.databases.find().sort({name: 1}).forEach(function(db) {
-        output("\t" + tojson(db, "", true));
+        output(2, tojson(db, "", true));
 
         if (db.partitioned) {
             configDB.collections.find({_id: new RegExp("^" + RegExp.escape(db._id) + "\.")})
                 .sort({_id: 1})
                 .forEach(function(coll) {
-                    output("\t\t" + coll._id + " chunks:");
+                    output(3, coll._id + " chunks:");
                     configDB.chunks.find({"ns": coll._id}).sort({min: 1}).forEach(function(chunk) {
                         var mydb = shards[chunk.shard].getDB(db._id);
                         var out = mydb.runCommand({
@@ -895,8 +914,9 @@ function printShardingSizes(configDB) {
                         delete out.millis;
                         delete out.ok;
 
-                        output("\t\t\t" + tojson(chunk.min) + " -->> " + tojson(chunk.max) +
-                               " on : " + chunk.shard + " " + tojson(out));
+                        output(4,
+                               tojson(chunk.min) + " -->> " + tojson(chunk.max) + " on : " +
+                                   chunk.shard + " " + tojson(out));
 
                     });
                 });
diff --git a/src/mongo/transport/SConscript b/src/mongo/transport/SConscript
index 8e414610b2c..e82b6bd28fa 100644
--- a/src/mongo/transport/SConscript
+++ b/src/mongo/transport/SConscript
@@ -100,7 +100,9 @@ env.CppUnitTest(
     ],
 )
 
-env.Library(
+messageCompressorEnv = env.Clone()
+messageCompressorEnv.InjectThirdPartyIncludePaths(libraries=['snappy'])
+messageCompressorEnv.Library(
     target='message_compressor',
     source=[
         'message_compressor_manager.cpp',
diff --git a/src/mongo/transport/message_compressor_manager.cpp b/src/mongo/transport/message_compressor_manager.cpp
index 669b091f422..9a5b5ca81a9 100644
--- a/src/mongo/transport/message_compressor_manager.cpp
+++ b/src/mongo/transport/message_compressor_manager.cpp
@@ -51,23 +51,31 @@ struct CompressionHeader {
     uint8_t compressorId;
 
     void serialize(DataRangeCursor* cursor) {
-        cursor->writeAndAdvance<LittleEndian<int32_t>>(originalOpCode);
-        cursor->writeAndAdvance<LittleEndian<int32_t>>(uncompressedSize);
-        cursor->writeAndAdvance<LittleEndian<uint8_t>>(compressorId);
+        uassertStatusOK(cursor->writeAndAdvance<LittleEndian<int32_t>>(originalOpCode));
+        uassertStatusOK(cursor->writeAndAdvance<LittleEndian<int32_t>>(uncompressedSize));
+        uassertStatusOK(cursor->writeAndAdvance<LittleEndian<uint8_t>>(compressorId));
     }
 
     CompressionHeader(int32_t _opcode, int32_t _size, uint8_t _id)
         : originalOpCode{_opcode}, uncompressedSize{_size}, compressorId{_id} {}
 
     CompressionHeader(ConstDataRangeCursor* cursor) {
-        originalOpCode = cursor->readAndAdvance<LittleEndian<std::int32_t>>().getValue();
-        uncompressedSize = cursor->readAndAdvance<LittleEndian<std::int32_t>>().getValue();
-        compressorId = cursor->readAndAdvance<LittleEndian<uint8_t>>().getValue();
+        originalOpCode = _readWithChecking<LittleEndian<std::int32_t>>(cursor);
+        uncompressedSize = _readWithChecking<LittleEndian<std::int32_t>>(cursor);
+        compressorId = _readWithChecking<LittleEndian<uint8_t>>(cursor);
     }
 
     static size_t size() {
         return sizeof(originalOpCode) + sizeof(uncompressedSize) + sizeof(compressorId);
     }
+
+private:
+    template <typename T>
+    T _readWithChecking(ConstDataRangeCursor* cursor) {
+        auto sw = cursor->readAndAdvance<T>();
+        uassertStatusOK(sw.getStatus());
+        return sw.getValue();
+    }
 };
 }  // namespace
 
@@ -124,6 +132,9 @@ StatusWith<Message> MessageCompressorManager::compressMessage(const Message& msg
 StatusWith<Message> MessageCompressorManager::decompressMessage(const Message& msg) {
     auto inputHeader = msg.header();
     ConstDataRangeCursor input(inputHeader.data(), inputHeader.data() + inputHeader.dataLen());
+    if (input.length() < CompressionHeader::size()) {
+        return {ErrorCodes::BadValue, "Invalid compressed message header"};
+    }
     CompressionHeader compressionHeader(&input);
 
     auto compressor = _registry->getCompressor(compressionHeader.compressorId);
@@ -132,7 +143,12 @@ StatusWith<Message> MessageCompressorManager::decompressMessage(const Message& m
                 "Compression algorithm specified in message is not available"};
     }
 
-    auto bufferSize = compressionHeader.uncompressedSize + MsgData::MsgDataHeaderSize;
+    size_t bufferSize = compressionHeader.uncompressedSize + MsgData::MsgDataHeaderSize;
+    if (bufferSize > MaxMessageSizeBytes) {
+        return {ErrorCodes::BadValue,
+                "Decompressed message would be larger than maximum message size"};
+    }
+
     auto outputMessageBuffer = SharedBuffer::allocate(bufferSize);
     MsgData::View outMessage(outputMessageBuffer.get());
     outMessage.setId(inputHeader.getId());
diff --git a/src/mongo/transport/message_compressor_manager_test.cpp b/src/mongo/transport/message_compressor_manager_test.cpp
index 383bb1e3260..5c533e45e75 100644
--- a/src/mongo/transport/message_compressor_manager_test.cpp
+++ b/src/mongo/transport/message_compressor_manager_test.cpp
@@ -26,6 +26,8 @@
  *    it in the license file.
  */
 
+#define MONGO_LOG_DEFAULT_COMPONENT ::mongo::logger::LogComponent::kNetwork
+
 #include "mongo/platform/basic.h"
 
 #include "mongo/bson/bsonobjbuilder.h"
@@ -33,7 +35,9 @@
 #include "mongo/transport/message_compressor_manager.h"
 #include "mongo/transport/message_compressor_noop.h"
 #include "mongo/transport/message_compressor_registry.h"
+#include "mongo/transport/message_compressor_snappy.h"
 #include "mongo/unittest/unittest.h"
+#include "mongo/util/log.h"
 #include "mongo/util/net/message.h"
 
 #include <string>
@@ -120,6 +124,41 @@ void checkFidelity(const Message& msg, std::unique_ptr<MessageCompressorBase> co
     ASSERT_EQ(memcmp(decompressedMsgView.data(), originalView.data(), originalView.dataLen()), 0);
 }
 
+void checkOverflow(std::unique_ptr<MessageCompressorBase> compressor) {
+    // This is our test data that we're going to try to compress/decompress into a buffer that's
+    // way too small.
+    const std::string data =
+        "We embrace reality. We apply high-quality thinking and rigor."
+        "We have courage in our convictions but work hard to ensure biases "
+        "or personal beliefs do not get in the way of finding the best solution.";
+    ConstDataRange input(data.data(), data.size());
+
+    // This is our tiny buffer that should cause an error.
+    std::array<char, 16> smallBuffer;
+    DataRange smallOutput(smallBuffer.data(), smallBuffer.size());
+
+    // This is a normal sized buffer that we can store a compressed version of our test data safely
+    std::vector<char> normalBuffer;
+    normalBuffer.resize(compressor->getMaxCompressedSize(data.size()));
+    auto sws = compressor->compressData(input, DataRange(normalBuffer.data(), normalBuffer.size()));
+    ASSERT_OK(sws);
+    DataRange normalRange = DataRange(normalBuffer.data(), sws.getValue());
+
+    // Check that compressing the test data into a small buffer fails
+    ASSERT_NOT_OK(compressor->compressData(input, smallOutput));
+
+    // Check that decompressing compressed test data into a small buffer fails
+    ASSERT_NOT_OK(compressor->decompressData(normalRange, smallOutput));
+
+    // Check that decompressing a valid buffer that's missing data doesn't overflow the
+    // source buffer.
+    std::vector<char> scratch;
+    scratch.resize(data.size());
+    ConstDataRange tooSmallRange(normalBuffer.data(), normalBuffer.size() / 2);
+    ASSERT_NOT_OK(
+        compressor->decompressData(tooSmallRange, DataRange(scratch.data(), scratch.size())));
+}
+
 Message buildMessage() {
     const auto data = std::string{"Hello, world!"};
     const auto bufferSize = MsgData::MsgDataHeaderSize + data.size();
@@ -179,8 +218,54 @@ TEST(NoopMessageCompressor, Fidelity) {
 
 TEST(SnappyMessageCompressor, Fidelity) {
     auto testMessage = buildMessage();
-    checkFidelity(testMessage, stdx::make_unique<NoopMessageCompressor>());
+    checkFidelity(testMessage, stdx::make_unique<SnappyMessageCompressor>());
+}
+
+TEST(SnappyMessageCompressor, Overflow) {
+    checkOverflow(stdx::make_unique<SnappyMessageCompressor>());
+}
+
+TEST(MessageCompressorManager, MessageSizeTooLarge) {
+    auto registry = buildRegistry();
+    MessageCompressorManager compManager(&registry);
+
+    auto badMessageBuffer = SharedBuffer::allocate(128);
+    MsgData::View badMessage(badMessageBuffer.get());
+    badMessage.setId(1);
+    badMessage.setResponseToMsgId(0);
+    badMessage.setOperation(dbCompressed);
+    badMessage.setLen(128);
+
+    DataRangeCursor cursor(badMessage.data(), badMessage.data() + badMessage.dataLen());
+    uassertStatusOK(cursor.writeAndAdvance<LittleEndian<int32_t>>(dbQuery));
+    uassertStatusOK(cursor.writeAndAdvance<LittleEndian<int32_t>>(MaxMessageSizeBytes + 1));
+    uassertStatusOK(
+        cursor.writeAndAdvance<LittleEndian<uint8_t>>(registry.getCompressor("noop")->getId()));
+
+    auto status = compManager.decompressMessage(Message(badMessageBuffer)).getStatus();
+    ASSERT_NOT_OK(status);
+}
+
+TEST(MessageCompressorManager, RuntMessage) {
+    auto registry = buildRegistry();
+    MessageCompressorManager compManager(&registry);
+
+    auto badMessageBuffer = SharedBuffer::allocate(128);
+    MsgData::View badMessage(badMessageBuffer.get());
+    badMessage.setId(1);
+    badMessage.setResponseToMsgId(0);
+    badMessage.setOperation(dbCompressed);
+    badMessage.setLen(MsgData::MsgDataHeaderSize + 8);
+
+    // This is a totally bogus compression header of just the orginal opcode + 0 byte uncompressed
+    // size
+    DataRangeCursor cursor(badMessage.data(), badMessage.data() + badMessage.dataLen());
+    uassertStatusOK(cursor.writeAndAdvance<LittleEndian<int32_t>>(dbQuery));
+    uassertStatusOK(cursor.writeAndAdvance<LittleEndian<int32_t>>(0));
+
+    auto status = compManager.decompressMessage(Message(badMessageBuffer)).getStatus();
+    ASSERT_NOT_OK(status);
 }
 
-}  // namespace mongo
 }  // namespace
+}  // namespace mongo
diff --git a/src/mongo/transport/message_compressor_snappy.cpp b/src/mongo/transport/message_compressor_snappy.cpp
index db1e0c9dfca..9d523fca661 100644
--- a/src/mongo/transport/message_compressor_snappy.cpp
+++ b/src/mongo/transport/message_compressor_snappy.cpp
@@ -30,12 +30,13 @@
 
 #include "mongo/platform/basic.h"
 
+#include "mongo/base/data_range_cursor.h"
 #include "mongo/base/init.h"
 #include "mongo/stdx/memory.h"
 #include "mongo/transport/message_compressor_registry.h"
 #include "mongo/transport/message_compressor_snappy.h"
 
-#include "third_party/snappy-1.1.3/snappy.h"
+#include <snappy.h>
 
 namespace mongo {
 
@@ -48,7 +49,10 @@ std::size_t SnappyMessageCompressor::getMaxCompressedSize(size_t inputSize) {
 
 StatusWith<std::size_t> SnappyMessageCompressor::compressData(ConstDataRange input,
                                                               DataRange output) {
-    size_t outLength;
+    size_t outLength = output.length();
+    if (output.length() < getMaxCompressedSize(input.length())) {
+        return {ErrorCodes::BadValue, "Output too small for max size of compressed input"};
+    }
     snappy::RawCompress(input.data(), input.length(), const_cast<char*>(output.data()), &outLength);
 
     counterHitCompress(input.length(), outLength);
@@ -57,10 +61,13 @@ StatusWith<std::size_t> SnappyMessageCompressor::compressData(ConstDataRange inp
 
 StatusWith<std::size_t> SnappyMessageCompressor::decompressData(ConstDataRange input,
                                                                 DataRange output) {
-    bool ret =
-        snappy::RawUncompress(input.data(), input.length(), const_cast<char*>(output.data()));
+    size_t expectedLength = 0;
+    if (!snappy::GetUncompressedLength(input.data(), input.length(), &expectedLength) ||
+        expectedLength != output.length()) {
+        return {ErrorCodes::BadValue, "Compressed message was invalid or corrupted"};
+    }
 
-    if (!ret) {
+    if (!snappy::RawUncompress(input.data(), input.length(), const_cast<char*>(output.data()))) {
         return Status{ErrorCodes::BadValue, "Compressed message was invalid or corrupted"};
     }
 
diff --git a/src/mongo/util/SConscript b/src/mongo/util/SConscript
index 3346a596763..fe0380c1345 100644
--- a/src/mongo/util/SConscript
+++ b/src/mongo/util/SConscript
@@ -500,6 +500,17 @@ env.CppUnitTest(
 )
 
 env.CppUnitTest(
+    target='producer_consumer_queue_test',
+    source=[
+        'producer_consumer_queue_test.cpp',
+    ],
+    LIBDEPS=[
+        '$BUILD_DIR/mongo/base',
+        '$BUILD_DIR/mongo/db/service_context',
+    ]
+)
+
+env.CppUnitTest(
     target='duration_test',
     source=[
         'duration_test.cpp',
diff --git a/src/mongo/util/assert_util.cpp b/src/mongo/util/assert_util.cpp
index 9d40b1f5d62..500a8cbb5eb 100644
--- a/src/mongo/util/assert_util.cpp
+++ b/src/mongo/util/assert_util.cpp
@@ -232,10 +232,9 @@ NOINLINE_DECL void msgassertedWithLocation(int msgid,
                                            const char* msg,
                                            const char* file,
                                            unsigned line) {
-    assertionCount.condrollover(++assertionCount.warning);
+    assertionCount.condrollover(++assertionCount.msg);
     log() << "Assertion: " << msgid << ":" << redact(msg) << ' ' << file << ' ' << dec << line
           << endl;
-    logContext();
     throw MsgAssertionException(msgid, msg);
 }
 
@@ -243,7 +242,7 @@ NOINLINE_DECL void msgassertedNoTraceWithLocation(int msgid,
                                                   const char* msg,
                                                   const char* file,
                                                   unsigned line) {
-    assertionCount.condrollover(++assertionCount.warning);
+    assertionCount.condrollover(++assertionCount.msg);
     log() << "Assertion: " << msgid << ":" << redact(msg) << ' ' << file << ' ' << dec << line
           << endl;
     throw MsgAssertionException(msgid, msg);
diff --git a/src/mongo/util/cmdline_utils/censor_cmdline.cpp b/src/mongo/util/cmdline_utils/censor_cmdline.cpp
index 044de375e56..e903242ac75 100644
--- a/src/mongo/util/cmdline_utils/censor_cmdline.cpp
+++ b/src/mongo/util/cmdline_utils/censor_cmdline.cpp
@@ -45,6 +45,7 @@ static bool _isPasswordArgument(const char* argumentName) {
         "net.ssl.clusterPassword",
         "processManagement.windowsService.servicePassword",
         "security.kmip.clientCertificatePassword",
+        "security.ldap.bind.queryPassword",
         NULL  // Last entry sentinel.
     };
     for (const char* const* current = passwordArguments; *current; ++current) {
@@ -60,6 +61,7 @@ static bool _isPasswordSwitch(const char* switchName) {
         "sslClusterPassword",
         "servicePassword",
         "kmipClientCertificatePassword",
+        "ldapQueryPassword",
         NULL  // Last entry sentinel.
     };
 
diff --git a/src/mongo/util/concurrency/with_lock.h b/src/mongo/util/concurrency/with_lock.h
new file mode 100644
index 00000000000..e1607e361cf
--- /dev/null
+++ b/src/mongo/util/concurrency/with_lock.h
@@ -0,0 +1,111 @@
+/**    Copyright 2017 MongoDB, Inc.
+ *
+ *    This program is free software: you can redistribute it and/or  modify
+ *    it under the terms of the GNU Affero General Public License, version 3,
+ *    as published by the Free Software Foundation.
+ *
+ *    This program is distributed in the hope that it will be useful,
+ *    but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *    GNU Affero General Public License for more details.
+ *
+ *    You should have received a copy of the GNU Affero General Public License
+ *    along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ *    As a special exception, the copyright holders give permission to link the
+ *    code of portions of this program with the OpenSSL library under certain
+ *    conditions as described in each individual source file and distribute
+ *    linked combinations including the program with the OpenSSL library. You
+ *    must comply with the GNU Affero General Public License in all respects
+ *    for all of the code used other than as permitted herein. If you modify
+ *    file(s) with this exception, you may extend this exception to your
+ *    version of the file(s), but you are not obligated to do so. If you do not
+ *    wish to do so, delete this exception statement from your version. If you
+ *    delete this exception statement from all source files in the program,
+ *    then also delete it in the license file.
+ */
+
+#pragma once
+
+#include "mongo/stdx/mutex.h"
+#include "mongo/util/assert_util.h"
+
+#include <utility>
+
+namespace mongo {
+
+/**
+ * WithLock is an attestation to pass as an argument to functions that must be called only while
+ * holding a lock, as a rigorous alternative to an unchecked naming convention and/or stern
+ * comments.  It helps prevent a common usage error.
+ *
+ * It may be used to modernize code from (something like) this
+ *
+ *     // Member _mutex MUST be held when calling this:
+ *     void _clobber_inlock(OperationContext* opCtx) {
+ *         _stuff = makeStuff(opCtx);
+ *     }
+ *
+ * into
+ *
+ *     void _clobber(WithLock, OperationContext* opCtx) {
+ *         _stuff = makeStuff(opCtx);
+ *     }
+ *
+ * A call to such a function looks like this:
+ *
+ *     stdx::lock_guard<stdx::mutex> lk(_mutex);
+ *     _clobber(lk, opCtx);  // instead of _clobber_inlock(opCtx)
+ *
+ * Note that the formal argument need not (and should not) be named unless it is needed to pass
+ * the attestation along to another function:
+ *
+ *     void _clobber(WithLock lock, OperationContext* opCtx) {
+ *         _really_clobber(lock, opCtx);
+ *     }
+ *
+ */
+struct WithLock {
+    template <typename Mutex>
+    WithLock(stdx::lock_guard<Mutex> const&) noexcept {}
+
+    template <typename Mutex>
+    WithLock(stdx::unique_lock<Mutex> const& lock) noexcept {
+        invariant(lock.owns_lock());
+    }
+
+    // Add constructors from any other lock types here.
+
+    // Pass by value is OK.
+    WithLock(WithLock const&) noexcept {}
+    WithLock(WithLock&&) noexcept {}
+
+    // No assigning WithLocks.
+    void operator=(WithLock const&) = delete;
+    void operator=(WithLock&&) = delete;
+
+    // No moving a lock_guard<> or unique_lock<> in.
+    template <typename Mutex>
+    WithLock(stdx::lock_guard<Mutex>&&) = delete;
+    template <typename Mutex>
+    WithLock(stdx::unique_lock<Mutex>&&) = delete;
+
+    /*
+     * Produces a WithLock without benefit of any actual lock, for use in cases where a lock is not
+     * really needed, such as in many (but not all!) constructors.
+     */
+    static WithLock withoutLock() noexcept {
+        return {};
+    }
+
+private:
+    WithLock() noexcept = default;
+};
+
+}  // namespace mongo
+
+namespace std {
+// No moving a WithLock:
+template <>
+mongo::WithLock&& move<mongo::WithLock>(mongo::WithLock&&) noexcept = delete;
+}  // namespace std
diff --git a/src/mongo/util/net/hostandport.cpp b/src/mongo/util/net/hostandport.cpp
index abd8f2d6ade..4b46938fbdb 100644
--- a/src/mongo/util/net/hostandport.cpp
+++ b/src/mongo/util/net/hostandport.cpp
@@ -95,7 +95,9 @@ void HostAndPort::append(StringBuilder& ss) const {
     } else {
         ss << host();
     }
-    ss << ':' << port();
+    if (host().find('/') == std::string::npos) {
+        ss << ':' << port();
+    }
 }
 
 bool HostAndPort::empty() const {
diff --git a/src/mongo/util/net/miniwebserver.cpp b/src/mongo/util/net/miniwebserver.cpp
index 5f4165d42f4..9fe847168ee 100644
--- a/src/mongo/util/net/miniwebserver.cpp
+++ b/src/mongo/util/net/miniwebserver.cpp
@@ -45,6 +45,7 @@
 namespace mongo {
 
 using std::shared_ptr;
+using std::string;
 using std::stringstream;
 using std::vector;
 
diff --git a/src/mongo/util/net/ssl_manager.cpp b/src/mongo/util/net/ssl_manager.cpp
index 187c0c105e2..4bb0befb4b0 100644
--- a/src/mongo/util/net/ssl_manager.cpp
+++ b/src/mongo/util/net/ssl_manager.cpp
@@ -37,6 +37,7 @@
 #include <boost/thread/thread.hpp>
 #include <iostream>
 #include <sstream>
+#include <stack>
 #include <string>
 #include <vector>
 
@@ -171,21 +172,28 @@ const STACK_OF(X509_EXTENSION) * X509_get0_extensions(const X509* peerCert) {
  * OpenSSL before version 1.1.0 requires applications provide a callback which emits a thread
  * identifier. This ID is used to store thread specific ERR information. When a thread is
  * terminated, it must call ERR_remove_state or ERR_remove_thread_state. These functions may
- * themselves invoke the application provided callback.
+ * themselves invoke the application provided callback. These IDs are stored in a hashtable with
+ * a questionable hash function. They must be uniformly distributed to prevent collisions.
  */
 class SSLThreadInfo {
 public:
     static unsigned long getID() {
-        enforceCleanupOnShutdown();
+        struct ThreadLocalState {
+            bool firstCall;
+            unsigned long id;
+        };
+        static MONGO_TRIVIALLY_CONSTRUCTIBLE_THREAD_LOCAL ThreadLocalState state{true, 0};
+        if (state.firstCall) {
+            state.id = _idManager.reserveID();
+            unsigned long id = state.id;
+            boost::this_thread::at_thread_exit([id] {
+                ERR_remove_state(0);
+                _idManager.releaseID(id);
+            });
+            state.firstCall = false;
+        }
 
-#ifdef _WIN32
-        return GetCurrentThreadId();
-#else
-        static_assert(sizeof(void*) == sizeof(unsigned long),
-                      "OpenSSL needs the address of a thread-unique object to be castable to"
-                      "unsigned long");
-        return reinterpret_cast<unsigned long>(&errno);
-#endif
+        return state.id;
     }
 
     static void lockingCallback(int mode, int type, const char* file, int line) {
@@ -208,21 +216,39 @@ public:
 private:
     SSLThreadInfo() = delete;
 
-    // When called, ensures that this thread will, on termination, call ERR_remove_state.
-    static void enforceCleanupOnShutdown() {
-        static MONGO_TRIVIALLY_CONSTRUCTIBLE_THREAD_LOCAL bool firstCall = true;
-        if (firstCall) {
-            boost::this_thread::at_thread_exit([] { ERR_remove_state(0); });
-            firstCall = false;
-        }
-    }
-
     // Note: see SERVER-8734 for why we are using a recursive mutex here.
     // Once the deadlock fix in OpenSSL is incorporated into most distros of
     // Linux, this can be changed back to a nonrecursive mutex.
     static std::vector<std::unique_ptr<stdx::recursive_mutex>> _mutex;
+
+    class ThreadIDManager {
+    public:
+        unsigned long reserveID() {
+            stdx::unique_lock<stdx::mutex> lock(_idMutex);
+            if (!_idLast.empty()) {
+                unsigned long ret = _idLast.top();
+                _idLast.pop();
+                return ret;
+            }
+            return ++_idNext;
+        }
+
+        void releaseID(unsigned long id) {
+            stdx::unique_lock<stdx::mutex> lock(_idMutex);
+            _idLast.push(id);
+        }
+
+    private:
+        // Machinery for producing IDs that are unique for the life of a thread.
+        stdx::mutex _idMutex;       // Protects _idNext and _idLast.
+        unsigned long _idNext = 0;  // Stores the next thread ID to use, if none already allocated.
+        std::stack<unsigned long, std::vector<unsigned long>>
+            _idLast;  // Stores old thread IDs, for reuse.
+    };
+    static ThreadIDManager _idManager;
 };
 std::vector<std::unique_ptr<stdx::recursive_mutex>> SSLThreadInfo::_mutex;
+SSLThreadInfo::ThreadIDManager SSLThreadInfo::_idManager;
 
 // We only want to free SSL_CTX objects if they have been populated. OpenSSL seems to perform this
 // check before freeing them, but because it does not document this, we should protect ourselves.
diff --git a/src/mongo/util/options_parser/options_parser.cpp b/src/mongo/util/options_parser/options_parser.cpp
index 5d03910a390..63bf184ec85 100644
--- a/src/mongo/util/options_parser/options_parser.cpp
+++ b/src/mongo/util/options_parser/options_parser.cpp
@@ -341,6 +341,11 @@ Status YAMLNodeToValue(const YAML::Node& YAMLNode,
         return Status::OK();
     }
 
+    if (!YAMLNode.IsScalar() && !YAMLNode.IsNull()) {
+        return Status(ErrorCodes::BadValue,
+                      str::stream() << "Scalar option '" << key << "' must be a single value");
+    }
+
     // Our YAML parser reads everything as a string, so we need to parse it ourselves.
     std::string stringVal = YAMLNode.Scalar();
     return stringToValue(stringVal, type, key, value);
diff --git a/src/mongo/util/options_parser/options_parser_test.cpp b/src/mongo/util/options_parser/options_parser_test.cpp
index a0d97c512bb..aeb4a875914 100644
--- a/src/mongo/util/options_parser/options_parser_test.cpp
+++ b/src/mongo/util/options_parser/options_parser_test.cpp
@@ -1834,6 +1834,82 @@ TEST(Parsing, BadConfigFileOption) {
     ASSERT_NOT_OK(parser.run(testOpts, argv, env_map, &environment));
 }
 
+TEST(Parsing, MapForScalarMismatch) {
+    OptionsParserTester parser;
+    moe::Environment environment;
+    moe::OptionSection testOpts;
+
+    testOpts.addOptionChaining("config", "config", moe::Int, "Config file to parse");
+    testOpts.addOptionChaining("str", "str", moe::String, "");
+
+    std::vector<std::string> argv;
+    argv.push_back("binaryname");
+    argv.push_back("--config");
+    argv.push_back("config.json");
+    std::map<std::string, std::string> env_map;
+
+    parser.setConfig("config.json", R"cfg({ str: { elem: "val" } })cfg");
+
+    ASSERT_NOT_OK(parser.run(testOpts, argv, env_map, &environment));
+}
+
+TEST(Parsing, ScalarForMapMismatch) {
+    OptionsParserTester parser;
+    moe::Environment environment;
+    moe::OptionSection testOpts;
+
+    testOpts.addOptionChaining("config", "config", moe::Int, "Config file to parse");
+    testOpts.addOptionChaining("strmap", "strmap", moe::StringMap, "");
+
+    std::vector<std::string> argv;
+    argv.push_back("binaryname");
+    argv.push_back("--config");
+    argv.push_back("config.json");
+    std::map<std::string, std::string> env_map;
+
+    parser.setConfig("config.json", R"cfg({ str: "val" })cfg");
+
+    ASSERT_NOT_OK(parser.run(testOpts, argv, env_map, &environment));
+}
+
+TEST(Parsing, ListForScalarMismatch) {
+    OptionsParserTester parser;
+    moe::Environment environment;
+    moe::OptionSection testOpts;
+
+    testOpts.addOptionChaining("config", "config", moe::Int, "Config file to parse");
+    testOpts.addOptionChaining("str", "str", moe::String, "");
+
+    std::vector<std::string> argv;
+    argv.push_back("binaryname");
+    argv.push_back("--config");
+    argv.push_back("config.json");
+    std::map<std::string, std::string> env_map;
+
+    parser.setConfig("config.json", R"cfg({ str: ["val"] })cfg");
+
+    ASSERT_NOT_OK(parser.run(testOpts, argv, env_map, &environment));
+}
+
+TEST(Parsing, ScalarForListMismatch) {
+    OptionsParserTester parser;
+    moe::Environment environment;
+    moe::OptionSection testOpts;
+
+    testOpts.addOptionChaining("config", "config", moe::Int, "Config file to parse");
+    testOpts.addOptionChaining("strlist", "strlist", moe::StringVector, "");
+
+    std::vector<std::string> argv;
+    argv.push_back("binaryname");
+    argv.push_back("--config");
+    argv.push_back("config.json");
+    std::map<std::string, std::string> env_map;
+
+    parser.setConfig("config.json", R"cfg({ str: "val" })cfg");
+
+    ASSERT_NOT_OK(parser.run(testOpts, argv, env_map, &environment));
+}
+
 TEST(ConfigFromFilesystem, JSONGood) {
     moe::OptionsParser parser;
     moe::Environment environment;
diff --git a/src/mongo/util/password.cpp b/src/mongo/util/password.cpp
index 11c936bac45..4814cbb3b12 100644
--- a/src/mongo/util/password.cpp
+++ b/src/mongo/util/password.cpp
@@ -46,7 +46,7 @@ namespace mongo {
 
 string askPassword() {
     std::string password;
-    cout << "Enter password: ";
+    cerr << "Enter password: ";
 #ifndef _WIN32
     const int stdinfd = 0;
     termios termio;
@@ -102,7 +102,7 @@ string askPassword() {
         return string();
     }
 #endif
-    cout << "\n";
+    cerr << "\n";
     return password;
 }
 }
diff --git a/src/mongo/util/producer_consumer_queue.h b/src/mongo/util/producer_consumer_queue.h
new file mode 100644
index 00000000000..f0e942a7d94
--- /dev/null
+++ b/src/mongo/util/producer_consumer_queue.h
@@ -0,0 +1,559 @@
+/**
+ * Copyright (C) 2018 MongoDB Inc.
+ *
+ * This program is free software: you can redistribute it and/or  modify
+ * it under the terms of the GNU Affero General Public License, version 3,
+ * as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU Affero General Public License for more details.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * As a special exception, the copyright holders give permission to link the
+ * code of portions of this program with the OpenSSL library under certain
+ * conditions as described in each individual source file and distribute
+ * linked combinations including the program with the OpenSSL library. You
+ * must comply with the GNU Affero General Public License in all respects
+ * for all of the code used other than as permitted herein. If you modify
+ * file(s) with this exception, you may extend this exception to your
+ * version of the file(s), but you are not obligated to do so. If you do not
+ * wish to do so, delete this exception statement from your version. If you
+ * delete this exception statement from all source files in the program,
+ * then also delete it in the license file.
+ */
+
+#pragma once
+
+#include <boost/optional.hpp>
+#include <deque>
+#include <list>
+#include <queue>
+#include <stack>
+
+#include "mongo/db/operation_context.h"
+#include "mongo/stdx/condition_variable.h"
+#include "mongo/stdx/mutex.h"
+#include "mongo/stdx/type_traits.h"
+#include "mongo/util/concurrency/with_lock.h"
+#include "mongo/util/scopeguard.h"
+
+namespace mongo {
+
+namespace producer_consumer_queue_detail {
+
+/**
+ * The default cost function for the producer consumer queue.
+ *
+ * By default, all items in the queue have equal weight.
+ */
+struct DefaultCostFunction {
+    template <typename T>
+    size_t operator()(const T&) const {
+        return 1;
+    }
+};
+
+// Various helpers to tighten down whether the args getting passed are valid interruption args.
+//
+// Whatever the caller passes in the interruption args, they need to be invocable on one of
+// these helpers. std::is_invocable would do the job in C++17
+std::false_type areInterruptionArgsHelper(...) {
+    return {};
+}
+
+std::true_type areInterruptionArgsHelper(OperationContext*) {
+    return {};
+}
+
+std::true_type areInterruptionArgsHelper(OperationContext*, Date_t) {
+    return {};
+}
+
+std::true_type areInterruptionArgsHelper(Milliseconds) {
+    return {};
+}
+
+std::true_type areInterruptionArgsHelper(Date_t) {
+    return {};
+}
+
+template <typename U, typename... InterruptionArgs>
+auto areInterruptionArgs(U&& u, InterruptionArgs&&... args)
+    -> decltype(areInterruptionArgsHelper(std::forward<U>(u),
+                                          std::forward<InterruptionArgs>(args)...)) {
+    return areInterruptionArgsHelper(std::forward<U>(u), std::forward<InterruptionArgs>(args)...);
+}
+
+std::true_type areInterruptionArgs() {
+    return {};
+}
+
+}  // namespace producer_consumer_queue_detail
+
+/**
+ * A bounded, blocking, thread safe, cost parametrizable, single producer, multi-consumer queue.
+ *
+ * Properties:
+ *   bounded - the queue can be limited in the number of items it can hold
+ *   blocking - when the queue is full, or has no entries, callers block
+ *   thread safe - the queue can be accessed safely from multiple threads at the same time
+ *   cost parametrizable - the cost of items in the queue need not be equal. I.e. your items could
+ *                          be discrete byte buffers and the queue depth measured in bytes, so that
+ *                          the queue could hold one large buffer, or many smaller ones
+ *   single producer - Only one thread may push work into the queue
+ *   multi-consumer - Any number of threads may pop work out of the queue
+ *
+ * Interruptibility:
+ *   All of the blocking methods on this type allow for 6 kinds of interruptibility. The matrix is
+ *   parameterized by (void|OperationContext*)|(void|Milliseconds|Date_t). These provide different
+ *   kinds of waiting based on whether the method should be interruptible via opCtx, and then
+ *   whether they should timeout via deadline or duration
+ *
+ *   A contrived example: pcq.pop(opCtx, Minutes(1)) would be warranted if:
+ *     - The caller is blocking on a client thread. (opCtx)
+ *     - The caller needs to act periodically on inactivity. (the duration)
+ *
+ * Exceptions include:
+ *   timeouts
+ *     ErrorCodes::ExceededTimeLimit exceptions
+ *   opCtx interrupts
+ *     ErrorCodes::Interrupted exceptions
+ *   closure of queue endpoints
+ *     ErrorCodes::ProducerConsumerQueueEndClosed
+ *   pushes with batches that exceed the max queue size
+ *     ErrorCodes::ProducerConsumerQueueBatchTooLarge
+ *
+ * Cost Function:
+ *   The cost function must have a call operator which takes a const T& and returns the cost in
+ *   size_t units. It must be pure across moves for a given T and never return zero. The intent of
+ *   the cost function is to express the kind of bounds the queue provides, rather than to
+ *   specialize behavior for a type. I.e. you should not specialize the default cost function and
+ *   the cost function should always be explicit in the type.
+ */
+template <typename T, typename CostFunc = producer_consumer_queue_detail::DefaultCostFunction>
+class ProducerConsumerQueue {
+
+public:
+    // By default the queue depth is unlimited
+    ProducerConsumerQueue()
+        : ProducerConsumerQueue(std::numeric_limits<size_t>::max(), CostFunc{}) {}
+
+    // Or it can be measured in whatever units your size function returns
+    explicit ProducerConsumerQueue(size_t size) : ProducerConsumerQueue(size, CostFunc{}) {}
+
+    // If your cost function has meaningful state, you may also pass a non-default constructed
+    // instance
+    explicit ProducerConsumerQueue(size_t size, CostFunc costFunc)
+        : _max(size), _costFunc(std::move(costFunc)) {}
+
+    ProducerConsumerQueue(const ProducerConsumerQueue&) = delete;
+    ProducerConsumerQueue& operator=(const ProducerConsumerQueue&) = delete;
+
+    ProducerConsumerQueue(ProducerConsumerQueue&&) = delete;
+    ProducerConsumerQueue& operator=(ProducerConsumerQueue&&) = delete;
+
+    ~ProducerConsumerQueue() {
+        invariant(!_producerWants);
+        invariant(!_consumers);
+    }
+
+    // Pushes the passed T into the queue
+    //
+    // Leaves T unchanged if an interrupt exception is thrown while waiting for space
+    template <
+        typename... InterruptionArgs,
+        typename = stdx::enable_if_t<decltype(producer_consumer_queue_detail::areInterruptionArgs(
+            std::declval<InterruptionArgs>()...))::value>>
+    void push(T&& t, InterruptionArgs&&... interruptionArgs) {
+        _pushRunner([&](stdx::unique_lock<stdx::mutex>& lk) {
+            auto cost = _invokeCostFunc(t, lk);
+            uassert(ErrorCodes::ProducerConsumerQueueBatchTooLarge,
+                    str::stream() << "cost of item (" << cost
+                                  << ") larger than maximum queue size ("
+                                  << _max
+                                  << ")",
+                    cost <= _max);
+
+            _waitForSpace(lk, cost, std::forward<InterruptionArgs>(interruptionArgs)...);
+            _push(lk, std::move(t));
+        });
+    }
+
+    // Pushes all Ts into the queue
+    //
+    // Blocks until all of the Ts can be pushed at once
+    //
+    // StartIterator must be ForwardIterator
+    //
+    // Leaves the values underneath the iterators unchanged if an interrupt exception is thrown
+    // while waiting for space
+    //
+    // Lifecycle methods of T must not throw if you want to use this method, as there's no obvious
+    // mechanism to see what was and was not pushed if those do throw
+    template <
+        typename StartIterator,
+        typename EndIterator,
+        typename... InterruptionArgs,
+        typename = stdx::enable_if_t<decltype(producer_consumer_queue_detail::areInterruptionArgs(
+            std::declval<InterruptionArgs>()...))::value>>
+    void pushMany(StartIterator start, EndIterator last, InterruptionArgs&&... interruptionArgs) {
+        return _pushRunner([&](stdx::unique_lock<stdx::mutex>& lk) {
+            size_t cost = 0;
+            for (auto iter = start; iter != last; ++iter) {
+                cost += _invokeCostFunc(*iter, lk);
+            }
+
+            uassert(ErrorCodes::ProducerConsumerQueueBatchTooLarge,
+                    str::stream() << "cost of items in batch (" << cost
+                                  << ") larger than maximum queue size ("
+                                  << _max
+                                  << ")",
+                    cost <= _max);
+
+            _waitForSpace(lk, cost, std::forward<InterruptionArgs>(interruptionArgs)...);
+
+            for (auto iter = start; iter != last; ++iter) {
+                _push(lk, std::move(*iter));
+            }
+        });
+    }
+
+    // Attempts a non-blocking push of a value
+    //
+    // Leaves T unchanged if it fails
+    bool tryPush(T&& t) {
+        return _pushRunner(
+            [&](stdx::unique_lock<stdx::mutex>& lk) { return _tryPush(lk, std::move(t)); });
+    }
+
+    // Pops one T out of the queue
+    template <
+        typename... InterruptionArgs,
+        typename = stdx::enable_if_t<decltype(producer_consumer_queue_detail::areInterruptionArgs(
+            std::declval<InterruptionArgs>()...))::value>>
+    T pop(InterruptionArgs&&... interruptionArgs) {
+        return _popRunner([&](stdx::unique_lock<stdx::mutex>& lk) {
+            _waitForNonEmpty(lk, std::forward<InterruptionArgs>(interruptionArgs)...);
+            return _pop(lk);
+        });
+    }
+
+    // Waits for at least one item in the queue, then pops items out of the queue until it would
+    // block
+    //
+    // OutputIterator must not throw on move assignment to *iter or popped values may be lost
+    // TODO: add sfinae to check to enforce
+    //
+    // Returns the cost value of the items extracted, along with the updated output iterator
+    template <
+        typename OutputIterator,
+        typename... InterruptionArgs,
+        typename = stdx::enable_if_t<decltype(producer_consumer_queue_detail::areInterruptionArgs(
+            std::declval<InterruptionArgs>()...))::value>>
+    std::pair<size_t, OutputIterator> popMany(OutputIterator iterator,
+                                              InterruptionArgs&&... interruptionArgs) {
+        return popManyUpTo(_max, iterator, std::forward<InterruptionArgs>(interruptionArgs)...);
+    }
+
+    // Waits for at least one item in the queue, then pops items out of the queue until it would
+    // block, or we've exceeded our budget
+    //
+    // OutputIterator must not throw on move assignment to *iter or popped values may be lost
+    // TODO: add sfinae to check to enforce
+    //
+    // Returns the cost value of the items extracted, along with the updated output iterator
+    template <
+        typename OutputIterator,
+        typename... InterruptionArgs,
+        typename = stdx::enable_if_t<decltype(producer_consumer_queue_detail::areInterruptionArgs(
+            std::declval<InterruptionArgs>()...))::value>>
+    std::pair<size_t, OutputIterator> popManyUpTo(size_t budget,
+                                                  OutputIterator iterator,
+                                                  InterruptionArgs&&... interruptionArgs) {
+        return _popRunner([&](stdx::unique_lock<stdx::mutex>& lk) {
+            size_t cost = 0;
+
+            _waitForNonEmpty(lk, std::forward<InterruptionArgs>(interruptionArgs)...);
+
+            while (auto out = _tryPop(lk)) {
+                cost += _invokeCostFunc(*out, lk);
+                *iterator = std::move(*out);
+                ++iterator;
+
+                if (cost >= budget) {
+                    break;
+                }
+            }
+
+            return std::make_pair(cost, iterator);
+        });
+    }
+
+    // Attempts a non-blocking pop of a value
+    boost::optional<T> tryPop() {
+        return _popRunner([&](stdx::unique_lock<stdx::mutex>& lk) { return _tryPop(lk); });
+    }
+
+    // Closes the producer end. Consumers will continue to consume until the queue is exhausted, at
+    // which time they will begin to throw with an interruption dbexception
+    void closeProducerEnd() {
+        stdx::lock_guard<stdx::mutex> lk(_mutex);
+
+        _producerEndClosed = true;
+
+        _notifyIfNecessary(lk);
+    }
+
+    // Closes the consumer end. This causes all callers to throw with an interruption dbexception
+    void closeConsumerEnd() {
+        stdx::lock_guard<stdx::mutex> lk(_mutex);
+
+        _consumerEndClosed = true;
+        _producerEndClosed = true;
+
+        _notifyIfNecessary(lk);
+    }
+
+    // TEST ONLY FUNCTIONS
+
+    // Returns the current depth of the queue in CostFunction units
+    size_t sizeForTest() const {
+        stdx::lock_guard<stdx::mutex> lk(_mutex);
+
+        return _current;
+    }
+
+    // Returns true if the queue is empty
+    bool emptyForTest() const {
+        return sizeForTest() == 0;
+    }
+
+private:
+    size_t _invokeCostFunc(const T& t, WithLock) {
+        auto cost = _costFunc(t);
+        invariant(cost);
+        return cost;
+    }
+
+    void _checkProducerClosed(WithLock) {
+        uassert(
+            ErrorCodes::ProducerConsumerQueueEndClosed, "Producer end closed", !_producerEndClosed);
+        uassert(
+            ErrorCodes::ProducerConsumerQueueEndClosed, "Consumer end closed", !_consumerEndClosed);
+    }
+
+    void _checkConsumerClosed(WithLock) {
+        uassert(
+            ErrorCodes::ProducerConsumerQueueEndClosed, "Consumer end closed", !_consumerEndClosed);
+        uassert(ErrorCodes::ProducerConsumerQueueEndClosed,
+                "Producer end closed and values exhausted",
+                !(_producerEndClosed && _queue.empty()));
+    }
+
+    void _notifyIfNecessary(WithLock) {
+        // If we've closed the consumer end, or if the production end is closed and we've exhausted
+        // the queue, wake everyone up and get out of here
+        if (_consumerEndClosed || (_queue.empty() && _producerEndClosed)) {
+            if (_consumers) {
+                _condvarConsumer.notify_all();
+            }
+
+            if (_producerWants) {
+                _condvarProducer.notify_one();
+            }
+
+            return;
+        }
+
+        // If a producer is queued, and we have enough space for it to push its work
+        if (_producerWants && _current + _producerWants <= _max) {
+            _condvarProducer.notify_one();
+
+            return;
+        }
+
+        // If we have consumers and anything in the queue, notify consumers
+        if (_consumers && _queue.size()) {
+            _condvarConsumer.notify_one();
+
+            return;
+        }
+    }
+
+    template <typename Callback>
+    auto _pushRunner(Callback&& cb) -> decltype(
+        cb(std::declval<std::add_lvalue_reference<stdx::unique_lock<stdx::mutex>>::type>())) {
+        stdx::unique_lock<stdx::mutex> lk(_mutex);
+
+        _checkProducerClosed(lk);
+
+        const auto guard = MakeGuard([&] { _notifyIfNecessary(lk); });
+
+        return cb(lk);
+    }
+
+    template <typename Callback>
+    auto _popRunner(Callback&& cb) -> decltype(
+        cb(std::declval<std::add_lvalue_reference<stdx::unique_lock<stdx::mutex>>::type>())) {
+        stdx::unique_lock<stdx::mutex> lk(_mutex);
+
+        _checkConsumerClosed(lk);
+
+        const auto guard = MakeGuard([&] { _notifyIfNecessary(lk); });
+
+        return cb(lk);
+    }
+
+    bool _tryPush(WithLock wl, T&& t) {
+        size_t cost = _invokeCostFunc(t, wl);
+        if (_current + cost <= _max) {
+            _queue.emplace(std::move(t));
+            _current += cost;
+            return true;
+        }
+
+        return false;
+    }
+
+    void _push(WithLock wl, T&& t) {
+        size_t cost = _invokeCostFunc(t, wl);
+        invariant(_current + cost <= _max);
+
+        _queue.emplace(std::move(t));
+        _current += cost;
+    }
+
+    boost::optional<T> _tryPop(WithLock wl) {
+        boost::optional<T> out;
+
+        if (!_queue.empty()) {
+            out.emplace(std::move(_queue.front()));
+            _queue.pop();
+            _current -= _invokeCostFunc(*out, wl);
+        }
+
+        return out;
+    }
+
+    T _pop(WithLock wl) {
+        invariant(_queue.size());
+
+        auto t = std::move(_queue.front());
+        _queue.pop();
+
+        _current -= _invokeCostFunc(t, wl);
+
+        return t;
+    }
+
+    template <typename... InterruptionArgs>
+    void _waitForSpace(stdx::unique_lock<stdx::mutex>& lk,
+                       size_t cost,
+                       InterruptionArgs&&... interruptionArgs) {
+        invariant(!_producerWants);
+
+        _producerWants = cost;
+        const auto guard = MakeGuard([&] { _producerWants = 0; });
+
+        _waitFor(lk,
+                 _condvarProducer,
+                 [&] {
+                     _checkProducerClosed(lk);
+                     return _current + cost <= _max;
+                 },
+                 std::forward<InterruptionArgs>(interruptionArgs)...);
+    }
+
+    template <typename... InterruptionArgs>
+    void _waitForNonEmpty(stdx::unique_lock<stdx::mutex>& lk,
+                          InterruptionArgs&&... interruptionArgs) {
+
+        _consumers++;
+        const auto guard = MakeGuard([&] { _consumers--; });
+
+        _waitFor(lk,
+                 _condvarConsumer,
+                 [&] {
+                     _checkConsumerClosed(lk);
+                     return _queue.size();
+                 },
+                 std::forward<InterruptionArgs>(interruptionArgs)...);
+    }
+
+    template <typename Callback>
+    void _waitFor(stdx::unique_lock<stdx::mutex>& lk,
+                  stdx::condition_variable& condvar,
+                  Callback&& pred,
+                  OperationContext* opCtx) {
+        opCtx->waitForConditionOrInterrupt(condvar, lk, pred);
+    }
+
+    template <typename Callback>
+    void _waitFor(stdx::unique_lock<stdx::mutex>& lk,
+                  stdx::condition_variable& condvar,
+                  Callback&& pred) {
+        condvar.wait(lk, pred);
+    }
+
+    template <typename Callback>
+    void _waitFor(stdx::unique_lock<stdx::mutex>& lk,
+                  stdx::condition_variable& condvar,
+                  Callback&& pred,
+                  OperationContext* opCtx,
+                  Date_t deadline) {
+        uassert(ErrorCodes::ExceededTimeLimit,
+                "exceeded timeout",
+                opCtx->waitForConditionOrInterruptUntil(condvar, lk, deadline, pred));
+    }
+
+    template <typename Callback>
+    void _waitFor(stdx::unique_lock<stdx::mutex>& lk,
+                  stdx::condition_variable& condvar,
+                  Callback&& pred,
+                  Date_t deadline) {
+        uassert(ErrorCodes::ExceededTimeLimit,
+                "exceeded timeout",
+                condvar.wait_until(lk, deadline.toSystemTimePoint(), pred));
+    }
+
+    template <typename Callback>
+    void _waitFor(stdx::unique_lock<stdx::mutex>& lk,
+                  stdx::condition_variable& condvar,
+                  Callback&& pred,
+                  Milliseconds duration) {
+        uassert(ErrorCodes::ExceededTimeLimit,
+                "exceeded timeout",
+                condvar.wait_for(lk, duration.toSystemDuration(), pred));
+    }
+
+    mutable stdx::mutex _mutex;
+    stdx::condition_variable _condvarConsumer;
+    stdx::condition_variable _condvarProducer;
+
+    // Max size of the queue
+    const size_t _max;
+
+    // User's cost function
+    CostFunc _costFunc;
+
+    // Current size of the queue
+    size_t _current = 0;
+
+    std::queue<T> _queue;
+
+    // Counter for consumers in the queue
+    size_t _consumers = 0;
+
+    // Size of batch the blocking producer wants to insert
+    size_t _producerWants = 0;
+
+    // Flags that we're shutting down the queue
+    bool _consumerEndClosed = false;
+    bool _producerEndClosed = false;
+};
+
+}  // namespace mongo
diff --git a/src/mongo/util/producer_consumer_queue_test.cpp b/src/mongo/util/producer_consumer_queue_test.cpp
new file mode 100644
index 00000000000..f824b5a63de
--- /dev/null
+++ b/src/mongo/util/producer_consumer_queue_test.cpp
@@ -0,0 +1,531 @@
+/**
+ * Copyright (C) 2018 MongoDB Inc.
+ *
+ * This program is free software: you can redistribute it and/or  modify
+ * it under the terms of the GNU Affero General Public License, version 3,
+ * as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU Affero General Public License for more details.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * As a special exception, the copyright holders give permission to link the
+ * code of portions of this program with the OpenSSL library under certain
+ * conditions as described in each individual source file and distribute
+ * linked combinations including the program with the OpenSSL library. You
+ * must comply with the GNU Affero General Public License in all respects
+ * for all of the code used other than as permitted herein. If you modify
+ * file(s) with this exception, you may extend this exception to your
+ * version of the file(s), but you are not obligated to do so. If you do not
+ * wish to do so, delete this exception statement from your version. If you
+ * delete this exception statement from all source files in the program,
+ * then also delete it in the license file.
+ */
+
+#include "mongo/platform/basic.h"
+
+#include "mongo/unittest/unittest.h"
+
+#include "mongo/util/producer_consumer_queue.h"
+
+#include "mongo/db/service_context_noop.h"
+#include "mongo/stdx/condition_variable.h"
+#include "mongo/stdx/memory.h"
+#include "mongo/stdx/mutex.h"
+#include "mongo/stdx/thread.h"
+#include "mongo/util/assert_util.h"
+
+namespace mongo {
+
+namespace {
+
+class ProducerConsumerQueueTest : public unittest::Test {
+public:
+    ProducerConsumerQueueTest() : _serviceCtx(stdx::make_unique<ServiceContextNoop>()) {}
+
+    template <typename Callback>
+    stdx::thread runThread(StringData name, Callback&& cb) {
+        return stdx::thread([this, name, cb] { cb(); });
+    }
+
+private:
+    std::unique_ptr<ServiceContext> _serviceCtx;
+};
+
+class MoveOnly {
+public:
+    struct CostFunc {
+        CostFunc() = default;
+        explicit CostFunc(size_t val) : val(val) {}
+
+        size_t operator()(const MoveOnly& mo) const {
+            return val + *mo._val;
+        }
+
+        const size_t val = 0;
+    };
+
+    explicit MoveOnly(int i) : _val(i) {}
+
+    MoveOnly(const MoveOnly&) = delete;
+    MoveOnly& operator=(const MoveOnly&) = delete;
+
+    MoveOnly(MoveOnly&& other) : _val(other._val) {
+        other._val.reset();
+    }
+
+    MoveOnly& operator=(MoveOnly&& other) {
+        if (&other == this) {
+            return *this;
+        }
+
+        _val = other._val;
+        other._val.reset();
+
+        return *this;
+    }
+
+    bool movedFrom() const {
+        return !_val;
+    }
+
+    friend bool operator==(const MoveOnly& lhs, const MoveOnly& rhs) {
+        return *lhs._val == *rhs._val;
+    }
+
+    friend bool operator!=(const MoveOnly& lhs, const MoveOnly& rhs) {
+        return !(lhs == rhs);
+    }
+
+    friend std::ostream& operator<<(std::ostream& os, const MoveOnly& mo) {
+        return (os << "MoveOnly(" << *mo._val << ")");
+    }
+
+private:
+    boost::optional<int> _val;
+};
+
+TEST_F(ProducerConsumerQueueTest, basicPushPop) {
+    ProducerConsumerQueue<MoveOnly> pcq{};
+
+    runThread("Producer", [&]() { pcq.push(MoveOnly(1)); }).join();
+
+    ASSERT_EQUALS(pcq.sizeForTest(), 1ul);
+
+    runThread("Consumer", [&]() { ASSERT_EQUALS(pcq.pop(), MoveOnly(1)); }).join();
+
+    ASSERT_TRUE(pcq.emptyForTest());
+}
+
+TEST_F(ProducerConsumerQueueTest, closeConsumerEnd) {
+    ProducerConsumerQueue<MoveOnly> pcq{1};
+
+    pcq.push(MoveOnly(1));
+
+    auto producer = runThread("Producer", [&]() {
+        ASSERT_THROWS_CODE(
+            pcq.push(MoveOnly(2)), DBException, ErrorCodes::ProducerConsumerQueueEndClosed);
+    });
+
+    ASSERT_EQUALS(pcq.sizeForTest(), 1ul);
+
+    pcq.closeConsumerEnd();
+
+    ASSERT_THROWS_CODE(pcq.pop(), DBException, ErrorCodes::ProducerConsumerQueueEndClosed);
+
+    producer.join();
+}
+
+TEST_F(ProducerConsumerQueueTest, closeProducerEndImmediate) {
+    ProducerConsumerQueue<MoveOnly> pcq{};
+
+    pcq.push(MoveOnly(1));
+    pcq.closeProducerEnd();
+
+    runThread("Consumer", [&]() {
+        ASSERT_EQUALS(pcq.pop(), MoveOnly(1));
+
+        ASSERT_THROWS_CODE(pcq.pop(), DBException, ErrorCodes::ProducerConsumerQueueEndClosed);
+    }).join();
+}
+
+TEST_F(ProducerConsumerQueueTest, closeProducerEndBlocking) {
+    ProducerConsumerQueue<MoveOnly> pcq{};
+
+    auto consumer = runThread("Consumer", [&]() {
+        ASSERT_THROWS_CODE(pcq.pop(), DBException, ErrorCodes::ProducerConsumerQueueEndClosed);
+    });
+
+    pcq.closeProducerEnd();
+
+    consumer.join();
+}
+
+TEST_F(ProducerConsumerQueueTest, popsWithTimeout) {
+    ProducerConsumerQueue<MoveOnly> pcq{};
+
+    runThread("Consumer", [&]() {
+        ASSERT_THROWS_CODE(pcq.pop(Milliseconds(100)), DBException, ErrorCodes::ExceededTimeLimit);
+
+        std::vector<MoveOnly> vec;
+        ASSERT_THROWS_CODE(pcq.popMany(std::back_inserter(vec), Milliseconds(100)),
+                           DBException,
+                           ErrorCodes::ExceededTimeLimit);
+
+        ASSERT_THROWS_CODE(pcq.popManyUpTo(1000, std::back_inserter(vec), Milliseconds(100)),
+                           DBException,
+                           ErrorCodes::ExceededTimeLimit);
+    }).join();
+
+    ASSERT_EQUALS(pcq.sizeForTest(), 0ul);
+}
+
+TEST_F(ProducerConsumerQueueTest, pushesWithTimeout) {
+    ProducerConsumerQueue<MoveOnly> pcq{1};
+
+    {
+        MoveOnly mo(1);
+        pcq.push(std::move(mo));
+        ASSERT(mo.movedFrom());
+    }
+
+    runThread("Consumer", [&]() {
+        {
+            MoveOnly mo(2);
+            ASSERT_THROWS_CODE(pcq.push(std::move(mo), Milliseconds(100)),
+                               DBException,
+                               ErrorCodes::ExceededTimeLimit);
+            ASSERT_EQUALS(pcq.sizeForTest(), 1ul);
+            ASSERT(!mo.movedFrom());
+            ASSERT_EQUALS(mo, MoveOnly(2));
+        }
+
+        {
+            std::vector<MoveOnly> vec;
+            vec.emplace_back(MoveOnly(2));
+
+            auto iter = begin(vec);
+            ASSERT_THROWS_CODE(pcq.pushMany(iter, end(vec), Milliseconds(100)),
+                               DBException,
+                               ErrorCodes::ExceededTimeLimit);
+            ASSERT_EQUALS(pcq.sizeForTest(), 1ul);
+            ASSERT(!vec[0].movedFrom());
+            ASSERT_EQUALS(vec[0], MoveOnly(2));
+        }
+    }).join();
+
+    ASSERT_EQUALS(pcq.sizeForTest(), 1ul);
+}
+
+TEST_F(ProducerConsumerQueueTest, basicPushPopWithBlocking) {
+    ProducerConsumerQueue<MoveOnly> pcq{};
+
+    auto consumer = runThread("Consumer", [&]() { ASSERT_EQUALS(pcq.pop(), MoveOnly(1)); });
+
+    auto producer = runThread("Producer", [&]() { pcq.push(MoveOnly(1)); });
+
+    consumer.join();
+    producer.join();
+
+    ASSERT_TRUE(pcq.emptyForTest());
+}
+
+TEST_F(ProducerConsumerQueueTest, multipleStepPushPopWithBlocking) {
+    ProducerConsumerQueue<MoveOnly> pcq{1};
+
+    auto consumer = runThread("Consumer", [&]() {
+        for (int i = 0; i < 10; ++i) {
+            ASSERT_EQUALS(pcq.pop(), MoveOnly(i));
+        }
+    });
+
+    auto producer = runThread("Producer", [&]() {
+        for (int i = 0; i < 10; ++i) {
+            pcq.push(MoveOnly(i));
+        }
+    });
+
+    consumer.join();
+    producer.join();
+
+    ASSERT_TRUE(pcq.emptyForTest());
+}
+
+
+TEST_F(ProducerConsumerQueueTest, pushTooLarge) {
+    {
+        ProducerConsumerQueue<MoveOnly, MoveOnly::CostFunc> pcq{1};
+
+        runThread("Producer", [&]() {
+            ASSERT_THROWS_CODE(
+                pcq.push(MoveOnly(2)), DBException, ErrorCodes::ProducerConsumerQueueBatchTooLarge);
+        }).join();
+    }
+
+    {
+        ProducerConsumerQueue<MoveOnly, MoveOnly::CostFunc> pcq{4};
+
+        std::vector<MoveOnly> vec;
+        vec.push_back(MoveOnly(3));
+        vec.push_back(MoveOnly(3));
+
+        runThread("Producer", [&]() {
+            ASSERT_THROWS_CODE(pcq.pushMany(begin(vec), end(vec)),
+                               DBException,
+                               ErrorCodes::ProducerConsumerQueueBatchTooLarge);
+        }).join();
+    }
+}
+
+TEST_F(ProducerConsumerQueueTest, pushManyPopWithoutBlocking) {
+    ProducerConsumerQueue<MoveOnly> pcq{};
+
+    runThread("Producer", [&]() {
+        std::vector<MoveOnly> vec;
+        for (int i = 0; i < 10; ++i) {
+            vec.emplace_back(MoveOnly(i));
+        }
+
+        pcq.pushMany(begin(vec), end(vec));
+    }).join();
+
+    runThread("Consumer", [&]() {
+        for (int i = 0; i < 10; ++i) {
+            ASSERT_EQUALS(pcq.pop(), MoveOnly(i));
+        }
+    }).join();
+
+    ASSERT_TRUE(pcq.emptyForTest());
+}
+
+TEST_F(ProducerConsumerQueueTest, popManyPopWithBlocking) {
+    ProducerConsumerQueue<MoveOnly> pcq{2};
+
+    auto consumer = runThread("Consumer", [&]() {
+        for (int i = 0; i < 10; i = i + 2) {
+            std::vector<MoveOnly> out;
+
+            pcq.popMany(std::back_inserter(out));
+
+            ASSERT_EQUALS(out.size(), 2ul);
+            ASSERT_EQUALS(out[0], MoveOnly(i));
+            ASSERT_EQUALS(out[1], MoveOnly(i + 1));
+        }
+    });
+
+    auto producer = runThread("Producer", [&]() {
+        std::vector<MoveOnly> vec;
+        for (int i = 0; i < 10; ++i) {
+            vec.emplace_back(MoveOnly(i));
+        }
+
+        for (auto iter = begin(vec); iter != end(vec); iter += 2) {
+            pcq.pushMany(iter, iter + 2);
+        }
+    });
+
+    consumer.join();
+    producer.join();
+
+    ASSERT_TRUE(pcq.emptyForTest());
+}
+
+TEST_F(ProducerConsumerQueueTest, popManyUpToPopWithBlocking) {
+    ProducerConsumerQueue<MoveOnly> pcq{4};
+
+    auto consumer = runThread("Consumer", [&]() {
+        for (int i = 0; i < 10; i = i + 2) {
+            std::vector<MoveOnly> out;
+
+            size_t spent;
+            std::tie(spent, std::ignore) = pcq.popManyUpTo(2, std::back_inserter(out));
+
+            ASSERT_EQUALS(spent, 2ul);
+            ASSERT_EQUALS(out.size(), 2ul);
+            ASSERT_EQUALS(out[0], MoveOnly(i));
+            ASSERT_EQUALS(out[1], MoveOnly(i + 1));
+        }
+    });
+
+    auto producer = runThread("Producer", [&]() {
+        std::vector<MoveOnly> vec;
+        for (int i = 0; i < 10; ++i) {
+            vec.emplace_back(MoveOnly(i));
+        }
+
+        for (auto iter = begin(vec); iter != end(vec); iter += 2) {
+            pcq.pushMany(iter, iter + 2);
+        }
+    });
+
+    consumer.join();
+    producer.join();
+
+    ASSERT_TRUE(pcq.emptyForTest());
+}
+
+TEST_F(ProducerConsumerQueueTest, popManyUpToPopWithBlockingWithSpecialCost) {
+    ProducerConsumerQueue<MoveOnly, MoveOnly::CostFunc> pcq{};
+
+    auto consumer = runThread("Consumer", [&]() {
+        {
+            std::vector<MoveOnly> out;
+            size_t spent;
+            std::tie(spent, std::ignore) = pcq.popManyUpTo(5, std::back_inserter(out));
+
+            ASSERT_EQUALS(spent, 6ul);
+            ASSERT_EQUALS(out.size(), 3ul);
+            ASSERT_EQUALS(out[0], MoveOnly(1));
+            ASSERT_EQUALS(out[1], MoveOnly(2));
+            ASSERT_EQUALS(out[2], MoveOnly(3));
+        }
+
+        {
+            std::vector<MoveOnly> out;
+            size_t spent;
+            std::tie(spent, std::ignore) = pcq.popManyUpTo(15, std::back_inserter(out));
+
+            ASSERT_EQUALS(spent, 9ul);
+            ASSERT_EQUALS(out.size(), 2ul);
+            ASSERT_EQUALS(out[0], MoveOnly(4));
+            ASSERT_EQUALS(out[1], MoveOnly(5));
+        }
+    });
+
+    auto producer = runThread("Producer", [&]() {
+        std::vector<MoveOnly> vec;
+        for (int i = 1; i < 6; ++i) {
+            vec.emplace_back(MoveOnly(i));
+        }
+
+        pcq.pushMany(begin(vec), end(vec));
+    });
+
+    consumer.join();
+    producer.join();
+
+    ASSERT_TRUE(pcq.emptyForTest());
+}
+
+TEST_F(ProducerConsumerQueueTest, singleProducerMultiConsumer) {
+    ProducerConsumerQueue<MoveOnly> pcq{};
+
+    stdx::mutex mutex;
+    size_t success = 0;
+    size_t failure = 0;
+
+    std::array<stdx::thread, 3> threads;
+    for (auto& thread : threads) {
+        thread = runThread("Consumer", [&]() {
+            {
+                try {
+                    pcq.pop();
+                    stdx::lock_guard<stdx::mutex> lk(mutex);
+                    success++;
+                } catch (const DBException& exception) {
+                    ASSERT_EQUALS(exception.getCode(), ErrorCodes::ProducerConsumerQueueEndClosed);
+                    stdx::lock_guard<stdx::mutex> lk(mutex);
+                    failure++;
+                }
+            }
+        });
+    }
+
+    pcq.push(MoveOnly(1));
+    pcq.push(MoveOnly(2));
+
+    pcq.closeProducerEnd();
+
+    for (auto& thread : threads) {
+        thread.join();
+    }
+
+    ASSERT_EQUALS(success, 2ul);
+    ASSERT_EQUALS(failure, 1ul);
+
+    ASSERT_TRUE(pcq.emptyForTest());
+}
+
+TEST_F(ProducerConsumerQueueTest, basicTryPop) {
+    ProducerConsumerQueue<MoveOnly> pcq{};
+
+    ASSERT_FALSE(pcq.tryPop());
+    ASSERT_TRUE(pcq.tryPush(MoveOnly(1)));
+    ASSERT_EQUALS(pcq.sizeForTest(), 1ul);
+
+    auto val = pcq.tryPop();
+
+    ASSERT_FALSE(pcq.tryPop());
+    ASSERT_TRUE(val);
+    ASSERT_EQUALS(*val, MoveOnly(1));
+
+    ASSERT_TRUE(pcq.emptyForTest());
+}
+
+TEST_F(ProducerConsumerQueueTest, basicTryPush) {
+    ProducerConsumerQueue<MoveOnly> pcq{1};
+
+    ASSERT_TRUE(pcq.tryPush(MoveOnly(1)));
+    ASSERT_FALSE(pcq.tryPush(MoveOnly(2)));
+
+    ASSERT_EQUALS(pcq.sizeForTest(), 1ul);
+
+    auto val = pcq.tryPop();
+    ASSERT_FALSE(pcq.tryPop());
+    ASSERT_TRUE(val);
+    ASSERT_EQUALS(*val, MoveOnly(1));
+
+    ASSERT_TRUE(pcq.emptyForTest());
+}
+
+TEST_F(ProducerConsumerQueueTest, tryPushWithSpecialCost) {
+    ProducerConsumerQueue<MoveOnly, MoveOnly::CostFunc> pcq{5};
+
+    ASSERT_TRUE(pcq.tryPush(MoveOnly(1)));
+    ASSERT_TRUE(pcq.tryPush(MoveOnly(2)));
+    ASSERT_FALSE(pcq.tryPush(MoveOnly(3)));
+
+    ASSERT_EQUALS(pcq.sizeForTest(), 3ul);
+
+    auto val1 = pcq.tryPop();
+    ASSERT_EQUALS(pcq.sizeForTest(), 2ul);
+    auto val2 = pcq.tryPop();
+    ASSERT_EQUALS(pcq.sizeForTest(), 0ul);
+    ASSERT_FALSE(pcq.tryPop());
+    ASSERT_TRUE(val1);
+    ASSERT_TRUE(val2);
+    ASSERT_EQUALS(*val1, MoveOnly(1));
+    ASSERT_EQUALS(*val2, MoveOnly(2));
+
+    ASSERT_TRUE(pcq.emptyForTest());
+}
+
+TEST_F(ProducerConsumerQueueTest, tryPushWithSpecialStatefulCost) {
+    ProducerConsumerQueue<MoveOnly, MoveOnly::CostFunc> pcq{5, MoveOnly::CostFunc(1)};
+
+    ASSERT_TRUE(pcq.tryPush(MoveOnly(1)));
+    ASSERT_TRUE(pcq.tryPush(MoveOnly(2)));
+    ASSERT_FALSE(pcq.tryPush(MoveOnly(3)));
+
+    ASSERT_EQUALS(pcq.sizeForTest(), 5ul);
+
+    auto val1 = pcq.tryPop();
+    ASSERT_EQUALS(pcq.sizeForTest(), 3ul);
+    auto val2 = pcq.tryPop();
+    ASSERT_EQUALS(pcq.sizeForTest(), 0ul);
+    ASSERT_FALSE(pcq.tryPop());
+    ASSERT_TRUE(val1);
+    ASSERT_TRUE(val2);
+    ASSERT_EQUALS(*val1, MoveOnly(1));
+    ASSERT_EQUALS(*val2, MoveOnly(2));
+
+    ASSERT_TRUE(pcq.emptyForTest());
+}
+
+}  // namespace
+
+}  // namespace mongo
diff --git a/src/mongo/util/tcmalloc_server_status_section.cpp b/src/mongo/util/tcmalloc_server_status_section.cpp
index 303076ba784..784f74e8cf8 100644
--- a/src/mongo/util/tcmalloc_server_status_section.cpp
+++ b/src/mongo/util/tcmalloc_server_status_section.cpp
@@ -38,6 +38,7 @@
 
 #include "mongo/base/init.h"
 #include "mongo/db/commands/server_status.h"
+#include "mongo/db/server_parameters.h"
 #include "mongo/db/service_context.h"
 #include "mongo/transport/transport_layer.h"
 #include "mongo/util/log.h"
@@ -55,11 +56,17 @@ const int kManyClients = 40;
 
 stdx::mutex tcmallocCleanupLock;
 
+MONGO_EXPORT_SERVER_PARAMETER(tcmallocEnableMarkThreadIdle, bool, true);
+
 /**
  *  Callback to allow TCMalloc to release freed memory to the central list at
  *  favorable times. Ideally would do some milder cleanup or scavenge...
  */
 void threadStateChange() {
+    if (!tcmallocEnableMarkThreadIdle.load()) {
+        return;
+    }
+
     if (getGlobalServiceContext()->getTransportLayer()->sessionStats().numOpenSessions <=
         kManyClients)
         return;
@@ -156,6 +163,25 @@ public:
             appendNumericPropertyIfAvailable(
                 sub, "aggressive_memory_decommit", "tcmalloc.aggressive_memory_decommit");
 
+            appendNumericPropertyIfAvailable(
+                sub, "pageheap_committed_bytes", "tcmalloc.pageheap_committed_bytes");
+            appendNumericPropertyIfAvailable(
+                sub, "pageheap_scavenge_count", "tcmalloc.pageheap_scavenge_count");
+            appendNumericPropertyIfAvailable(
+                sub, "pageheap_commit_count", "tcmalloc.pageheap_commit_count");
+            appendNumericPropertyIfAvailable(
+                sub, "pageheap_total_commit_bytes", "tcmalloc.pageheap_total_commit_bytes");
+            appendNumericPropertyIfAvailable(
+                sub, "pageheap_decommit_count", "tcmalloc.pageheap_decommit_count");
+            appendNumericPropertyIfAvailable(
+                sub, "pageheap_total_decommit_bytes", "tcmalloc.pageheap_total_decommit_bytes");
+            appendNumericPropertyIfAvailable(
+                sub, "pageheap_reserve_count", "tcmalloc.pageheap_reserve_count");
+            appendNumericPropertyIfAvailable(
+                sub, "pageheap_total_reserve_bytes", "tcmalloc.pageheap_total_reserve_bytes");
+            appendNumericPropertyIfAvailable(
+                sub, "spinlock_total_delay_ns", "tcmalloc.spinlock_total_delay_ns");
+
 #if MONGO_HAVE_GPERFTOOLS_SIZE_CLASS_STATS
             if (verbosity >= 2) {
                 // Size class information
diff --git a/src/third_party/SConscript b/src/third_party/SConscript
index 3a820399677..add20816b83 100644
--- a/src/third_party/SConscript
+++ b/src/third_party/SConscript
@@ -6,7 +6,7 @@ Import("wiredtiger")
 boostSuffix = "-1.60.0"
 snappySuffix = '-1.1.3'
 zlibSuffix = '-1.2.8'
-pcreSuffix = "-8.39"
+pcreSuffix = "-8.41"
 mozjsSuffix = '-45'
 yamlSuffix = '-0.5.3'
 icuSuffix = '-57.1'
diff --git a/src/third_party/gperftools-2.5/src/base/spinlock.cc b/src/third_party/gperftools-2.5/src/base/spinlock.cc
index 85ff21ed404..acb01c8537a 100644
--- a/src/third_party/gperftools-2.5/src/base/spinlock.cc
+++ b/src/third_party/gperftools-2.5/src/base/spinlock.cc
@@ -49,6 +49,47 @@ const base::LinkerInitialized SpinLock::LINKER_INITIALIZED =
     base::LINKER_INITIALIZED;
 
 namespace {
+
+const int64_t kNanoPerSec = uint64_t(1000) * 1000 * 1000;
+
+#ifdef _WIN32
+int64_t frequency;
+
+void InitTimer() {
+    LARGE_INTEGER large_freq;
+    QueryPerformanceFrequency(&large_freq);
+    frequency = large_freq.QuadPart;
+}
+
+int64_t NowMonotonic() {
+  LARGE_INTEGER time_value;
+  QueryPerformanceCounter(&time_value);
+  return time_value.QuadPart;
+}
+
+int64_t TicksToNanos(int64_t timer_value) {
+  return timer_value * kNanoPerSec / frequency;
+}
+
+#else // _WIN32
+void InitTimer() {}
+
+int64_t NowMonotonic() {
+  struct timespec t;
+
+  if (clock_gettime(CLOCK_MONOTONIC, &t)) {
+    return 0;
+  }
+
+  return (static_cast<double>(t.tv_sec) * kNanoPerSec) + t.tv_nsec;
+}
+
+int64_t TicksToNanos(int64_t timer_value) {
+  return timer_value;
+}
+
+#endif // _WIN32
+
 struct SpinLock_InitHelper {
   SpinLock_InitHelper() {
     // On multi-cpu machines, spin for longer before yielding
@@ -56,6 +97,8 @@ struct SpinLock_InitHelper {
     if (GetSystemCPUsCount() > 1) {
       adaptive_spin_count = 1000;
     }
+
+    InitTimer();
   }
 };
 
@@ -69,6 +112,31 @@ inline void SpinlockPause(void) {
   __asm__ __volatile__("rep; nop" : : );
 #endif
 }
+  
+// The current version of atomic_ops.h lacks base::subtle::Barrier_AtomicIncrement
+// so a CAS loop is used instead.
+void NoBarrier_AtomicAdd(volatile base::subtle::Atomic64* ptr,
+                         base::subtle::Atomic64 increment) {
+  base::subtle::Atomic64 base_value = base::subtle::NoBarrier_Load(ptr);
+  while (true) {
+    base::subtle::Atomic64 new_value;
+    base::subtle::NoBarrier_Store(&new_value, base::subtle::NoBarrier_Load(&base_value)
+          + base::subtle::NoBarrier_Load(&increment));
+
+    // Swap in the new incremented value.
+    base::subtle::Atomic64 cas_result = base::subtle::Acquire_CompareAndSwap(
+          ptr, base_value, new_value);
+
+    // Check if the increment succeeded.
+    if (cas_result == base_value) {
+      return;
+    }
+
+    // If the increment failed, just use the previous value as the value to
+    // add our increment to.
+    base_value = cas_result;
+  };
+}
 
 }  // unnamed namespace
 
@@ -84,10 +152,13 @@ Atomic32 SpinLock::SpinLoop() {
                                               kSpinLockSleeper);
 }
 
+base::subtle::Atomic64 SpinLock::totalDelayNanos_ = 0;
+
 void SpinLock::SlowLock() {
   Atomic32 lock_value = SpinLoop();
 
   int lock_wait_call_count = 0;
+  int64_t start = 0;
   while (lock_value != kSpinLockFree) {
     // If the lock is currently held, but not marked as having a sleeper, mark
     // it as having a sleeper.
@@ -115,12 +186,18 @@ void SpinLock::SlowLock() {
     }
 
     // Wait for an OS specific delay.
+    start = NowMonotonic();
     base::internal::SpinLockDelay(&lockword_, lock_value,
                                   ++lock_wait_call_count);
     // Spin again after returning from the wait routine to give this thread
     // some chance of obtaining the lock.
     lock_value = SpinLoop();
   }
+
+  if (start) {
+    NoBarrier_AtomicAdd(&totalDelayNanos_, static_cast<base::subtle::Atomic64>(
+        TicksToNanos(NowMonotonic() - start)));
+  }
 }
 
 void SpinLock::SlowUnlock() {
diff --git a/src/third_party/gperftools-2.5/src/base/spinlock.h b/src/third_party/gperftools-2.5/src/base/spinlock.h
index 7243aeaaefd..c5ad8f048f6 100644
--- a/src/third_party/gperftools-2.5/src/base/spinlock.h
+++ b/src/third_party/gperftools-2.5/src/base/spinlock.h
@@ -107,6 +107,10 @@ class LOCKABLE SpinLock {
     return base::subtle::NoBarrier_Load(&lockword_) != kSpinLockFree;
   }
 
+  static base::subtle::Atomic64 GetTotalDelayNanos() {
+    return base::subtle::NoBarrier_Load(&totalDelayNanos_);
+  }
+
   static const base::LinkerInitialized LINKER_INITIALIZED;  // backwards compat
  private:
   enum { kSpinLockFree = 0 };
@@ -115,6 +119,8 @@ class LOCKABLE SpinLock {
 
   volatile Atomic32 lockword_;
 
+  static base::subtle::Atomic64 totalDelayNanos_;
+
   void SlowLock();
   void SlowUnlock();
   Atomic32 SpinLoop();
diff --git a/src/third_party/gperftools-2.5/src/page_heap.cc b/src/third_party/gperftools-2.5/src/page_heap.cc
index f52ae2af029..eaa33255f2c 100644
--- a/src/third_party/gperftools-2.5/src/page_heap.cc
+++ b/src/third_party/gperftools-2.5/src/page_heap.cc
@@ -245,16 +245,22 @@ Span* PageHeap::Split(Span* span, Length n) {
 }
 
 void PageHeap::CommitSpan(Span* span) {
+  ++stats_.commit_count;
+
   TCMalloc_SystemCommit(reinterpret_cast<void*>(span->start << kPageShift),
                         static_cast<size_t>(span->length << kPageShift));
   stats_.committed_bytes += span->length << kPageShift;
+  stats_.total_commit_bytes += (span->length << kPageShift);
 }
 
 bool PageHeap::DecommitSpan(Span* span) {
+  ++stats_.decommit_count;
+
   bool rv = TCMalloc_SystemRelease(reinterpret_cast<void*>(span->start << kPageShift),
                                    static_cast<size_t>(span->length << kPageShift));
   if (rv) {
     stats_.committed_bytes -= span->length << kPageShift;
+    stats_.total_decommit_bytes += (span->length << kPageShift);
   }
 
   return rv;
@@ -441,6 +447,8 @@ void PageHeap::IncrementalScavenge(Length n) {
     return;
   }
 
+  ++stats_.scavenge_count;
+
   Length released_pages = ReleaseAtLeastNPages(1);
 
   if (released_pages == 0) {
@@ -616,9 +624,16 @@ bool PageHeap::GrowHeap(Length n) {
   ask = actual_size >> kPageShift;
   RecordGrowth(ask << kPageShift);
 
+  ++stats_.reserve_count;
+  ++stats_.commit_count;
+
   uint64_t old_system_bytes = stats_.system_bytes;
   stats_.system_bytes += (ask << kPageShift);
   stats_.committed_bytes += (ask << kPageShift);
+
+  stats_.total_commit_bytes += (ask << kPageShift);
+  stats_.total_reserve_bytes += (ask << kPageShift);
+
   const PageID p = reinterpret_cast<uintptr_t>(ptr) >> kPageShift;
   ASSERT(p > 0);
 
diff --git a/src/third_party/gperftools-2.5/src/page_heap.h b/src/third_party/gperftools-2.5/src/page_heap.h
index 18abed1974a..5c6607bbf16 100644
--- a/src/third_party/gperftools-2.5/src/page_heap.h
+++ b/src/third_party/gperftools-2.5/src/page_heap.h
@@ -143,12 +143,24 @@ class PERFTOOLS_DLL_DECL PageHeap {
 
   // Page heap statistics
   struct Stats {
-    Stats() : system_bytes(0), free_bytes(0), unmapped_bytes(0), committed_bytes(0) {}
+    Stats() : system_bytes(0), free_bytes(0), unmapped_bytes(0), committed_bytes(0),
+        scavenge_count(0), commit_count(0), decommit_count(0),
+        total_commit_bytes(0), total_decommit_bytes(0),
+        reserve_count(0), total_reserve_bytes(0) {}
     uint64_t system_bytes;    // Total bytes allocated from system
     uint64_t free_bytes;      // Total bytes on normal freelists
     uint64_t unmapped_bytes;  // Total bytes on returned freelists
     uint64_t committed_bytes;  // Bytes committed, always <= system_bytes_.
 
+    uint64_t scavenge_count;   // Number of times scavagened flush pages
+
+    uint64_t commit_count;          // Number of virtual memory commits
+    uint64_t total_commit_bytes;    // Bytes committed in lifetime of process
+    uint64_t decommit_count;        // Number of virtual memory decommits
+    uint64_t total_decommit_bytes;  // Bytes decommitted in lifetime of process
+
+    uint64_t reserve_count;         // Number of virtual memory reserves
+    uint64_t total_reserve_bytes;   // Bytes reserved in lifetime of process
   };
   inline Stats stats() const { return stats_; }
 
diff --git a/src/third_party/gperftools-2.5/src/tcmalloc.cc b/src/third_party/gperftools-2.5/src/tcmalloc.cc
index ba6c06e9d89..b573b4a9821 100644
--- a/src/third_party/gperftools-2.5/src/tcmalloc.cc
+++ b/src/third_party/gperftools-2.5/src/tcmalloc.cc
@@ -703,6 +703,54 @@ class TCMallocImplementation : public MallocExtension {
       return true;
     }
 
+    if (strcmp(name, "tcmalloc.pageheap_committed_bytes") == 0) {
+      SpinLockHolder l(Static::pageheap_lock());
+      *value = Static::pageheap()->stats().committed_bytes;
+      return true;
+    }
+
+    if (strcmp(name, "tcmalloc.pageheap_scavenge_count") == 0) {
+      SpinLockHolder l(Static::pageheap_lock());
+      *value = Static::pageheap()->stats().scavenge_count;
+      return true;
+    }
+
+    if (strcmp(name, "tcmalloc.pageheap_commit_count") == 0) {
+      SpinLockHolder l(Static::pageheap_lock());
+      *value = Static::pageheap()->stats().commit_count;
+      return true;
+    }
+
+    if (strcmp(name, "tcmalloc.pageheap_total_commit_bytes") == 0) {
+      SpinLockHolder l(Static::pageheap_lock());
+      *value = Static::pageheap()->stats().total_commit_bytes;
+      return true;
+    }
+
+    if (strcmp(name, "tcmalloc.pageheap_decommit_count") == 0) {
+      SpinLockHolder l(Static::pageheap_lock());
+      *value = Static::pageheap()->stats().decommit_count;
+      return true;
+    }
+
+    if (strcmp(name, "tcmalloc.pageheap_total_decommit_bytes") == 0) {
+      SpinLockHolder l(Static::pageheap_lock());
+      *value = Static::pageheap()->stats().total_decommit_bytes;
+      return true;
+    }
+
+    if (strcmp(name, "tcmalloc.pageheap_reserve_count") == 0) {
+      SpinLockHolder l(Static::pageheap_lock());
+      *value = Static::pageheap()->stats().reserve_count;
+      return true;
+    }
+
+    if (strcmp(name, "tcmalloc.pageheap_total_reserve_bytes") == 0) {
+        SpinLockHolder l(Static::pageheap_lock());
+        *value = Static::pageheap()->stats().total_reserve_bytes;
+        return true;
+    }
+
     if (strcmp(name, "tcmalloc.max_total_thread_cache_bytes") == 0) {
       SpinLockHolder l(Static::pageheap_lock());
       *value = ThreadCache::overall_thread_cache_size();
@@ -721,6 +769,11 @@ class TCMallocImplementation : public MallocExtension {
       return true;
     }
 
+    if (strcmp(name, "tcmalloc.spinlock_total_delay_ns") == 0) {
+      *value = SpinLock::GetTotalDelayNanos();
+      return true;
+    }
+
     return false;
   }
 
diff --git a/src/third_party/pcre-8.39/132html b/src/third_party/pcre-8.41/132html
index e00024970a5..e00024970a5 100755
--- a/src/third_party/pcre-8.39/132html
+++ b/src/third_party/pcre-8.41/132html
diff --git a/src/third_party/pcre-8.39/AUTHORS b/src/third_party/pcre-8.41/AUTHORS
index 342417a8a19..291657caef1 100644
--- a/src/third_party/pcre-8.39/AUTHORS
+++ b/src/third_party/pcre-8.41/AUTHORS
@@ -8,7 +8,7 @@ Email domain:     cam.ac.uk
 University of Cambridge Computing Service,
 Cambridge, England.
 
-Copyright (c) 1997-2016 University of Cambridge
+Copyright (c) 1997-2017 University of Cambridge
 All rights reserved
 
 
@@ -19,7 +19,7 @@ Written by:       Zoltan Herczeg
 Email local part: hzmester
 Emain domain:     freemail.hu
 
-Copyright(c) 2010-2016 Zoltan Herczeg
+Copyright(c) 2010-2017 Zoltan Herczeg
 All rights reserved.
 
 
@@ -30,7 +30,7 @@ Written by:       Zoltan Herczeg
 Email local part: hzmester
 Emain domain:     freemail.hu
 
-Copyright(c) 2009-2016 Zoltan Herczeg
+Copyright(c) 2009-2017 Zoltan Herczeg
 All rights reserved.
 
 
diff --git a/src/third_party/pcre-8.39/COPYING b/src/third_party/pcre-8.41/COPYING
index 58eed01b61d..58eed01b61d 100644
--- a/src/third_party/pcre-8.39/COPYING
+++ b/src/third_party/pcre-8.41/COPYING
diff --git a/src/third_party/pcre-8.39/ChangeLog b/src/third_party/pcre-8.41/ChangeLog
index a34f845f8a1..590a7542885 100644
--- a/src/third_party/pcre-8.39/ChangeLog
+++ b/src/third_party/pcre-8.41/ChangeLog
@@ -4,6 +4,100 @@ ChangeLog for PCRE
 Note that the PCRE 8.xx series (PCRE1) is now in a bugfix-only state. All
 development is happening in the PCRE2 10.xx series.
 
+Version 8.41 05-July-2017
+-------------------------
+
+1.  Fixed typo in CMakeLists.txt (wrong number of arguments for
+PCRE_STATIC_RUNTIME (affects MSVC only).
+
+2.  Issue 1 for 8.40 below was not correctly fixed. If pcregrep in multiline
+mode with --only-matching matched several lines, it restarted scanning at the
+next line instead of moving on to the end of the matched string, which can be
+several lines after the start.
+
+3.  Fix a missing else in the JIT compiler reported by 'idaifish'.
+
+4.  A (?# style comment is now ignored between a basic quantifier and a
+following '+' or '?' (example: /X+(?#comment)?Y/.
+
+5.  Avoid use of a potentially overflowing buffer in pcregrep (patch by Petr
+Pisar).
+
+6.  Fuzzers have reported issues in pcretest. These are NOT serious (it is,
+after all, just a test program). However, to stop the reports, some easy ones
+are fixed:
+
+    (a) Check for values < 256 when calling isprint() in pcretest.
+    (b) Give an error for too big a number after \O.
+
+7.  In the 32-bit library in non-UTF mode, an attempt to find a Unicode
+property for a character with a code point greater than 0x10ffff (the Unicode
+maximum) caused a crash.
+
+8. The alternative matching function, pcre_dfa_exec() misbehaved if it
+encountered a character class with a possessive repeat, for example [a-f]{3}+.
+
+9. When pcretest called pcre_copy_substring() in 32-bit mode, it set the buffer
+length incorrectly, which could result in buffer overflow.
+
+10. Remove redundant line of code (accidentally left in ages ago).
+
+11. Applied C++ patch from Irfan Adilovic to guard 'using std::' directives
+with namespace pcrecpp (Bugzilla #2084).
+
+12. Remove a duplication typo in pcre_tables.c.
+
+13. Fix returned offsets from regexec() when REG_STARTEND is used with a
+starting offset greater than zero.
+
+
+Version 8.40 11-January-2017
+----------------------------
+
+1.  Using -o with -M in pcregrep could cause unnecessary repeated output when
+    the match extended over a line boundary.
+
+2.  Applied Chris Wilson's second patch (Bugzilla #1681) to CMakeLists.txt for
+    MSVC static compilation, putting the first patch under a new option.
+
+3.  Fix register overwite in JIT when SSE2 acceleration is enabled.
+
+4.  Ignore "show all captures" (/=) for DFA matching.
+
+5.  Fix JIT unaligned accesses on x86. Patch by Marc Mutz.
+
+6.  In any wide-character mode (8-bit UTF or any 16-bit or 32-bit mode),
+    without PCRE_UCP set, a negative character type such as \D in a positive
+    class should cause all characters greater than 255 to match, whatever else
+    is in the class. There was a bug that caused this not to happen if a
+    Unicode property item was added to such a class, for example [\D\P{Nd}] or
+    [\W\pL].
+
+7.  When pcretest was outputing information from a callout, the caret indicator
+    for the current position in the subject line was incorrect if it was after
+    an escape sequence for a character whose code point was greater than
+    \x{ff}.
+
+8.  A pattern such as (?<RA>abc)(?(R)xyz) was incorrectly compiled such that
+    the conditional was interpreted as a reference to capturing group 1 instead
+    of a test for recursion. Any group whose name began with R was
+    misinterpreted in this way. (The reference interpretation should only
+    happen if the group's name is precisely "R".)
+
+9.  A number of bugs have been mended relating to match start-up optimizations
+    when the first thing in a pattern is a positive lookahead. These all
+    applied only when PCRE_NO_START_OPTIMIZE was *not* set:
+
+    (a) A pattern such as (?=.*X)X$ was incorrectly optimized as if it needed
+        both an initial 'X' and a following 'X'.
+    (b) Some patterns starting with an assertion that started with .* were
+        incorrectly optimized as having to match at the start of the subject or
+        after a newline. There are cases where this is not true, for example,
+        (?=.*[A-Z])(?=.{8,16})(?!.*[\s]) matches after the start in lines that
+        start with spaces. Starting .* in an assertion is no longer taken as an
+        indication of matching at the start (or after a newline).
+
+
 Version 8.39 14-June-2016
 -------------------------
 
diff --git a/src/third_party/pcre-8.39/CheckMan b/src/third_party/pcre-8.41/CheckMan
index 480d735481e..480d735481e 100755
--- a/src/third_party/pcre-8.39/CheckMan
+++ b/src/third_party/pcre-8.41/CheckMan
diff --git a/src/third_party/pcre-8.39/CleanTxt b/src/third_party/pcre-8.41/CleanTxt
index 1f42519c8d7..1f42519c8d7 100755
--- a/src/third_party/pcre-8.39/CleanTxt
+++ b/src/third_party/pcre-8.41/CleanTxt
diff --git a/src/third_party/pcre-8.39/Detrail b/src/third_party/pcre-8.41/Detrail
index 1c5c7e9cae1..1c5c7e9cae1 100755
--- a/src/third_party/pcre-8.39/Detrail
+++ b/src/third_party/pcre-8.41/Detrail
diff --git a/src/third_party/pcre-8.39/HACKING b/src/third_party/pcre-8.41/HACKING
index 691b7a14e50..691b7a14e50 100644
--- a/src/third_party/pcre-8.39/HACKING
+++ b/src/third_party/pcre-8.41/HACKING
diff --git a/src/third_party/pcre-8.39/INSTALL b/src/third_party/pcre-8.41/INSTALL
index 2099840756e..2099840756e 100644
--- a/src/third_party/pcre-8.39/INSTALL
+++ b/src/third_party/pcre-8.41/INSTALL
diff --git a/src/third_party/pcre-8.39/LICENCE b/src/third_party/pcre-8.41/LICENCE
index dd977af971b..dd9071a8dd8 100644
--- a/src/third_party/pcre-8.39/LICENCE
+++ b/src/third_party/pcre-8.41/LICENCE
@@ -25,7 +25,7 @@ Email domain:     cam.ac.uk
 University of Cambridge Computing Service,
 Cambridge, England.
 
-Copyright (c) 1997-2016 University of Cambridge
+Copyright (c) 1997-2017 University of Cambridge
 All rights reserved.
 
 
@@ -36,7 +36,7 @@ Written by:       Zoltan Herczeg
 Email local part: hzmester
 Emain domain:     freemail.hu
 
-Copyright(c) 2010-2016 Zoltan Herczeg
+Copyright(c) 2010-2017 Zoltan Herczeg
 All rights reserved.
 
 
@@ -47,7 +47,7 @@ Written by:       Zoltan Herczeg
 Email local part: hzmester
 Emain domain:     freemail.hu
 
-Copyright(c) 2009-2016 Zoltan Herczeg
+Copyright(c) 2009-2017 Zoltan Herczeg
 All rights reserved.
 
 
diff --git a/src/third_party/pcre-8.39/NEWS b/src/third_party/pcre-8.41/NEWS
index 0ca1bab2a4c..36be07cb880 100644
--- a/src/third_party/pcre-8.39/NEWS
+++ b/src/third_party/pcre-8.41/NEWS
@@ -1,6 +1,18 @@
 News about PCRE releases
 ------------------------
 
+Release 8.41 13-June-2017
+-------------------------
+
+This is a bug-fix release.
+
+
+Release 8.40 11-January-2017
+----------------------------
+
+This is a bug-fix release.
+
+
 Release 8.39 14-June-2016
 -------------------------
 
diff --git a/src/third_party/pcre-8.39/NON-AUTOTOOLS-BUILD b/src/third_party/pcre-8.41/NON-AUTOTOOLS-BUILD
index 3910059106b..3910059106b 100644
--- a/src/third_party/pcre-8.39/NON-AUTOTOOLS-BUILD
+++ b/src/third_party/pcre-8.41/NON-AUTOTOOLS-BUILD
diff --git a/src/third_party/pcre-8.39/NON-UNIX-USE b/src/third_party/pcre-8.41/NON-UNIX-USE
index a25546b6ff5..a25546b6ff5 100644
--- a/src/third_party/pcre-8.39/NON-UNIX-USE
+++ b/src/third_party/pcre-8.41/NON-UNIX-USE
diff --git a/src/third_party/pcre-8.39/PrepareRelease b/src/third_party/pcre-8.41/PrepareRelease
index 9891e08d67f..9891e08d67f 100755
--- a/src/third_party/pcre-8.39/PrepareRelease
+++ b/src/third_party/pcre-8.41/PrepareRelease
diff --git a/src/third_party/pcre-8.39/README b/src/third_party/pcre-8.41/README
index 4887ebf350e..4887ebf350e 100644
--- a/src/third_party/pcre-8.39/README
+++ b/src/third_party/pcre-8.41/README
diff --git a/src/third_party/pcre-8.39/RunGrepTest b/src/third_party/pcre-8.41/RunGrepTest
index 721ec5184ab..721ec5184ab 100755
--- a/src/third_party/pcre-8.39/RunGrepTest
+++ b/src/third_party/pcre-8.41/RunGrepTest
diff --git a/src/third_party/pcre-8.39/RunTest b/src/third_party/pcre-8.41/RunTest
index 357a739b75f..357a739b75f 100755
--- a/src/third_party/pcre-8.39/RunTest
+++ b/src/third_party/pcre-8.41/RunTest
diff --git a/src/third_party/pcre-8.39/RunTest.bat b/src/third_party/pcre-8.41/RunTest.bat
index 35d7f71f9e8..35d7f71f9e8 100644
--- a/src/third_party/pcre-8.39/RunTest.bat
+++ b/src/third_party/pcre-8.41/RunTest.bat
diff --git a/src/third_party/pcre-8.39/SConscript b/src/third_party/pcre-8.41/SConscript
index e7b9a59a550..e7b9a59a550 100644
--- a/src/third_party/pcre-8.39/SConscript
+++ b/src/third_party/pcre-8.41/SConscript
diff --git a/src/third_party/pcre-8.39/ar-lib b/src/third_party/pcre-8.41/ar-lib
index 463b9ec0206..463b9ec0206 100755
--- a/src/third_party/pcre-8.39/ar-lib
+++ b/src/third_party/pcre-8.41/ar-lib
diff --git a/src/third_party/pcre-8.39/build_posix/config.h b/src/third_party/pcre-8.41/build_posix/config.h
index 9cfa5a4d70f..8af8c7add1e 100644
--- a/src/third_party/pcre-8.39/build_posix/config.h
+++ b/src/third_party/pcre-8.41/build_posix/config.h
@@ -221,7 +221,7 @@ sure both macros are undefined; an emulation function will then be used. */
 #define PACKAGE_NAME "PCRE"
 
 /* Define to the full name and version of this package. */
-#define PACKAGE_STRING "PCRE 8.39"
+#define PACKAGE_STRING "PCRE 8.41"
 
 /* Define to the one symbol short name of this package. */
 #define PACKAGE_TARNAME "pcre"
@@ -230,7 +230,7 @@ sure both macros are undefined; an emulation function will then be used. */
 #define PACKAGE_URL ""
 
 /* Define to the version of this package. */
-#define PACKAGE_VERSION "8.39"
+#define PACKAGE_VERSION "8.41"
 
 /* The value of PARENS_NEST_LIMIT specifies the maximum depth of nested
    parentheses (of any kind) in a pattern. This limits the amount of system
@@ -335,7 +335,7 @@ sure both macros are undefined; an emulation function will then be used. */
 /* #undef SUPPORT_VALGRIND */
 
 /* Version number of package */
-#define VERSION "8.39"
+#define VERSION "8.41"
 
 /* Define to empty if `const' does not conform to ANSI C. */
 /* #undef const */
diff --git a/src/third_party/pcre-8.39/build_solaris/config.h b/src/third_party/pcre-8.41/build_solaris/config.h
index a391fd1cee5..164bd8a2e39 100644
--- a/src/third_party/pcre-8.39/build_solaris/config.h
+++ b/src/third_party/pcre-8.41/build_solaris/config.h
@@ -221,7 +221,7 @@ sure both macros are undefined; an emulation function will then be used. */
 #define PACKAGE_NAME "PCRE"
 
 /* Define to the full name and version of this package. */
-#define PACKAGE_STRING "PCRE 8.39"
+#define PACKAGE_STRING "PCRE 8.41"
 
 /* Define to the one symbol short name of this package. */
 #define PACKAGE_TARNAME "pcre"
@@ -230,7 +230,7 @@ sure both macros are undefined; an emulation function will then be used. */
 #define PACKAGE_URL ""
 
 /* Define to the version of this package. */
-#define PACKAGE_VERSION "8.39"
+#define PACKAGE_VERSION "8.41"
 
 /* The value of PARENS_NEST_LIMIT specifies the maximum depth of nested
    parentheses (of any kind) in a pattern. This limits the amount of system
@@ -335,7 +335,7 @@ sure both macros are undefined; an emulation function will then be used. */
 /* #undef SUPPORT_VALGRIND */
 
 /* Version number of package */
-#define VERSION "8.39"
+#define VERSION "8.41"
 
 /* Define to empty if `const' does not conform to ANSI C. */
 /* #undef const */
diff --git a/src/third_party/pcre-8.39/build_windows/config.h b/src/third_party/pcre-8.41/build_windows/config.h
index 2e8d4ec7b6a..2e8d4ec7b6a 100755
--- a/src/third_party/pcre-8.39/build_windows/config.h
+++ b/src/third_party/pcre-8.41/build_windows/config.h
diff --git a/src/third_party/pcre-8.39/dftables.c b/src/third_party/pcre-8.41/dftables.c
index 1fdc8e0f231..1fdc8e0f231 100644
--- a/src/third_party/pcre-8.39/dftables.c
+++ b/src/third_party/pcre-8.41/dftables.c
diff --git a/src/third_party/pcre-8.39/libpcre.pc.in b/src/third_party/pcre-8.41/libpcre.pc.in
index 0a35da87f85..0a35da87f85 100644
--- a/src/third_party/pcre-8.39/libpcre.pc.in
+++ b/src/third_party/pcre-8.41/libpcre.pc.in
diff --git a/src/third_party/pcre-8.39/libpcre16.pc.in b/src/third_party/pcre-8.41/libpcre16.pc.in
index 080c9dcfe8a..080c9dcfe8a 100644
--- a/src/third_party/pcre-8.39/libpcre16.pc.in
+++ b/src/third_party/pcre-8.41/libpcre16.pc.in
diff --git a/src/third_party/pcre-8.39/libpcre32.pc.in b/src/third_party/pcre-8.41/libpcre32.pc.in
index a3ae0e11fcf..a3ae0e11fcf 100644
--- a/src/third_party/pcre-8.39/libpcre32.pc.in
+++ b/src/third_party/pcre-8.41/libpcre32.pc.in
diff --git a/src/third_party/pcre-8.39/libpcrecpp.pc.in b/src/third_party/pcre-8.41/libpcrecpp.pc.in
index ef006fe47aa..ef006fe47aa 100644
--- a/src/third_party/pcre-8.39/libpcrecpp.pc.in
+++ b/src/third_party/pcre-8.41/libpcrecpp.pc.in
diff --git a/src/third_party/pcre-8.39/libpcreposix.pc.in b/src/third_party/pcre-8.41/libpcreposix.pc.in
index c6c0b0c6c66..c6c0b0c6c66 100644
--- a/src/third_party/pcre-8.39/libpcreposix.pc.in
+++ b/src/third_party/pcre-8.41/libpcreposix.pc.in
diff --git a/src/third_party/pcre-8.39/makevp.bat b/src/third_party/pcre-8.41/makevp.bat
index 5f795487eb9..5f795487eb9 100644
--- a/src/third_party/pcre-8.39/makevp.bat
+++ b/src/third_party/pcre-8.41/makevp.bat
diff --git a/src/third_party/pcre-8.39/makevp_c.txt b/src/third_party/pcre-8.41/makevp_c.txt
index 5648115405d..5648115405d 100644
--- a/src/third_party/pcre-8.39/makevp_c.txt
+++ b/src/third_party/pcre-8.41/makevp_c.txt
diff --git a/src/third_party/pcre-8.39/makevp_l.txt b/src/third_party/pcre-8.41/makevp_l.txt
index 9b071e0f90a..9b071e0f90a 100644
--- a/src/third_party/pcre-8.39/makevp_l.txt
+++ b/src/third_party/pcre-8.41/makevp_l.txt
diff --git a/src/third_party/pcre-8.39/pcre-config.in b/src/third_party/pcre-8.41/pcre-config.in
index ac06a3325bc..ac06a3325bc 100644
--- a/src/third_party/pcre-8.39/pcre-config.in
+++ b/src/third_party/pcre-8.41/pcre-config.in
diff --git a/src/third_party/pcre-8.39/pcre.h b/src/third_party/pcre-8.41/pcre.h
index 70559700654..442c6bdb5fe 100644
--- a/src/third_party/pcre-8.39/pcre.h
+++ b/src/third_party/pcre-8.41/pcre.h
@@ -42,9 +42,9 @@ POSSIBILITY OF SUCH DAMAGE.
 /* The current PCRE version information. */
 
 #define PCRE_MAJOR          8
-#define PCRE_MINOR          39
+#define PCRE_MINOR          41
 #define PCRE_PRERELEASE     
-#define PCRE_DATE           2016-06-14
+#define PCRE_DATE           2017-07-05
 
 /* When an application links to a PCRE DLL in Windows, the symbols that are
 imported have to be identified as such. When building PCRE, the appropriate
diff --git a/src/third_party/pcre-8.39/pcre.h.generic b/src/third_party/pcre-8.41/pcre.h.generic
index 70559700654..442c6bdb5fe 100644
--- a/src/third_party/pcre-8.39/pcre.h.generic
+++ b/src/third_party/pcre-8.41/pcre.h.generic
@@ -42,9 +42,9 @@ POSSIBILITY OF SUCH DAMAGE.
 /* The current PCRE version information. */
 
 #define PCRE_MAJOR          8
-#define PCRE_MINOR          39
+#define PCRE_MINOR          41
 #define PCRE_PRERELEASE     
-#define PCRE_DATE           2016-06-14
+#define PCRE_DATE           2017-07-05
 
 /* When an application links to a PCRE DLL in Windows, the symbols that are
 imported have to be identified as such. When building PCRE, the appropriate
diff --git a/src/third_party/pcre-8.39/pcre.h.in b/src/third_party/pcre-8.41/pcre.h.in
index 667a45ed575..667a45ed575 100644
--- a/src/third_party/pcre-8.39/pcre.h.in
+++ b/src/third_party/pcre-8.41/pcre.h.in
diff --git a/src/third_party/pcre-8.39/pcre16_byte_order.c b/src/third_party/pcre-8.41/pcre16_byte_order.c
index 11d2973a3db..11d2973a3db 100644
--- a/src/third_party/pcre-8.39/pcre16_byte_order.c
+++ b/src/third_party/pcre-8.41/pcre16_byte_order.c
diff --git a/src/third_party/pcre-8.39/pcre16_chartables.c b/src/third_party/pcre-8.41/pcre16_chartables.c
index 7c0ff35f5e0..7c0ff35f5e0 100644
--- a/src/third_party/pcre-8.39/pcre16_chartables.c
+++ b/src/third_party/pcre-8.41/pcre16_chartables.c
diff --git a/src/third_party/pcre-8.39/pcre16_compile.c b/src/third_party/pcre-8.41/pcre16_compile.c
index e499b670877..e499b670877 100644
--- a/src/third_party/pcre-8.39/pcre16_compile.c
+++ b/src/third_party/pcre-8.41/pcre16_compile.c
diff --git a/src/third_party/pcre-8.39/pcre16_config.c b/src/third_party/pcre-8.41/pcre16_config.c
index b52138764f6..b52138764f6 100644
--- a/src/third_party/pcre-8.39/pcre16_config.c
+++ b/src/third_party/pcre-8.41/pcre16_config.c
diff --git a/src/third_party/pcre-8.39/pcre16_dfa_exec.c b/src/third_party/pcre-8.41/pcre16_dfa_exec.c
index 2ba740e972b..2ba740e972b 100644
--- a/src/third_party/pcre-8.39/pcre16_dfa_exec.c
+++ b/src/third_party/pcre-8.41/pcre16_dfa_exec.c
diff --git a/src/third_party/pcre-8.39/pcre16_exec.c b/src/third_party/pcre-8.41/pcre16_exec.c
index 7417b1770c6..7417b1770c6 100644
--- a/src/third_party/pcre-8.39/pcre16_exec.c
+++ b/src/third_party/pcre-8.41/pcre16_exec.c
diff --git a/src/third_party/pcre-8.39/pcre16_fullinfo.c b/src/third_party/pcre-8.41/pcre16_fullinfo.c
index 544dca6ed5c..544dca6ed5c 100644
--- a/src/third_party/pcre-8.39/pcre16_fullinfo.c
+++ b/src/third_party/pcre-8.41/pcre16_fullinfo.c
diff --git a/src/third_party/pcre-8.39/pcre16_get.c b/src/third_party/pcre-8.41/pcre16_get.c
index 3ded08c622c..3ded08c622c 100644
--- a/src/third_party/pcre-8.39/pcre16_get.c
+++ b/src/third_party/pcre-8.41/pcre16_get.c
diff --git a/src/third_party/pcre-8.39/pcre16_globals.c b/src/third_party/pcre-8.41/pcre16_globals.c
index a136b3d8c22..a136b3d8c22 100644
--- a/src/third_party/pcre-8.39/pcre16_globals.c
+++ b/src/third_party/pcre-8.41/pcre16_globals.c
diff --git a/src/third_party/pcre-8.39/pcre16_jit_compile.c b/src/third_party/pcre-8.41/pcre16_jit_compile.c
index ab0cacd7646..ab0cacd7646 100644
--- a/src/third_party/pcre-8.39/pcre16_jit_compile.c
+++ b/src/third_party/pcre-8.41/pcre16_jit_compile.c
diff --git a/src/third_party/pcre-8.39/pcre16_maketables.c b/src/third_party/pcre-8.41/pcre16_maketables.c
index b1cd1c579d6..b1cd1c579d6 100644
--- a/src/third_party/pcre-8.39/pcre16_maketables.c
+++ b/src/third_party/pcre-8.41/pcre16_maketables.c
diff --git a/src/third_party/pcre-8.39/pcre16_newline.c b/src/third_party/pcre-8.41/pcre16_newline.c
index 7fe201400f5..7fe201400f5 100644
--- a/src/third_party/pcre-8.39/pcre16_newline.c
+++ b/src/third_party/pcre-8.41/pcre16_newline.c
diff --git a/src/third_party/pcre-8.39/pcre16_ord2utf16.c b/src/third_party/pcre-8.41/pcre16_ord2utf16.c
index 8e2ce5ea6c5..8e2ce5ea6c5 100644
--- a/src/third_party/pcre-8.39/pcre16_ord2utf16.c
+++ b/src/third_party/pcre-8.41/pcre16_ord2utf16.c
diff --git a/src/third_party/pcre-8.39/pcre16_printint.c b/src/third_party/pcre-8.41/pcre16_printint.c
index 33d8c340200..33d8c340200 100644
--- a/src/third_party/pcre-8.39/pcre16_printint.c
+++ b/src/third_party/pcre-8.41/pcre16_printint.c
diff --git a/src/third_party/pcre-8.39/pcre16_refcount.c b/src/third_party/pcre-8.41/pcre16_refcount.c
index d3d15439737..d3d15439737 100644
--- a/src/third_party/pcre-8.39/pcre16_refcount.c
+++ b/src/third_party/pcre-8.41/pcre16_refcount.c
diff --git a/src/third_party/pcre-8.39/pcre16_string_utils.c b/src/third_party/pcre-8.41/pcre16_string_utils.c
index 382c40799fb..382c40799fb 100644
--- a/src/third_party/pcre-8.39/pcre16_string_utils.c
+++ b/src/third_party/pcre-8.41/pcre16_string_utils.c
diff --git a/src/third_party/pcre-8.39/pcre16_study.c b/src/third_party/pcre-8.41/pcre16_study.c
index f87de081fc4..f87de081fc4 100644
--- a/src/third_party/pcre-8.39/pcre16_study.c
+++ b/src/third_party/pcre-8.41/pcre16_study.c
diff --git a/src/third_party/pcre-8.39/pcre16_tables.c b/src/third_party/pcre-8.41/pcre16_tables.c
index d84297093a4..d84297093a4 100644
--- a/src/third_party/pcre-8.39/pcre16_tables.c
+++ b/src/third_party/pcre-8.41/pcre16_tables.c
diff --git a/src/third_party/pcre-8.39/pcre16_ucd.c b/src/third_party/pcre-8.41/pcre16_ucd.c
index ee23439a013..ee23439a013 100644
--- a/src/third_party/pcre-8.39/pcre16_ucd.c
+++ b/src/third_party/pcre-8.41/pcre16_ucd.c
diff --git a/src/third_party/pcre-8.39/pcre16_utf16_utils.c b/src/third_party/pcre-8.41/pcre16_utf16_utils.c
index 49ced0c0b1c..49ced0c0b1c 100644
--- a/src/third_party/pcre-8.39/pcre16_utf16_utils.c
+++ b/src/third_party/pcre-8.41/pcre16_utf16_utils.c
diff --git a/src/third_party/pcre-8.39/pcre16_valid_utf16.c b/src/third_party/pcre-8.41/pcre16_valid_utf16.c
index 09076539d09..09076539d09 100644
--- a/src/third_party/pcre-8.39/pcre16_valid_utf16.c
+++ b/src/third_party/pcre-8.41/pcre16_valid_utf16.c
diff --git a/src/third_party/pcre-8.39/pcre16_version.c b/src/third_party/pcre-8.41/pcre16_version.c
index e991b1a8cfd..e991b1a8cfd 100644
--- a/src/third_party/pcre-8.39/pcre16_version.c
+++ b/src/third_party/pcre-8.41/pcre16_version.c
diff --git a/src/third_party/pcre-8.39/pcre16_xclass.c b/src/third_party/pcre-8.41/pcre16_xclass.c
index 5aac2a36c68..5aac2a36c68 100644
--- a/src/third_party/pcre-8.39/pcre16_xclass.c
+++ b/src/third_party/pcre-8.41/pcre16_xclass.c
diff --git a/src/third_party/pcre-8.39/pcre32_byte_order.c b/src/third_party/pcre-8.41/pcre32_byte_order.c
index 9cf5362730a..9cf5362730a 100644
--- a/src/third_party/pcre-8.39/pcre32_byte_order.c
+++ b/src/third_party/pcre-8.41/pcre32_byte_order.c
diff --git a/src/third_party/pcre-8.39/pcre32_chartables.c b/src/third_party/pcre-8.41/pcre32_chartables.c
index b5d8c23dbf1..b5d8c23dbf1 100644
--- a/src/third_party/pcre-8.39/pcre32_chartables.c
+++ b/src/third_party/pcre-8.41/pcre32_chartables.c
diff --git a/src/third_party/pcre-8.39/pcre32_compile.c b/src/third_party/pcre-8.41/pcre32_compile.c
index d781eb377e0..d781eb377e0 100644
--- a/src/third_party/pcre-8.39/pcre32_compile.c
+++ b/src/third_party/pcre-8.41/pcre32_compile.c
diff --git a/src/third_party/pcre-8.39/pcre32_config.c b/src/third_party/pcre-8.41/pcre32_config.c
index d63f3e9ea23..d63f3e9ea23 100644
--- a/src/third_party/pcre-8.39/pcre32_config.c
+++ b/src/third_party/pcre-8.41/pcre32_config.c
diff --git a/src/third_party/pcre-8.39/pcre32_dfa_exec.c b/src/third_party/pcre-8.41/pcre32_dfa_exec.c
index b0bfd34f04d..b0bfd34f04d 100644
--- a/src/third_party/pcre-8.39/pcre32_dfa_exec.c
+++ b/src/third_party/pcre-8.41/pcre32_dfa_exec.c
diff --git a/src/third_party/pcre-8.39/pcre32_exec.c b/src/third_party/pcre-8.41/pcre32_exec.c
index 8170ed77d35..8170ed77d35 100644
--- a/src/third_party/pcre-8.39/pcre32_exec.c
+++ b/src/third_party/pcre-8.41/pcre32_exec.c
diff --git a/src/third_party/pcre-8.39/pcre32_fullinfo.c b/src/third_party/pcre-8.41/pcre32_fullinfo.c
index 6ecc5209a08..6ecc5209a08 100644
--- a/src/third_party/pcre-8.39/pcre32_fullinfo.c
+++ b/src/third_party/pcre-8.41/pcre32_fullinfo.c
diff --git a/src/third_party/pcre-8.39/pcre32_get.c b/src/third_party/pcre-8.41/pcre32_get.c
index d35deee0cd4..d35deee0cd4 100644
--- a/src/third_party/pcre-8.39/pcre32_get.c
+++ b/src/third_party/pcre-8.41/pcre32_get.c
diff --git a/src/third_party/pcre-8.39/pcre32_globals.c b/src/third_party/pcre-8.41/pcre32_globals.c
index 32e0914ca6d..32e0914ca6d 100644
--- a/src/third_party/pcre-8.39/pcre32_globals.c
+++ b/src/third_party/pcre-8.41/pcre32_globals.c
diff --git a/src/third_party/pcre-8.39/pcre32_jit_compile.c b/src/third_party/pcre-8.41/pcre32_jit_compile.c
index 2e7c6f97c96..2e7c6f97c96 100644
--- a/src/third_party/pcre-8.39/pcre32_jit_compile.c
+++ b/src/third_party/pcre-8.41/pcre32_jit_compile.c
diff --git a/src/third_party/pcre-8.39/pcre32_maketables.c b/src/third_party/pcre-8.41/pcre32_maketables.c
index 5d1b1c64c96..5d1b1c64c96 100644
--- a/src/third_party/pcre-8.39/pcre32_maketables.c
+++ b/src/third_party/pcre-8.41/pcre32_maketables.c
diff --git a/src/third_party/pcre-8.39/pcre32_newline.c b/src/third_party/pcre-8.41/pcre32_newline.c
index 7f8d5360cdc..7f8d5360cdc 100644
--- a/src/third_party/pcre-8.39/pcre32_newline.c
+++ b/src/third_party/pcre-8.41/pcre32_newline.c
diff --git a/src/third_party/pcre-8.39/pcre32_ord2utf32.c b/src/third_party/pcre-8.41/pcre32_ord2utf32.c
index 606bcb3d7ba..606bcb3d7ba 100644
--- a/src/third_party/pcre-8.39/pcre32_ord2utf32.c
+++ b/src/third_party/pcre-8.41/pcre32_ord2utf32.c
diff --git a/src/third_party/pcre-8.39/pcre32_printint.c b/src/third_party/pcre-8.41/pcre32_printint.c
index f3fd7b25e2c..f3fd7b25e2c 100644
--- a/src/third_party/pcre-8.39/pcre32_printint.c
+++ b/src/third_party/pcre-8.41/pcre32_printint.c
diff --git a/src/third_party/pcre-8.39/pcre32_refcount.c b/src/third_party/pcre-8.41/pcre32_refcount.c
index dbdf432d82a..dbdf432d82a 100644
--- a/src/third_party/pcre-8.39/pcre32_refcount.c
+++ b/src/third_party/pcre-8.41/pcre32_refcount.c
diff --git a/src/third_party/pcre-8.39/pcre32_string_utils.c b/src/third_party/pcre-8.41/pcre32_string_utils.c
index e37b3d4805f..e37b3d4805f 100644
--- a/src/third_party/pcre-8.39/pcre32_string_utils.c
+++ b/src/third_party/pcre-8.41/pcre32_string_utils.c
diff --git a/src/third_party/pcre-8.39/pcre32_study.c b/src/third_party/pcre-8.41/pcre32_study.c
index d3a3afed791..d3a3afed791 100644
--- a/src/third_party/pcre-8.39/pcre32_study.c
+++ b/src/third_party/pcre-8.41/pcre32_study.c
diff --git a/src/third_party/pcre-8.39/pcre32_tables.c b/src/third_party/pcre-8.41/pcre32_tables.c
index 3d94cca33a1..3d94cca33a1 100644
--- a/src/third_party/pcre-8.39/pcre32_tables.c
+++ b/src/third_party/pcre-8.41/pcre32_tables.c
diff --git a/src/third_party/pcre-8.39/pcre32_ucd.c b/src/third_party/pcre-8.41/pcre32_ucd.c
index befe22d3435..befe22d3435 100644
--- a/src/third_party/pcre-8.39/pcre32_ucd.c
+++ b/src/third_party/pcre-8.41/pcre32_ucd.c
diff --git a/src/third_party/pcre-8.39/pcre32_utf32_utils.c b/src/third_party/pcre-8.41/pcre32_utf32_utils.c
index f844e237165..f844e237165 100644
--- a/src/third_party/pcre-8.39/pcre32_utf32_utils.c
+++ b/src/third_party/pcre-8.41/pcre32_utf32_utils.c
diff --git a/src/third_party/pcre-8.39/pcre32_valid_utf32.c b/src/third_party/pcre-8.41/pcre32_valid_utf32.c
index 94cda1a2c4c..94cda1a2c4c 100644
--- a/src/third_party/pcre-8.39/pcre32_valid_utf32.c
+++ b/src/third_party/pcre-8.41/pcre32_valid_utf32.c
diff --git a/src/third_party/pcre-8.39/pcre32_version.c b/src/third_party/pcre-8.41/pcre32_version.c
index fdaad9b0859..fdaad9b0859 100644
--- a/src/third_party/pcre-8.39/pcre32_version.c
+++ b/src/third_party/pcre-8.41/pcre32_version.c
diff --git a/src/third_party/pcre-8.39/pcre32_xclass.c b/src/third_party/pcre-8.41/pcre32_xclass.c
index 5662408ad5f..5662408ad5f 100644
--- a/src/third_party/pcre-8.39/pcre32_xclass.c
+++ b/src/third_party/pcre-8.41/pcre32_xclass.c
diff --git a/src/third_party/pcre-8.39/pcre_byte_order.c b/src/third_party/pcre-8.41/pcre_byte_order.c
index cf5f12b04ea..cf5f12b04ea 100644
--- a/src/third_party/pcre-8.39/pcre_byte_order.c
+++ b/src/third_party/pcre-8.41/pcre_byte_order.c
diff --git a/src/third_party/pcre-8.39/pcre_chartables.c b/src/third_party/pcre-8.41/pcre_chartables.c
index 1e20ec29d05..1e20ec29d05 100644
--- a/src/third_party/pcre-8.39/pcre_chartables.c
+++ b/src/third_party/pcre-8.41/pcre_chartables.c
diff --git a/src/third_party/pcre-8.39/pcre_chartables.c.dist b/src/third_party/pcre-8.41/pcre_chartables.c.dist
index 1e20ec29d05..1e20ec29d05 100644
--- a/src/third_party/pcre-8.39/pcre_chartables.c.dist
+++ b/src/third_party/pcre-8.41/pcre_chartables.c.dist
diff --git a/src/third_party/pcre-8.39/pcre_compile.c b/src/third_party/pcre-8.41/pcre_compile.c
index 7cd39501230..42f204cdfff 100644
--- a/src/third_party/pcre-8.39/pcre_compile.c
+++ b/src/third_party/pcre-8.41/pcre_compile.c
@@ -5579,6 +5579,34 @@ for (;; ptr++)
 #endif
 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
       {
+      /* For non-UCP wide characters, in a non-negative class containing \S or
+      similar (should_flip_negation is set), all characters greater than 255
+      must be in the class. */
+
+      if (
+#if defined COMPILE_PCRE8
+           utf &&
+#endif
+           should_flip_negation && !negate_class && (options & PCRE_UCP) == 0)
+        {
+        *class_uchardata++ = XCL_RANGE;
+        if (utf)   /* Will always be utf in the 8-bit library */
+          {
+          class_uchardata += PRIV(ord2utf)(0x100, class_uchardata);
+          class_uchardata += PRIV(ord2utf)(0x10ffff, class_uchardata);
+          }
+        else       /* Can only happen for the 16-bit & 32-bit libraries */
+          {
+#if defined COMPILE_PCRE16
+          *class_uchardata++ = 0x100;
+          *class_uchardata++ = 0xffffu;
+#elif defined COMPILE_PCRE32
+          *class_uchardata++ = 0x100;
+          *class_uchardata++ = 0xffffffffu;
+#endif
+          }
+        }
+
       *class_uchardata++ = XCL_END;    /* Marks the end of extra data */
       *code++ = OP_XCLASS;
       code += LINK_SIZE;
@@ -5711,6 +5739,21 @@ for (;; ptr++)
       ptr = p - 1;    /* Character before the next significant one. */
       }
 
+    /* We also need to skip over (?# comments, which are not dependent on
+    extended mode. */
+
+    if (ptr[1] == CHAR_LEFT_PARENTHESIS && ptr[2] == CHAR_QUESTION_MARK &&
+        ptr[3] == CHAR_NUMBER_SIGN)
+      {
+      ptr += 4;
+      while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
+      if (*ptr == CHAR_NULL)
+        {
+        *errorcodeptr = ERR18;
+        goto FAILED;
+        }
+      }
+
     /* If the next character is '+', we have a possessive quantifier. This
     implies greediness, whatever the setting of the PCRE_UNGREEDY option.
     If the next character is '?' this is a minimizing repeat, by default,
@@ -6923,7 +6966,8 @@ for (;; ptr++)
         slot = cd->name_table;
         for (i = 0; i < cd->names_found; i++)
           {
-          if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) == 0) break;
+          if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) == 0 &&
+            slot[IMM2_SIZE+namelen] == 0) break;
           slot += cd->name_entry_size;
           }
 
@@ -7889,15 +7933,17 @@ for (;; ptr++)
         }
       }
 
-    /* For a forward assertion, we take the reqchar, if set. This can be
-    helpful if the pattern that follows the assertion doesn't set a different
-    char. For example, it's useful for /(?=abcde).+/. We can't set firstchar
-    for an assertion, however because it leads to incorrect effect for patterns
-    such as /(?=a)a.+/ when the "real" "a" would then become a reqchar instead
-    of a firstchar. This is overcome by a scan at the end if there's no
-    firstchar, looking for an asserted first char. */
-
-    else if (bravalue == OP_ASSERT && subreqcharflags >= 0)
+    /* For a forward assertion, we take the reqchar, if set, provided that the
+    group has also set a first char. This can be helpful if the pattern that
+    follows the assertion doesn't set a different char. For example, it's
+    useful for /(?=abcde).+/. We can't set firstchar for an assertion, however
+    because it leads to incorrect effect for patterns such as /(?=a)a.+/ when
+    the "real" "a" would then become a reqchar instead of a firstchar. This is
+    overcome by a scan at the end if there's no firstchar, looking for an
+    asserted first char. */
+
+    else if (bravalue == OP_ASSERT && subreqcharflags >= 0 &&
+             subfirstcharflags >= 0)
       {
       reqchar = subreqchar;
       reqcharflags = subreqcharflags;
@@ -8179,7 +8225,6 @@ for (;; ptr++)
 
       if (mclength == 1 || req_caseopt == 0)
         {
-        firstchar = mcbuffer[0] | req_caseopt;
         firstchar = mcbuffer[0];
         firstcharflags = req_caseopt;
 
@@ -8686,8 +8731,8 @@ matching and for non-DOTALL patterns that start with .* (which must start at
 the beginning or after \n). As in the case of is_anchored() (see above), we
 have to take account of back references to capturing brackets that contain .*
 because in that case we can't make the assumption. Also, the appearance of .*
-inside atomic brackets or in a pattern that contains *PRUNE or *SKIP does not
-count, because once again the assumption no longer holds.
+inside atomic brackets or in an assertion, or in a pattern that contains *PRUNE
+or *SKIP does not count, because once again the assumption no longer holds.
 
 Arguments:
   code           points to start of expression (the bracket)
@@ -8696,13 +8741,14 @@ Arguments:
                   the less precise approach
   cd             points to the compile data
   atomcount      atomic group level
+  inassert       TRUE if in an assertion
 
 Returns:         TRUE or FALSE
 */
 
 static BOOL
 is_startline(const pcre_uchar *code, unsigned int bracket_map,
-  compile_data *cd, int atomcount)
+  compile_data *cd, int atomcount, BOOL inassert)
 {
 do {
    const pcre_uchar *scode = first_significant_code(
@@ -8729,7 +8775,7 @@ do {
        return FALSE;
 
        default:     /* Assertion */
-       if (!is_startline(scode, bracket_map, cd, atomcount)) return FALSE;
+       if (!is_startline(scode, bracket_map, cd, atomcount, TRUE)) return FALSE;
        do scode += GET(scode, 1); while (*scode == OP_ALT);
        scode += 1 + LINK_SIZE;
        break;
@@ -8743,7 +8789,7 @@ do {
    if (op == OP_BRA  || op == OP_BRAPOS ||
        op == OP_SBRA || op == OP_SBRAPOS)
      {
-     if (!is_startline(scode, bracket_map, cd, atomcount)) return FALSE;
+     if (!is_startline(scode, bracket_map, cd, atomcount, inassert)) return FALSE;
      }
 
    /* Capturing brackets */
@@ -8753,33 +8799,33 @@ do {
      {
      int n = GET2(scode, 1+LINK_SIZE);
      int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
-     if (!is_startline(scode, new_map, cd, atomcount)) return FALSE;
+     if (!is_startline(scode, new_map, cd, atomcount, inassert)) return FALSE;
      }
 
    /* Positive forward assertions */
 
    else if (op == OP_ASSERT)
      {
-     if (!is_startline(scode, bracket_map, cd, atomcount)) return FALSE;
+     if (!is_startline(scode, bracket_map, cd, atomcount, TRUE)) return FALSE;
      }
 
    /* Atomic brackets */
 
    else if (op == OP_ONCE || op == OP_ONCE_NC)
      {
-     if (!is_startline(scode, bracket_map, cd, atomcount + 1)) return FALSE;
+     if (!is_startline(scode, bracket_map, cd, atomcount + 1, inassert)) return FALSE;
      }
 
    /* .* means "start at start or after \n" if it isn't in atomic brackets or
-   brackets that may be referenced, as long as the pattern does not contain
-   *PRUNE or *SKIP, because these break the feature. Consider, for example,
-   /.*?a(*PRUNE)b/ with the subject "aab", which matches "ab", i.e. not at the
-   start of a line. */
+   brackets that may be referenced or an assertion, as long as the pattern does
+   not contain *PRUNE or *SKIP, because these break the feature. Consider, for
+   example, /.*?a(*PRUNE)b/ with the subject "aab", which matches "ab", i.e.
+   not at the start of a line. */
 
    else if (op == OP_TYPESTAR || op == OP_TYPEMINSTAR || op == OP_TYPEPOSSTAR)
      {
      if (scode[1] != OP_ANY || (bracket_map & cd->backref_map) != 0 ||
-         atomcount > 0 || cd->had_pruneorskip)
+         atomcount > 0 || cd->had_pruneorskip || inassert)
        return FALSE;
      }
 
@@ -9634,7 +9680,7 @@ if ((re->options & PCRE_ANCHORED) == 0)
       re->flags |= PCRE_FIRSTSET;
       }
 
-    else if (is_startline(codestart, 0, cd, 0)) re->flags |= PCRE_STARTLINE;
+    else if (is_startline(codestart, 0, cd, 0, FALSE)) re->flags |= PCRE_STARTLINE;
     }
   }
 
diff --git a/src/third_party/pcre-8.39/pcre_config.c b/src/third_party/pcre-8.41/pcre_config.c
index 1cbdd9c960c..1cbdd9c960c 100644
--- a/src/third_party/pcre-8.39/pcre_config.c
+++ b/src/third_party/pcre-8.41/pcre_config.c
diff --git a/src/third_party/pcre-8.39/pcre_dfa_exec.c b/src/third_party/pcre-8.41/pcre_dfa_exec.c
index 170ce6a0016..bc09ced3a7c 100644
--- a/src/third_party/pcre-8.39/pcre_dfa_exec.c
+++ b/src/third_party/pcre-8.41/pcre_dfa_exec.c
@@ -7,7 +7,7 @@ and semantics are as close as possible to those of the Perl 5 language (but see
 below for why this module is different).
 
                        Written by Philip Hazel
-           Copyright (c) 1997-2014 University of Cambridge
+           Copyright (c) 1997-2017 University of Cambridge
 
 -----------------------------------------------------------------------------
 Redistribution and use in source and binary forms, with or without
@@ -2625,7 +2625,7 @@ for (;;)
           if (isinclass)
             {
             int max = (int)GET2(ecode, 1 + IMM2_SIZE);
-            if (*ecode == OP_CRPOSRANGE)
+            if (*ecode == OP_CRPOSRANGE && count >= (int)GET2(ecode, 1))
               {
               active_count--;           /* Remove non-match possibility */
               next_active_state--;
diff --git a/src/third_party/pcre-8.39/pcre_exec.c b/src/third_party/pcre-8.41/pcre_exec.c
index 24b23ca2864..1a9bdd546ee 100644
--- a/src/third_party/pcre-8.39/pcre_exec.c
+++ b/src/third_party/pcre-8.41/pcre_exec.c
@@ -669,7 +669,7 @@ if (ecode == NULL)
     return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
   else
     {
-    int len = (char *)&rdepth - (char *)eptr;
+    int len = (int)((char *)&rdepth - (char *)eptr);
     return (len > 0)? -len : len;
     }
   }
diff --git a/src/third_party/pcre-8.39/pcre_fullinfo.c b/src/third_party/pcre-8.41/pcre_fullinfo.c
index a6c2ece6ca5..a6c2ece6ca5 100644
--- a/src/third_party/pcre-8.39/pcre_fullinfo.c
+++ b/src/third_party/pcre-8.41/pcre_fullinfo.c
diff --git a/src/third_party/pcre-8.39/pcre_get.c b/src/third_party/pcre-8.41/pcre_get.c
index 9475d5e88cd..9475d5e88cd 100644
--- a/src/third_party/pcre-8.39/pcre_get.c
+++ b/src/third_party/pcre-8.41/pcre_get.c
diff --git a/src/third_party/pcre-8.39/pcre_globals.c b/src/third_party/pcre-8.41/pcre_globals.c
index 0f106aa9013..0f106aa9013 100644
--- a/src/third_party/pcre-8.39/pcre_globals.c
+++ b/src/third_party/pcre-8.41/pcre_globals.c
diff --git a/src/third_party/pcre-8.39/pcre_internal.h b/src/third_party/pcre-8.41/pcre_internal.h
index 2923b29f82d..97ff55d03b3 100644
--- a/src/third_party/pcre-8.39/pcre_internal.h
+++ b/src/third_party/pcre-8.41/pcre_internal.h
@@ -2772,6 +2772,9 @@ extern const pcre_uint8  PRIV(ucd_stage1)[];
 extern const pcre_uint16 PRIV(ucd_stage2)[];
 extern const pcre_uint32 PRIV(ucp_gentype)[];
 extern const pcre_uint32 PRIV(ucp_gbtable)[];
+#ifdef COMPILE_PCRE32
+extern const ucd_record  PRIV(dummy_ucd_record)[];
+#endif
 #ifdef SUPPORT_JIT
 extern const int         PRIV(ucp_typerange)[];
 #endif
@@ -2780,10 +2783,16 @@ extern const int         PRIV(ucp_typerange)[];
 /* UCD access macros */
 
 #define UCD_BLOCK_SIZE 128
-#define GET_UCD(ch) (PRIV(ucd_records) + \
+#define REAL_GET_UCD(ch) (PRIV(ucd_records) + \
         PRIV(ucd_stage2)[PRIV(ucd_stage1)[(int)(ch) / UCD_BLOCK_SIZE] * \
         UCD_BLOCK_SIZE + (int)(ch) % UCD_BLOCK_SIZE])
 
+#ifdef COMPILE_PCRE32
+#define GET_UCD(ch) ((ch > 0x10ffff)? PRIV(dummy_ucd_record) : REAL_GET_UCD(ch))
+#else
+#define GET_UCD(ch) REAL_GET_UCD(ch)
+#endif
+
 #define UCD_CHARTYPE(ch)    GET_UCD(ch)->chartype
 #define UCD_SCRIPT(ch)      GET_UCD(ch)->script
 #define UCD_CATEGORY(ch)    PRIV(ucp_gentype)[UCD_CHARTYPE(ch)]
diff --git a/src/third_party/pcre-8.39/pcre_jit_compile.c b/src/third_party/pcre-8.41/pcre_jit_compile.c
index 4f15a27ac28..249edbe8e7f 100644
--- a/src/third_party/pcre-8.39/pcre_jit_compile.c
+++ b/src/third_party/pcre-8.41/pcre_jit_compile.c
@@ -487,7 +487,7 @@ typedef struct compare_context {
 #undef CMP
 
 /* Used for accessing the elements of the stack. */
-#define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_sw))
+#define STACK(i)      ((i) * (int)sizeof(sljit_sw))
 
 #define TMP1          SLJIT_R0
 #define TMP2          SLJIT_R2
@@ -552,13 +552,15 @@ the start pointers when the end of the capturing group has not yet reached. */
   sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))
 #define CMPTO(type, src1, src1w, src2, src2w, label) \
   sljit_set_label(sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w)), (label))
-#define OP_FLAGS(op, dst, dstw, src, srcw, type) \
-  sljit_emit_op_flags(compiler, (op), (dst), (dstw), (src), (srcw), (type))
+#define OP_FLAGS(op, dst, dstw, type) \
+  sljit_emit_op_flags(compiler, (op), (dst), (dstw), (type))
 #define GET_LOCAL_BASE(dst, dstw, offset) \
   sljit_get_local_base(compiler, (dst), (dstw), (offset))
 
 #define READ_CHAR_MAX 0x7fffffff
 
+#define INVALID_UTF_CHAR 888
+
 static pcre_uchar *bracketend(pcre_uchar *cc)
 {
 SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));
@@ -784,7 +786,7 @@ switch(*cc)
 
   default:
   /* All opcodes are supported now! */
-  SLJIT_ASSERT_STOP();
+  SLJIT_UNREACHABLE();
   return NULL;
   }
 }
@@ -1660,9 +1662,9 @@ while (cc < ccend)
       {
       OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(0));
       OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
-      stackpos += (int)sizeof(sljit_sw);
+      stackpos -= (int)sizeof(sljit_sw);
       OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
-      stackpos += (int)sizeof(sljit_sw);
+      stackpos -= (int)sizeof(sljit_sw);
       setsom_found = TRUE;
       }
     cc += 1;
@@ -1676,9 +1678,9 @@ while (cc < ccend)
       {
       OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->mark_ptr);
       OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
-      stackpos += (int)sizeof(sljit_sw);
+      stackpos -= (int)sizeof(sljit_sw);
       OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
-      stackpos += (int)sizeof(sljit_sw);
+      stackpos -= (int)sizeof(sljit_sw);
       setmark_found = TRUE;
       }
     cc += 1 + 2 + cc[1];
@@ -1689,27 +1691,27 @@ while (cc < ccend)
       {
       OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(0));
       OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
-      stackpos += (int)sizeof(sljit_sw);
+      stackpos -= (int)sizeof(sljit_sw);
       OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
-      stackpos += (int)sizeof(sljit_sw);
+      stackpos -= (int)sizeof(sljit_sw);
       setsom_found = TRUE;
       }
     if (common->mark_ptr != 0 && !setmark_found)
       {
       OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->mark_ptr);
       OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
-      stackpos += (int)sizeof(sljit_sw);
+      stackpos -= (int)sizeof(sljit_sw);
       OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
-      stackpos += (int)sizeof(sljit_sw);
+      stackpos -= (int)sizeof(sljit_sw);
       setmark_found = TRUE;
       }
     if (common->capture_last_ptr != 0 && !capture_last_found)
       {
       OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->capture_last_ptr);
       OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
-      stackpos += (int)sizeof(sljit_sw);
+      stackpos -= (int)sizeof(sljit_sw);
       OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
-      stackpos += (int)sizeof(sljit_sw);
+      stackpos -= (int)sizeof(sljit_sw);
       capture_last_found = TRUE;
       }
     cc += 1 + LINK_SIZE;
@@ -1723,20 +1725,20 @@ while (cc < ccend)
       {
       OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->capture_last_ptr);
       OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
-      stackpos += (int)sizeof(sljit_sw);
+      stackpos -= (int)sizeof(sljit_sw);
       OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
-      stackpos += (int)sizeof(sljit_sw);
+      stackpos -= (int)sizeof(sljit_sw);
       capture_last_found = TRUE;
       }
     offset = (GET2(cc, 1 + LINK_SIZE)) << 1;
     OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));
-    stackpos += (int)sizeof(sljit_sw);
+    stackpos -= (int)sizeof(sljit_sw);
     OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(offset));
     OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(offset + 1));
     OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
-    stackpos += (int)sizeof(sljit_sw);
+    stackpos -= (int)sizeof(sljit_sw);
     OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);
-    stackpos += (int)sizeof(sljit_sw);
+    stackpos -= (int)sizeof(sljit_sw);
 
     cc += 1 + LINK_SIZE + IMM2_SIZE;
     break;
@@ -1887,18 +1889,17 @@ BOOL tmp1empty = TRUE;
 BOOL tmp2empty = TRUE;
 pcre_uchar *alternative;
 enum {
-  start,
   loop,
   end
 } status;
 
-status = save ? start : loop;
-stackptr = STACK(stackptr - 2);
+status = loop;
+stackptr = STACK(stackptr);
 stacktop = STACK(stacktop - 1);
 
 if (!save)
   {
-  stackptr += (needs_control_head ? 2 : 1) * sizeof(sljit_sw);
+  stacktop -= (needs_control_head ? 2 : 1) * sizeof(sljit_sw);
   if (stackptr < stacktop)
     {
     OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
@@ -1914,196 +1915,186 @@ if (!save)
   /* The tmp1next must be TRUE in either way. */
   }
 
+SLJIT_ASSERT(common->recursive_head_ptr != 0);
+
 do
   {
   count = 0;
-  switch(status)
+  if (cc >= ccend)
     {
-    case start:
-    SLJIT_ASSERT(save && common->recursive_head_ptr != 0);
+    if (!save)
+      break;
+
     count = 1;
     srcw[0] = common->recursive_head_ptr;
     if (needs_control_head)
       {
       SLJIT_ASSERT(common->control_head_ptr != 0);
       count = 2;
-      srcw[1] = common->control_head_ptr;
+      srcw[0] = common->control_head_ptr;
+      srcw[1] = common->recursive_head_ptr;
+      }
+    status = end;
+    }
+  else switch(*cc)
+    {
+    case OP_KET:
+    if (PRIVATE_DATA(cc) != 0)
+      {
+      count = 1;
+      srcw[0] = PRIVATE_DATA(cc);
+      SLJIT_ASSERT(PRIVATE_DATA(cc + 1) != 0);
+      cc += PRIVATE_DATA(cc + 1);
       }
-    status = loop;
+    cc += 1 + LINK_SIZE;
+    break;
+
+    case OP_ASSERT:
+    case OP_ASSERT_NOT:
+    case OP_ASSERTBACK:
+    case OP_ASSERTBACK_NOT:
+    case OP_ONCE:
+    case OP_ONCE_NC:
+    case OP_BRAPOS:
+    case OP_SBRA:
+    case OP_SBRAPOS:
+    case OP_SCOND:
+    count = 1;
+    srcw[0] = PRIVATE_DATA(cc);
+    SLJIT_ASSERT(srcw[0] != 0);
+    cc += 1 + LINK_SIZE;
     break;
 
-    case loop:
-    if (cc >= ccend)
+    case OP_CBRA:
+    case OP_SCBRA:
+    if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
       {
-      status = end;
-      break;
+      count = 1;
+      srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
       }
+    cc += 1 + LINK_SIZE + IMM2_SIZE;
+    break;
 
-    switch(*cc)
-      {
-      case OP_KET:
-      if (PRIVATE_DATA(cc) != 0)
-        {
-        count = 1;
-        srcw[0] = PRIVATE_DATA(cc);
-        SLJIT_ASSERT(PRIVATE_DATA(cc + 1) != 0);
-        cc += PRIVATE_DATA(cc + 1);
-        }
-      cc += 1 + LINK_SIZE;
-      break;
+    case OP_CBRAPOS:
+    case OP_SCBRAPOS:
+    count = 2;
+    srcw[0] = PRIVATE_DATA(cc);
+    srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
+    SLJIT_ASSERT(srcw[0] != 0 && srcw[1] != 0);
+    cc += 1 + LINK_SIZE + IMM2_SIZE;
+    break;
 
-      case OP_ASSERT:
-      case OP_ASSERT_NOT:
-      case OP_ASSERTBACK:
-      case OP_ASSERTBACK_NOT:
-      case OP_ONCE:
-      case OP_ONCE_NC:
-      case OP_BRAPOS:
-      case OP_SBRA:
-      case OP_SBRAPOS:
-      case OP_SCOND:
+    case OP_COND:
+    /* Might be a hidden SCOND. */
+    alternative = cc + GET(cc, 1);
+    if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
+      {
       count = 1;
       srcw[0] = PRIVATE_DATA(cc);
       SLJIT_ASSERT(srcw[0] != 0);
-      cc += 1 + LINK_SIZE;
-      break;
-
-      case OP_CBRA:
-      case OP_SCBRA:
-      if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
-        {
-        count = 1;
-        srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
-        }
-      cc += 1 + LINK_SIZE + IMM2_SIZE;
-      break;
+      }
+    cc += 1 + LINK_SIZE;
+    break;
 
-      case OP_CBRAPOS:
-      case OP_SCBRAPOS:
-      count = 2;
+    CASE_ITERATOR_PRIVATE_DATA_1
+    if (PRIVATE_DATA(cc))
+      {
+      count = 1;
       srcw[0] = PRIVATE_DATA(cc);
-      srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
-      SLJIT_ASSERT(srcw[0] != 0 && srcw[1] != 0);
-      cc += 1 + LINK_SIZE + IMM2_SIZE;
-      break;
-
-      case OP_COND:
-      /* Might be a hidden SCOND. */
-      alternative = cc + GET(cc, 1);
-      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
-        {
-        count = 1;
-        srcw[0] = PRIVATE_DATA(cc);
-        SLJIT_ASSERT(srcw[0] != 0);
-        }
-      cc += 1 + LINK_SIZE;
-      break;
-
-      CASE_ITERATOR_PRIVATE_DATA_1
-      if (PRIVATE_DATA(cc))
-        {
-        count = 1;
-        srcw[0] = PRIVATE_DATA(cc);
-        }
-      cc += 2;
+      }
+    cc += 2;
 #ifdef SUPPORT_UTF
-      if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
+    if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
 #endif
-      break;
+    break;
 
-      CASE_ITERATOR_PRIVATE_DATA_2A
-      if (PRIVATE_DATA(cc))
-        {
-        count = 2;
-        srcw[0] = PRIVATE_DATA(cc);
-        srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
-        }
-      cc += 2;
+    CASE_ITERATOR_PRIVATE_DATA_2A
+    if (PRIVATE_DATA(cc))
+      {
+      count = 2;
+      srcw[0] = PRIVATE_DATA(cc);
+      srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
+      }
+    cc += 2;
 #ifdef SUPPORT_UTF
-      if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
+    if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
 #endif
-      break;
+    break;
 
-      CASE_ITERATOR_PRIVATE_DATA_2B
-      if (PRIVATE_DATA(cc))
-        {
-        count = 2;
-        srcw[0] = PRIVATE_DATA(cc);
-        srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
-        }
-      cc += 2 + IMM2_SIZE;
+    CASE_ITERATOR_PRIVATE_DATA_2B
+    if (PRIVATE_DATA(cc))
+      {
+      count = 2;
+      srcw[0] = PRIVATE_DATA(cc);
+      srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
+      }
+    cc += 2 + IMM2_SIZE;
 #ifdef SUPPORT_UTF
-      if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
+    if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
 #endif
-      break;
+    break;
 
-      CASE_ITERATOR_TYPE_PRIVATE_DATA_1
-      if (PRIVATE_DATA(cc))
+    CASE_ITERATOR_TYPE_PRIVATE_DATA_1
+    if (PRIVATE_DATA(cc))
+      {
+      count = 1;
+      srcw[0] = PRIVATE_DATA(cc);
+      }
+    cc += 1;
+    break;
+
+    CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
+    if (PRIVATE_DATA(cc))
+      {
+      count = 2;
+      srcw[0] = PRIVATE_DATA(cc);
+      srcw[1] = srcw[0] + sizeof(sljit_sw);
+      }
+    cc += 1;
+    break;
+
+    CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
+    if (PRIVATE_DATA(cc))
+      {
+      count = 2;
+      srcw[0] = PRIVATE_DATA(cc);
+      srcw[1] = srcw[0] + sizeof(sljit_sw);
+      }
+    cc += 1 + IMM2_SIZE;
+    break;
+
+    case OP_CLASS:
+    case OP_NCLASS:
+#if defined SUPPORT_UTF || !defined COMPILE_PCRE8
+    case OP_XCLASS:
+    size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
+#else
+    size = 1 + 32 / (int)sizeof(pcre_uchar);
+#endif
+    if (PRIVATE_DATA(cc))
+      switch(get_class_iterator_size(cc + size))
         {
+        case 1:
         count = 1;
         srcw[0] = PRIVATE_DATA(cc);
-        }
-      cc += 1;
-      break;
+        break;
 
-      CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
-      if (PRIVATE_DATA(cc))
-        {
+        case 2:
         count = 2;
         srcw[0] = PRIVATE_DATA(cc);
         srcw[1] = srcw[0] + sizeof(sljit_sw);
-        }
-      cc += 1;
-      break;
+        break;
 
-      CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
-      if (PRIVATE_DATA(cc))
-        {
-        count = 2;
-        srcw[0] = PRIVATE_DATA(cc);
-        srcw[1] = srcw[0] + sizeof(sljit_sw);
+        default:
+        SLJIT_UNREACHABLE();
+        break;
         }
-      cc += 1 + IMM2_SIZE;
-      break;
-
-      case OP_CLASS:
-      case OP_NCLASS:
-#if defined SUPPORT_UTF || !defined COMPILE_PCRE8
-      case OP_XCLASS:
-      size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
-#else
-      size = 1 + 32 / (int)sizeof(pcre_uchar);
-#endif
-      if (PRIVATE_DATA(cc))
-        switch(get_class_iterator_size(cc + size))
-          {
-          case 1:
-          count = 1;
-          srcw[0] = PRIVATE_DATA(cc);
-          break;
-
-          case 2:
-          count = 2;
-          srcw[0] = PRIVATE_DATA(cc);
-          srcw[1] = srcw[0] + sizeof(sljit_sw);
-          break;
-
-          default:
-          SLJIT_ASSERT_STOP();
-          break;
-          }
-      cc += size;
-      break;
-
-      default:
-      cc = next_opcode(common, cc);
-      SLJIT_ASSERT(cc != NULL);
-      break;
-      }
+    cc += size;
     break;
 
-    case end:
-    SLJIT_ASSERT_STOP();
+    default:
+    cc = next_opcode(common, cc);
+    SLJIT_ASSERT(cc != NULL);
     break;
     }
 
@@ -2312,7 +2303,7 @@ static SLJIT_INLINE void count_match(compiler_common *common)
 {
 DEFINE_COMPILER;
 
-OP2(SLJIT_SUB | SLJIT_SET_E, COUNT_MATCH, 0, COUNT_MATCH, 0, SLJIT_IMM, 1);
+OP2(SLJIT_SUB | SLJIT_SET_Z, COUNT_MATCH, 0, COUNT_MATCH, 0, SLJIT_IMM, 1);
 add_jump(compiler, &common->calllimit, JUMP(SLJIT_ZERO));
 }
 
@@ -2322,7 +2313,7 @@ static SLJIT_INLINE void allocate_stack(compiler_common *common, int size)
 DEFINE_COMPILER;
 
 SLJIT_ASSERT(size > 0);
-OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_sw));
+OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_sw));
 #ifdef DESTROY_REGISTERS
 OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);
 OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
@@ -2330,7 +2321,7 @@ OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);
 OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), LOCALS0, TMP1, 0);
 OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), LOCALS1, TMP1, 0);
 #endif
-add_stub(common, CMP(SLJIT_GREATER, STACK_TOP, 0, STACK_LIMIT, 0));
+add_stub(common, CMP(SLJIT_LESS, STACK_TOP, 0, STACK_LIMIT, 0));
 }
 
 static SLJIT_INLINE void free_stack(compiler_common *common, int size)
@@ -2338,7 +2329,7 @@ static SLJIT_INLINE void free_stack(compiler_common *common, int size)
 DEFINE_COMPILER;
 
 SLJIT_ASSERT(size > 0);
-OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_sw));
+OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_sw));
 }
 
 static sljit_uw * allocate_read_only_data(compiler_common *common, sljit_uw size)
@@ -2396,7 +2387,7 @@ else
   OP1(SLJIT_MOV, SLJIT_R2, 0, SLJIT_IMM, length - 1);
   loop = LABEL();
   OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_R1), sizeof(sljit_sw), SLJIT_R0, 0);
-  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_IMM, 1);
+  OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_IMM, 1);
   JUMPTO(SLJIT_NOT_ZERO, loop);
   }
 }
@@ -2434,7 +2425,7 @@ else
   OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_IMM, length - 2);
   loop = LABEL();
   OP1(SLJIT_MOVU, SLJIT_MEM1(TMP2), sizeof(sljit_sw), TMP1, 0);
-  OP2(SLJIT_SUB | SLJIT_SET_E, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, 1);
+  OP2(SLJIT_SUB | SLJIT_SET_Z, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, 1);
   JUMPTO(SLJIT_NOT_ZERO, loop);
   }
 
@@ -2452,22 +2443,22 @@ static sljit_sw SLJIT_CALL do_search_mark(sljit_sw *current, const pcre_uchar *s
 {
 while (current != NULL)
   {
-  switch (current[-2])
+  switch (current[1])
     {
     case type_then_trap:
     break;
 
     case type_mark:
-    if (STRCMP_UC_UC(skip_arg, (pcre_uchar *)current[-3]) == 0)
-      return current[-4];
+    if (STRCMP_UC_UC(skip_arg, (pcre_uchar *)current[2]) == 0)
+      return current[3];
     break;
 
     default:
-    SLJIT_ASSERT_STOP();
+    SLJIT_UNREACHABLE();
     break;
     }
-  SLJIT_ASSERT(current > (sljit_sw*)current[-1]);
-  current = (sljit_sw*)current[-1];
+  SLJIT_ASSERT(current[0] == 0 || current < (sljit_sw*)current[0]);
+  current = (sljit_sw*)current[0];
   }
 return -1;
 }
@@ -2501,7 +2492,7 @@ OP2(SLJIT_ADD, SLJIT_S0, 0, SLJIT_S0, 0, SLJIT_IMM, sizeof(sljit_sw));
 OP2(SLJIT_ASHR, SLJIT_S1, 0, SLJIT_S1, 0, SLJIT_IMM, UCHAR_SHIFT);
 #endif
 OP1(SLJIT_MOVU_S32, SLJIT_MEM1(SLJIT_R2), sizeof(int), SLJIT_S1, 0);
-OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_R1, 0, SLJIT_R1, 0, SLJIT_IMM, 1);
+OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_R1, 0, SLJIT_R1, 0, SLJIT_IMM, 1);
 JUMPTO(SLJIT_NOT_ZERO, loop);
 JUMPHERE(early_quit);
 
@@ -3106,8 +3097,8 @@ if (common->utf)
   OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
   /* Skip low surrogate if necessary. */
   OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
-  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xdc00);
-  OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
+  OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xdc00);
+  OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_EQUAL);
   OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
   OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
   return;
@@ -3126,6 +3117,7 @@ struct sljit_jump *jump;
 if (nltype == NLTYPE_ANY)
   {
   add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));
+  sljit_set_current_flags(compiler, SLJIT_SET_Z);
   add_jump(compiler, backtracks, JUMP(jumpifmatch ? SLJIT_NOT_ZERO : SLJIT_ZERO));
   }
 else if (nltype == NLTYPE_ANYCRLF)
@@ -3167,7 +3159,7 @@ OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
 
 /* Searching for the first zero. */
-OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
+OP2(SLJIT_AND | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
 jump = JUMP(SLJIT_NOT_ZERO);
 /* Two byte sequence. */
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
@@ -3181,7 +3173,7 @@ OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
 OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
 
-OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10000);
+OP2(SLJIT_AND | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10000);
 jump = JUMP(SLJIT_NOT_ZERO);
 /* Three byte sequence. */
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
@@ -3215,15 +3207,15 @@ OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
 
 /* Searching for the first zero. */
-OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
+OP2(SLJIT_AND | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
 jump = JUMP(SLJIT_NOT_ZERO);
 /* Two byte sequence. */
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
 
 JUMPHERE(jump);
-OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x400);
-OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_NOT_ZERO);
+OP2(SLJIT_AND | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x400);
+OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_NOT_ZERO);
 /* This code runs only in 8 bit mode. No need to shift the value. */
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
 OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
@@ -3246,7 +3238,7 @@ struct sljit_jump *compare;
 
 sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
 
-OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0x20);
+OP2(SLJIT_AND | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0x20);
 jump = JUMP(SLJIT_NOT_ZERO);
 /* Two byte sequence. */
 OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
@@ -3287,10 +3279,30 @@ static void do_getucd(compiler_common *common)
 /* Search the UCD record for the character comes in TMP1.
 Returns chartype in TMP1 and UCD offset in TMP2. */
 DEFINE_COMPILER;
+#ifdef COMPILE_PCRE32
+struct sljit_jump *jump;
+#endif
+
+#if defined SLJIT_DEBUG && SLJIT_DEBUG
+/* dummy_ucd_record */
+const ucd_record *record = GET_UCD(INVALID_UTF_CHAR);
+SLJIT_ASSERT(record->script == ucp_Common && record->chartype == ucp_Cn && record->gbprop == ucp_gbOther);
+SLJIT_ASSERT(record->caseset == 0 && record->other_case == 0);
+#endif
 
 SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && sizeof(ucd_record) == 8);
 
 sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
+
+#ifdef COMPILE_PCRE32
+if (!common->utf)
+  {
+  jump = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0x10ffff + 1);
+  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, INVALID_UTF_CHAR);
+  JUMPHERE(jump);
+  }
+#endif
+
 OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
 OP1(SLJIT_MOV_U8, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(ucd_stage1));
 OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);
@@ -3365,8 +3377,8 @@ if (newlinecheck)
   OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
   end = CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
   OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
-  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, common->newline & 0xff);
-  OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
+  OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, common->newline & 0xff);
+  OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_EQUAL);
 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
   OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, UCHAR_SHIFT);
 #endif
@@ -3403,8 +3415,8 @@ if (common->utf)
   {
   singlechar = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xd800);
   OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
-  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
-  OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
+  OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
+  OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_EQUAL);
   OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
   OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
   JUMPHERE(singlechar);
@@ -3853,7 +3865,7 @@ while (TRUE)
   }
 }
 
-#if (defined SLJIT_CONFIG_X86 && SLJIT_CONFIG_X86)
+#if (defined SLJIT_CONFIG_X86 && SLJIT_CONFIG_X86) && !(defined SUPPORT_VALGRIND)
 
 static sljit_s32 character_to_int32(pcre_uchar chr)
 {
@@ -4004,12 +4016,12 @@ sljit_emit_op_custom(compiler, instruction, 4);
 
 if (load_twice)
   {
-  OP1(SLJIT_MOV, TMP3, 0, TMP2, 0);
+  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP2, 0);
   instruction[3] = 0xc0 | (tmp2_ind << 3) | 1;
   sljit_emit_op_custom(compiler, instruction, 4);
 
   OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
-  OP1(SLJIT_MOV, TMP2, 0, TMP3, 0);
+  OP1(SLJIT_MOV, TMP2, 0, RETURN_ADDR, 0);
   }
 
 OP2(SLJIT_ASHR, TMP1, 0, TMP1, 0, TMP2, 0);
@@ -4019,6 +4031,7 @@ instruction[0] = 0x0f;
 instruction[1] = 0xbc;
 instruction[2] = 0xc0 | (tmp1_ind << 3) | tmp1_ind;
 sljit_emit_op_custom(compiler, instruction, 3);
+sljit_set_current_flags(compiler, SLJIT_SET_Z);
 
 nomatch = JUMP(SLJIT_ZERO);
 
@@ -4119,6 +4132,7 @@ instruction[0] = 0x0f;
 instruction[1] = 0xbc;
 instruction[2] = 0xc0 | (tmp1_ind << 3) | tmp1_ind;
 sljit_emit_op_custom(compiler, instruction, 3);
+sljit_set_current_flags(compiler, SLJIT_SET_Z);
 
 JUMPTO(SLJIT_ZERO, start);
 
@@ -4155,18 +4169,8 @@ if (has_match_end)
   OP1(SLJIT_MOV, TMP3, 0, STR_END, 0);
 
   OP2(SLJIT_ADD, STR_END, 0, SLJIT_MEM1(SLJIT_SP), common->match_end_ptr, SLJIT_IMM, IN_UCHARS(offset + 1));
-#if (defined SLJIT_CONFIG_X86 && SLJIT_CONFIG_X86)
-  if (sljit_x86_is_cmov_available())
-    {
-    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, STR_END, 0, TMP3, 0);
-    sljit_x86_emit_cmov(compiler, SLJIT_GREATER, STR_END, TMP3, 0);
-    }
-#endif
-    {
-    quit = CMP(SLJIT_LESS_EQUAL, STR_END, 0, TMP3, 0);
-    OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);
-    JUMPHERE(quit);
-    }
+  OP2(SLJIT_SUB | SLJIT_SET_GREATER, SLJIT_UNUSED, 0, STR_END, 0, TMP3, 0);
+  sljit_emit_cmov(compiler, SLJIT_GREATER, STR_END, TMP3, 0);
   }
 
 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
@@ -4174,11 +4178,11 @@ if (common->utf && offset > 0)
   utf_start = LABEL();
 #endif
 
-#if (defined SLJIT_CONFIG_X86 && SLJIT_CONFIG_X86)
+#if (defined SLJIT_CONFIG_X86 && SLJIT_CONFIG_X86) && !(defined SUPPORT_VALGRIND)
 
 /* SSE2 accelerated first character search. */
 
-if (sljit_x86_is_sse2_available())
+if (sljit_has_cpu_feature(SLJIT_HAS_SSE2))
   {
   fast_forward_first_char2_sse2(common, char1, char2);
 
@@ -4213,16 +4217,16 @@ if (sljit_x86_is_sse2_available())
     if (offset > 0)
       OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(offset));
     }
-  else if (sljit_x86_is_cmov_available())
-    {
-    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, STR_PTR, 0, STR_END, 0);
-    sljit_x86_emit_cmov(compiler, SLJIT_GREATER_EQUAL, STR_PTR, has_match_end ? SLJIT_MEM1(SLJIT_SP) : STR_END, has_match_end ? common->match_end_ptr : 0);
-    }
   else
     {
-    quit = CMP(SLJIT_LESS, STR_PTR, 0, STR_END, 0);
-    OP1(SLJIT_MOV, STR_PTR, 0, has_match_end ? SLJIT_MEM1(SLJIT_SP) : STR_END, has_match_end ? common->match_end_ptr : 0);
-    JUMPHERE(quit);
+    OP2(SLJIT_SUB | SLJIT_SET_GREATER_EQUAL, SLJIT_UNUSED, 0, STR_PTR, 0, STR_END, 0);
+    if (has_match_end)
+      {
+      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->match_end_ptr);
+      sljit_emit_cmov(compiler, SLJIT_GREATER_EQUAL, STR_PTR, TMP1, 0);
+      }
+    else
+      sljit_emit_cmov(compiler, SLJIT_GREATER_EQUAL, STR_PTR, STR_END, 0);
     }
 
   if (has_match_end)
@@ -4249,10 +4253,10 @@ else
     }
   else
     {
-    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, char1);
-    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
-    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, char2);
-    OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_EQUAL);
+    OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, char1);
+    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_EQUAL);
+    OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, char2);
+    OP_FLAGS(SLJIT_OR | SLJIT_SET_Z, TMP2, 0, SLJIT_EQUAL);
     found = JUMP(SLJIT_NOT_ZERO);
     }
   }
@@ -4571,8 +4575,8 @@ if (common->nltype == NLTYPE_FIXED && common->newline > 255)
   firstchar = CMP(SLJIT_LESS_EQUAL, STR_PTR, 0, TMP2, 0);
 
   OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, IN_UCHARS(2));
-  OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, STR_PTR, 0, TMP1, 0);
-  OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_GREATER_EQUAL);
+  OP2(SLJIT_SUB | SLJIT_SET_GREATER_EQUAL, SLJIT_UNUSED, 0, STR_PTR, 0, TMP1, 0);
+  OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_GREATER_EQUAL);
 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
   OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCHAR_SHIFT);
 #endif
@@ -4616,8 +4620,8 @@ if (common->nltype == NLTYPE_ANY || common->nltype == NLTYPE_ANYCRLF)
   JUMPHERE(foundcr);
   notfoundnl = CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
   OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
-  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_NL);
-  OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
+  OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_NL);
+  OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_EQUAL);
 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
   OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, UCHAR_SHIFT);
 #endif
@@ -4670,7 +4674,7 @@ if (!check_class_ranges(common, start_bits, (start_bits[31] & 0x80) != 0, TRUE,
   OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 3);
   OP1(SLJIT_MOV_U8, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)start_bits);
   OP2(SLJIT_SHL, TMP2, 0, SLJIT_IMM, 1, TMP2, 0);
-  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);
+  OP2(SLJIT_AND | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);
   found = JUMP(SLJIT_NOT_ZERO);
   }
 
@@ -4692,8 +4696,8 @@ if (common->utf)
   {
   CMPTO(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xd800, start);
   OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
-  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
-  OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
+  OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
+  OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_EQUAL);
   OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
   OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
   }
@@ -4780,31 +4784,31 @@ struct sljit_jump *jump;
 struct sljit_label *mainloop;
 
 sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
-OP1(SLJIT_MOV, TMP1, 0, STACK_TOP, 0);
-GET_LOCAL_BASE(TMP3, 0, 0);
+OP1(SLJIT_MOV, TMP3, 0, STACK_TOP, 0);
+GET_LOCAL_BASE(TMP1, 0, 0);
 
 /* Drop frames until we reach STACK_TOP. */
 mainloop = LABEL();
-OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(TMP1), 0);
-OP2(SLJIT_SUB | SLJIT_SET_S, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0);
-jump = JUMP(SLJIT_SIG_LESS_EQUAL);
-
-OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, TMP3, 0);
-OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), 0, SLJIT_MEM1(TMP1), sizeof(sljit_sw));
-OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), sizeof(sljit_sw), SLJIT_MEM1(TMP1), 2 * sizeof(sljit_sw));
-OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 3 * sizeof(sljit_sw));
+OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), -sizeof(sljit_sw));
+jump = CMP(SLJIT_SIG_LESS_EQUAL, TMP2, 0, SLJIT_IMM, 0);
+
+OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, TMP1, 0);
+OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), 0, SLJIT_MEM1(STACK_TOP), -2 * sizeof(sljit_sw));
+OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), sizeof(sljit_sw), SLJIT_MEM1(STACK_TOP), -3 * sizeof(sljit_sw));
+OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, 3 * sizeof(sljit_sw));
 JUMPTO(SLJIT_JUMP, mainloop);
 
 JUMPHERE(jump);
-jump = JUMP(SLJIT_SIG_LESS);
-/* End of dropping frames. */
+jump = CMP(SLJIT_NOT_ZERO /* SIG_LESS */, TMP2, 0, SLJIT_IMM, 0);
+/* End of reverting values. */
+OP1(SLJIT_MOV, STACK_TOP, 0, TMP3, 0);
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
 
 JUMPHERE(jump);
 OP1(SLJIT_NEG, TMP2, 0, TMP2, 0);
-OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, TMP3, 0);
-OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), 0, SLJIT_MEM1(TMP1), sizeof(sljit_sw));
-OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 2 * sizeof(sljit_sw));
+OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, TMP1, 0);
+OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), 0, SLJIT_MEM1(STACK_TOP), -2 * sizeof(sljit_sw));
+OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, 2 * sizeof(sljit_sw));
 JUMPTO(SLJIT_JUMP, mainloop);
 }
 
@@ -4837,11 +4841,11 @@ if (common->use_ucp)
   jump = CMP(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_UNDERSCORE);
   add_jump(compiler, &common->getucd, JUMP(SLJIT_FAST_CALL));
   OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Ll);
-  OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_Lu - ucp_Ll);
-  OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_LESS_EQUAL);
+  OP2(SLJIT_SUB | SLJIT_SET_LESS_EQUAL, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_Lu - ucp_Ll);
+  OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_LESS_EQUAL);
   OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Nd - ucp_Ll);
-  OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_No - ucp_Nd);
-  OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_LESS_EQUAL);
+  OP2(SLJIT_SUB | SLJIT_SET_LESS_EQUAL, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_No - ucp_Nd);
+  OP_FLAGS(SLJIT_OR, TMP2, 0, SLJIT_LESS_EQUAL);
   JUMPHERE(jump);
   OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), LOCALS1, TMP2, 0);
   }
@@ -4881,11 +4885,11 @@ if (common->use_ucp)
   jump = CMP(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_UNDERSCORE);
   add_jump(compiler, &common->getucd, JUMP(SLJIT_FAST_CALL));
   OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Ll);
-  OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_Lu - ucp_Ll);
-  OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_LESS_EQUAL);
+  OP2(SLJIT_SUB | SLJIT_SET_LESS_EQUAL, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_Lu - ucp_Ll);
+  OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_LESS_EQUAL);
   OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Nd - ucp_Ll);
-  OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_No - ucp_Nd);
-  OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_LESS_EQUAL);
+  OP2(SLJIT_SUB | SLJIT_SET_LESS_EQUAL, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_No - ucp_Nd);
+  OP_FLAGS(SLJIT_OR, TMP2, 0, SLJIT_LESS_EQUAL);
   JUMPHERE(jump);
   }
 else
@@ -4913,7 +4917,7 @@ else
   }
 set_jumps(skipread_list, LABEL());
 
-OP2(SLJIT_XOR | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_MEM1(SLJIT_SP), LOCALS1);
+OP2(SLJIT_XOR | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_MEM1(SLJIT_SP), LOCALS1);
 sljit_emit_fast_return(compiler, SLJIT_MEM1(SLJIT_SP), LOCALS0);
 }
 
@@ -5064,7 +5068,7 @@ switch(length)
   return TRUE;
 
   default:
-  SLJIT_ASSERT_STOP();
+  SLJIT_UNREACHABLE();
   return FALSE;
   }
 }
@@ -5077,22 +5081,22 @@ DEFINE_COMPILER;
 sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
 
 OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x0a);
-OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x0d - 0x0a);
-OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_LESS_EQUAL);
-OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x85 - 0x0a);
+OP2(SLJIT_SUB | SLJIT_SET_LESS_EQUAL, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x0d - 0x0a);
+OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_LESS_EQUAL);
+OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x85 - 0x0a);
 #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
 #ifdef COMPILE_PCRE8
 if (common->utf)
   {
 #endif
-  OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_EQUAL);
+  OP_FLAGS(SLJIT_OR, TMP2, 0, SLJIT_EQUAL);
   OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x1);
-  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x2029 - 0x0a);
+  OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x2029 - 0x0a);
 #ifdef COMPILE_PCRE8
   }
 #endif
 #endif /* SUPPORT_UTF || COMPILE_PCRE16 || COMPILE_PCRE32 */
-OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_EQUAL);
+OP_FLAGS(SLJIT_OR | SLJIT_SET_Z, TMP2, 0, SLJIT_EQUAL);
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
 }
 
@@ -5103,34 +5107,34 @@ DEFINE_COMPILER;
 
 sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
 
-OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x09);
-OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
-OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x20);
-OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_EQUAL);
-OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xa0);
+OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x09);
+OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_EQUAL);
+OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x20);
+OP_FLAGS(SLJIT_OR, TMP2, 0, SLJIT_EQUAL);
+OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xa0);
 #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
 #ifdef COMPILE_PCRE8
 if (common->utf)
   {
 #endif
-  OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_EQUAL);
-  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x1680);
-  OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_EQUAL);
-  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x180e);
-  OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_EQUAL);
+  OP_FLAGS(SLJIT_OR, TMP2, 0, SLJIT_EQUAL);
+  OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x1680);
+  OP_FLAGS(SLJIT_OR, TMP2, 0, SLJIT_EQUAL);
+  OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x180e);
+  OP_FLAGS(SLJIT_OR, TMP2, 0, SLJIT_EQUAL);
   OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x2000);
-  OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x200A - 0x2000);
-  OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_LESS_EQUAL);
-  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x202f - 0x2000);
-  OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_EQUAL);
-  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x205f - 0x2000);
-  OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_EQUAL);
-  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x3000 - 0x2000);
+  OP2(SLJIT_SUB | SLJIT_SET_LESS_EQUAL, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x200A - 0x2000);
+  OP_FLAGS(SLJIT_OR, TMP2, 0, SLJIT_LESS_EQUAL);
+  OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x202f - 0x2000);
+  OP_FLAGS(SLJIT_OR, TMP2, 0, SLJIT_EQUAL);
+  OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x205f - 0x2000);
+  OP_FLAGS(SLJIT_OR, TMP2, 0, SLJIT_EQUAL);
+  OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x3000 - 0x2000);
 #ifdef COMPILE_PCRE8
   }
 #endif
 #endif /* SUPPORT_UTF || COMPILE_PCRE16 || COMPILE_PCRE32 */
-OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_EQUAL);
+OP_FLAGS(SLJIT_OR | SLJIT_SET_Z, TMP2, 0, SLJIT_EQUAL);
 
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
 }
@@ -5143,22 +5147,22 @@ DEFINE_COMPILER;
 sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
 
 OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x0a);
-OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x0d - 0x0a);
-OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_LESS_EQUAL);
-OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x85 - 0x0a);
+OP2(SLJIT_SUB | SLJIT_SET_LESS_EQUAL, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x0d - 0x0a);
+OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_LESS_EQUAL);
+OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x85 - 0x0a);
 #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
 #ifdef COMPILE_PCRE8
 if (common->utf)
   {
 #endif
-  OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_EQUAL);
+  OP_FLAGS(SLJIT_OR, TMP2, 0, SLJIT_EQUAL);
   OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x1);
-  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x2029 - 0x0a);
+  OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x2029 - 0x0a);
 #ifdef COMPILE_PCRE8
   }
 #endif
 #endif /* SUPPORT_UTF || COMPILE_PCRE16 || COMPILE_PCRE32 */
-OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_EQUAL);
+OP_FLAGS(SLJIT_OR | SLJIT_SET_Z, TMP2, 0, SLJIT_EQUAL);
 
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
 }
@@ -5183,7 +5187,7 @@ label = LABEL();
 OP1(MOVU_UCHAR, CHAR1, 0, SLJIT_MEM1(TMP1), IN_UCHARS(1));
 OP1(MOVU_UCHAR, CHAR2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
 jump = CMP(SLJIT_NOT_EQUAL, CHAR1, 0, CHAR2, 0);
-OP2(SLJIT_SUB | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_IMM, IN_UCHARS(1));
+OP2(SLJIT_SUB | SLJIT_SET_Z, TMP2, 0, TMP2, 0, SLJIT_IMM, IN_UCHARS(1));
 JUMPTO(SLJIT_NOT_ZERO, label);
 
 JUMPHERE(jump);
@@ -5227,7 +5231,7 @@ OP1(SLJIT_MOV_U8, CHAR2, 0, SLJIT_MEM2(LCC_TABLE, CHAR2), 0);
 JUMPHERE(jump);
 #endif
 jump = CMP(SLJIT_NOT_EQUAL, CHAR1, 0, CHAR2, 0);
-OP2(SLJIT_SUB | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_IMM, IN_UCHARS(1));
+OP2(SLJIT_SUB | SLJIT_SET_Z, TMP2, 0, TMP2, 0, SLJIT_IMM, IN_UCHARS(1));
 JUMPTO(SLJIT_NOT_ZERO, label);
 
 JUMPHERE(jump);
@@ -5394,7 +5398,7 @@ do
 #endif
 
       default:
-      SLJIT_ASSERT_STOP();
+      SLJIT_UNREACHABLE();
       break;
       }
     context->ucharptr = 0;
@@ -5568,7 +5572,7 @@ while (*cc != XCL_END)
       break;
 
       default:
-      SLJIT_ASSERT_STOP();
+      SLJIT_UNREACHABLE();
       break;
       }
     cc += 2;
@@ -5592,7 +5596,7 @@ if ((cc[-1] & XCL_HASPROP) == 0)
       OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 3);
       OP1(SLJIT_MOV_U8, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)cc);
       OP2(SLJIT_SHL, TMP2, 0, SLJIT_IMM, 1, TMP2, 0);
-      OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);
+      OP2(SLJIT_AND | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);
       add_jump(compiler, &found, JUMP(SLJIT_NOT_ZERO));
       }
 
@@ -5625,7 +5629,7 @@ else if ((cc[-1] & XCL_MAP) != 0)
     OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 3);
     OP1(SLJIT_MOV_U8, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)cc);
     OP2(SLJIT_SHL, TMP2, 0, SLJIT_IMM, 1, TMP2, 0);
-    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);
+    OP2(SLJIT_AND | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);
     add_jump(compiler, list, JUMP(SLJIT_NOT_ZERO));
 
 #ifdef COMPILE_PCRE8
@@ -5644,6 +5648,15 @@ if (needstype || needsscript)
   if (needschar && !charsaved)
     OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);
 
+#ifdef COMPILE_PCRE32
+  if (!common->utf)
+    {
+    jump = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0x10ffff + 1);
+    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, INVALID_UTF_CHAR);
+    JUMPHERE(jump);
+    }
+#endif
+
   OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
   OP1(SLJIT_MOV_U8, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(ucd_stage1));
   OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);
@@ -5735,14 +5748,14 @@ while (*cc != XCL_END)
 
     if (numberofcmps < 3 && (*cc == XCL_SINGLE || *cc == XCL_RANGE))
       {
-      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(c - charoffset));
-      OP_FLAGS(numberofcmps == 0 ? SLJIT_MOV : SLJIT_OR, TMP2, 0, numberofcmps == 0 ? SLJIT_UNUSED : TMP2, 0, SLJIT_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(c - charoffset));
+      OP_FLAGS(numberofcmps == 0 ? SLJIT_MOV : SLJIT_OR, TMP2, 0, SLJIT_EQUAL);
       numberofcmps++;
       }
     else if (numberofcmps > 0)
       {
-      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(c - charoffset));
-      OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(c - charoffset));
+      OP_FLAGS(SLJIT_OR | SLJIT_SET_Z, TMP2, 0, SLJIT_EQUAL);
       jump = JUMP(SLJIT_NOT_ZERO ^ invertcmp);
       numberofcmps = 0;
       }
@@ -5761,14 +5774,14 @@ while (*cc != XCL_END)
 
     if (numberofcmps < 3 && (*cc == XCL_SINGLE || *cc == XCL_RANGE))
       {
-      OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(c - charoffset));
-      OP_FLAGS(numberofcmps == 0 ? SLJIT_MOV : SLJIT_OR, TMP2, 0, numberofcmps == 0 ? SLJIT_UNUSED : TMP2, 0, SLJIT_LESS_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_LESS_EQUAL, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(c - charoffset));
+      OP_FLAGS(numberofcmps == 0 ? SLJIT_MOV : SLJIT_OR, TMP2, 0, SLJIT_LESS_EQUAL);
       numberofcmps++;
       }
     else if (numberofcmps > 0)
       {
-      OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(c - charoffset));
-      OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_LESS_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_LESS_EQUAL, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(c - charoffset));
+      OP_FLAGS(SLJIT_OR | SLJIT_SET_Z, TMP2, 0, SLJIT_LESS_EQUAL);
       jump = JUMP(SLJIT_NOT_ZERO ^ invertcmp);
       numberofcmps = 0;
       }
@@ -5793,12 +5806,12 @@ while (*cc != XCL_END)
       break;
 
       case PT_LAMP:
-      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_Lu - typeoffset);
-      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
-      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_Ll - typeoffset);
-      OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_EQUAL);
-      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_Lt - typeoffset);
-      OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_Lu - typeoffset);
+      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_Ll - typeoffset);
+      OP_FLAGS(SLJIT_OR, TMP2, 0, SLJIT_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_Lt - typeoffset);
+      OP_FLAGS(SLJIT_OR | SLJIT_SET_Z, TMP2, 0, SLJIT_EQUAL);
       jump = JUMP(SLJIT_NOT_ZERO ^ invertcmp);
       break;
 
@@ -5820,33 +5833,33 @@ while (*cc != XCL_END)
       case PT_SPACE:
       case PT_PXSPACE:
       SET_CHAR_OFFSET(9);
-      OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd - 0x9);
-      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_LESS_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_LESS_EQUAL, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd - 0x9);
+      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_LESS_EQUAL);
 
-      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x85 - 0x9);
-      OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x85 - 0x9);
+      OP_FLAGS(SLJIT_OR, TMP2, 0, SLJIT_EQUAL);
 
-      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x180e - 0x9);
-      OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x180e - 0x9);
+      OP_FLAGS(SLJIT_OR, TMP2, 0, SLJIT_EQUAL);
 
       SET_TYPE_OFFSET(ucp_Zl);
-      OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_Zs - ucp_Zl);
-      OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_LESS_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_LESS_EQUAL, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_Zs - ucp_Zl);
+      OP_FLAGS(SLJIT_OR | SLJIT_SET_Z, TMP2, 0, SLJIT_LESS_EQUAL);
       jump = JUMP(SLJIT_NOT_ZERO ^ invertcmp);
       break;
 
       case PT_WORD:
-      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(CHAR_UNDERSCORE - charoffset));
-      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(CHAR_UNDERSCORE - charoffset));
+      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_EQUAL);
       /* Fall through. */
 
       case PT_ALNUM:
       SET_TYPE_OFFSET(ucp_Ll);
-      OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_Lu - ucp_Ll);
-      OP_FLAGS((*cc == PT_ALNUM) ? SLJIT_MOV : SLJIT_OR, TMP2, 0, (*cc == PT_ALNUM) ? SLJIT_UNUSED : TMP2, 0, SLJIT_LESS_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_LESS_EQUAL, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_Lu - ucp_Ll);
+      OP_FLAGS((*cc == PT_ALNUM) ? SLJIT_MOV : SLJIT_OR, TMP2, 0, SLJIT_LESS_EQUAL);
       SET_TYPE_OFFSET(ucp_Nd);
-      OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_No - ucp_Nd);
-      OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_LESS_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_LESS_EQUAL, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_No - ucp_Nd);
+      OP_FLAGS(SLJIT_OR | SLJIT_SET_Z, TMP2, 0, SLJIT_LESS_EQUAL);
       jump = JUMP(SLJIT_NOT_ZERO ^ invertcmp);
       break;
 
@@ -5868,8 +5881,8 @@ while (*cc != XCL_END)
           OP2(SLJIT_ADD, TMP2, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)charoffset);
           OP2(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_IMM, other_cases[1] ^ other_cases[0]);
           }
-        OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, other_cases[1]);
-        OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
+        OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, other_cases[1]);
+        OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_EQUAL);
         other_cases += 2;
         }
       else if (is_powerof2(other_cases[2] ^ other_cases[1]))
@@ -5881,63 +5894,63 @@ while (*cc != XCL_END)
           OP2(SLJIT_ADD, TMP2, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)charoffset);
           OP2(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_IMM, other_cases[1] ^ other_cases[0]);
           }
-        OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, other_cases[2]);
-        OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
+        OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, other_cases[2]);
+        OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_EQUAL);
 
-        OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(other_cases[0] - charoffset));
-        OP_FLAGS(SLJIT_OR | ((other_cases[3] == NOTACHAR) ? SLJIT_SET_E : 0), TMP2, 0, TMP2, 0, SLJIT_EQUAL);
+        OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(other_cases[0] - charoffset));
+        OP_FLAGS(SLJIT_OR | ((other_cases[3] == NOTACHAR) ? SLJIT_SET_Z : 0), TMP2, 0, SLJIT_EQUAL);
 
         other_cases += 3;
         }
       else
         {
-        OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(*other_cases++ - charoffset));
-        OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
+        OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(*other_cases++ - charoffset));
+        OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_EQUAL);
         }
 
       while (*other_cases != NOTACHAR)
         {
-        OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(*other_cases++ - charoffset));
-        OP_FLAGS(SLJIT_OR | ((*other_cases == NOTACHAR) ? SLJIT_SET_E : 0), TMP2, 0, TMP2, 0, SLJIT_EQUAL);
+        OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(*other_cases++ - charoffset));
+        OP_FLAGS(SLJIT_OR | ((*other_cases == NOTACHAR) ? SLJIT_SET_Z : 0), TMP2, 0, SLJIT_EQUAL);
         }
       jump = JUMP(SLJIT_NOT_ZERO ^ invertcmp);
       break;
 
       case PT_UCNC:
-      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(CHAR_DOLLAR_SIGN - charoffset));
-      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
-      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(CHAR_COMMERCIAL_AT - charoffset));
-      OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_EQUAL);
-      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(CHAR_GRAVE_ACCENT - charoffset));
-      OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(CHAR_DOLLAR_SIGN - charoffset));
+      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(CHAR_COMMERCIAL_AT - charoffset));
+      OP_FLAGS(SLJIT_OR, TMP2, 0, SLJIT_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(CHAR_GRAVE_ACCENT - charoffset));
+      OP_FLAGS(SLJIT_OR, TMP2, 0, SLJIT_EQUAL);
 
       SET_CHAR_OFFSET(0xa0);
-      OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(0xd7ff - charoffset));
-      OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_LESS_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_LESS_EQUAL, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(0xd7ff - charoffset));
+      OP_FLAGS(SLJIT_OR, TMP2, 0, SLJIT_LESS_EQUAL);
       SET_CHAR_OFFSET(0);
-      OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xe000 - 0);
-      OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_GREATER_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_GREATER_EQUAL, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xe000 - 0);
+      OP_FLAGS(SLJIT_OR | SLJIT_SET_Z, TMP2, 0, SLJIT_GREATER_EQUAL);
       jump = JUMP(SLJIT_NOT_ZERO ^ invertcmp);
       break;
 
       case PT_PXGRAPH:
       /* C and Z groups are the farthest two groups. */
       SET_TYPE_OFFSET(ucp_Ll);
-      OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_So - ucp_Ll);
-      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_GREATER);
+      OP2(SLJIT_SUB | SLJIT_SET_GREATER, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_So - ucp_Ll);
+      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_GREATER);
 
       jump = CMP(SLJIT_NOT_EQUAL, typereg, 0, SLJIT_IMM, ucp_Cf - ucp_Ll);
 
       /* In case of ucp_Cf, we overwrite the result. */
       SET_CHAR_OFFSET(0x2066);
-      OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x2069 - 0x2066);
-      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_LESS_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_LESS_EQUAL, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x2069 - 0x2066);
+      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_LESS_EQUAL);
 
-      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x061c - 0x2066);
-      OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x061c - 0x2066);
+      OP_FLAGS(SLJIT_OR, TMP2, 0, SLJIT_EQUAL);
 
-      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x180e - 0x2066);
-      OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x180e - 0x2066);
+      OP_FLAGS(SLJIT_OR, TMP2, 0, SLJIT_EQUAL);
 
       JUMPHERE(jump);
       jump = CMP(SLJIT_ZERO ^ invertcmp, TMP2, 0, SLJIT_IMM, 0);
@@ -5946,21 +5959,21 @@ while (*cc != XCL_END)
       case PT_PXPRINT:
       /* C and Z groups are the farthest two groups. */
       SET_TYPE_OFFSET(ucp_Ll);
-      OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_So - ucp_Ll);
-      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_GREATER);
+      OP2(SLJIT_SUB | SLJIT_SET_GREATER, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_So - ucp_Ll);
+      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_GREATER);
 
-      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_Zs - ucp_Ll);
-      OP_FLAGS(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_NOT_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_Zs - ucp_Ll);
+      OP_FLAGS(SLJIT_AND, TMP2, 0, SLJIT_NOT_EQUAL);
 
       jump = CMP(SLJIT_NOT_EQUAL, typereg, 0, SLJIT_IMM, ucp_Cf - ucp_Ll);
 
       /* In case of ucp_Cf, we overwrite the result. */
       SET_CHAR_OFFSET(0x2066);
-      OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x2069 - 0x2066);
-      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_LESS_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_LESS_EQUAL, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x2069 - 0x2066);
+      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_LESS_EQUAL);
 
-      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x061c - 0x2066);
-      OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x061c - 0x2066);
+      OP_FLAGS(SLJIT_OR, TMP2, 0, SLJIT_EQUAL);
 
       JUMPHERE(jump);
       jump = CMP(SLJIT_ZERO ^ invertcmp, TMP2, 0, SLJIT_IMM, 0);
@@ -5968,21 +5981,21 @@ while (*cc != XCL_END)
 
       case PT_PXPUNCT:
       SET_TYPE_OFFSET(ucp_Sc);
-      OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_So - ucp_Sc);
-      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_LESS_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_LESS_EQUAL, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_So - ucp_Sc);
+      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_LESS_EQUAL);
 
       SET_CHAR_OFFSET(0);
-      OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x7f);
-      OP_FLAGS(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_LESS_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_LESS_EQUAL, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x7f);
+      OP_FLAGS(SLJIT_AND, TMP2, 0, SLJIT_LESS_EQUAL);
 
       SET_TYPE_OFFSET(ucp_Pc);
-      OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_Ps - ucp_Pc);
-      OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_LESS_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_LESS_EQUAL, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_Ps - ucp_Pc);
+      OP_FLAGS(SLJIT_OR | SLJIT_SET_Z, TMP2, 0, SLJIT_LESS_EQUAL);
       jump = JUMP(SLJIT_NOT_ZERO ^ invertcmp);
       break;
 
       default:
-      SLJIT_ASSERT_STOP();
+      SLJIT_UNREACHABLE();
       break;
       }
     cc += 2;
@@ -6028,6 +6041,7 @@ switch(type)
   case OP_NOT_WORD_BOUNDARY:
   case OP_WORD_BOUNDARY:
   add_jump(compiler, &common->wordboundary, JUMP(SLJIT_FAST_CALL));
+  sljit_set_current_flags(compiler, SLJIT_SET_Z);
   add_jump(compiler, backtracks, JUMP(type == OP_NOT_WORD_BOUNDARY ? SLJIT_NOT_ZERO : SLJIT_ZERO));
   return cc;
 
@@ -6043,10 +6057,10 @@ switch(type)
     else
       {
       jump[1] = CMP(SLJIT_EQUAL, TMP2, 0, STR_END, 0);
-      OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP2, 0, STR_END, 0);
-      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_LESS);
-      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff);
-      OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_NOT_EQUAL);
+      OP2(SLJIT_SUB | SLJIT_SET_LESS, SLJIT_UNUSED, 0, TMP2, 0, STR_END, 0);
+      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_LESS);
+      OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff);
+      OP_FLAGS(SLJIT_OR | SLJIT_SET_Z, TMP2, 0, SLJIT_NOT_EQUAL);
       add_jump(compiler, backtracks, JUMP(SLJIT_NOT_EQUAL));
       check_partial(common, TRUE);
       add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));
@@ -6068,9 +6082,9 @@ switch(type)
     OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
     jump[1] = CMP(SLJIT_NOT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR);
     OP2(SLJIT_ADD, TMP2, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
-    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP2, 0, STR_END, 0);
+    OP2(SLJIT_SUB | SLJIT_SET_Z | SLJIT_SET_GREATER, SLJIT_UNUSED, 0, TMP2, 0, STR_END, 0);
     jump[2] = JUMP(SLJIT_GREATER);
-    add_jump(compiler, backtracks, JUMP(SLJIT_LESS));
+    add_jump(compiler, backtracks, JUMP(SLJIT_NOT_EQUAL) /* LESS */);
     /* Equal. */
     OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
     jump[3] = CMP(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL);
@@ -6089,6 +6103,7 @@ switch(type)
       read_char_range(common, common->nlmin, common->nlmax, TRUE);
       add_jump(compiler, backtracks, CMP(SLJIT_NOT_EQUAL, STR_PTR, 0, STR_END, 0));
       add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));
+      sljit_set_current_flags(compiler, SLJIT_SET_Z);
       add_jump(compiler, backtracks, JUMP(SLJIT_ZERO));
       OP1(SLJIT_MOV, STR_PTR, 0, SLJIT_MEM1(SLJIT_SP), LOCALS1);
       }
@@ -6204,7 +6219,7 @@ switch(type)
     label = LABEL();
     add_jump(compiler, backtracks, CMP(SLJIT_LESS_EQUAL, STR_PTR, 0, TMP3, 0));
     skip_char_back(common);
-    OP2(SLJIT_SUB | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_IMM, 1);
+    OP2(SLJIT_SUB | SLJIT_SET_Z, TMP2, 0, TMP2, 0, SLJIT_IMM, 1);
     JUMPTO(SLJIT_NOT_ZERO, label);
     }
   else
@@ -6217,7 +6232,7 @@ switch(type)
   check_start_used_ptr(common);
   return cc + LINK_SIZE;
   }
-SLJIT_ASSERT_STOP();
+SLJIT_UNREACHABLE();
 return cc;
 }
 
@@ -6250,7 +6265,7 @@ switch(type)
 #endif
     read_char8_type(common, type == OP_NOT_DIGIT);
     /* Flip the starting bit in the negative case. */
-  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ctype_digit);
+  OP2(SLJIT_AND | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ctype_digit);
   add_jump(compiler, backtracks, JUMP(type == OP_DIGIT ? SLJIT_ZERO : SLJIT_NOT_ZERO));
   return cc;
 
@@ -6264,7 +6279,7 @@ switch(type)
   else
 #endif
     read_char8_type(common, type == OP_NOT_WHITESPACE);
-  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ctype_space);
+  OP2(SLJIT_AND | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ctype_space);
   add_jump(compiler, backtracks, JUMP(type == OP_WHITESPACE ? SLJIT_ZERO : SLJIT_NOT_ZERO));
   return cc;
 
@@ -6278,7 +6293,7 @@ switch(type)
   else
 #endif
     read_char8_type(common, type == OP_NOT_WORDCHAR);
-  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ctype_word);
+  OP2(SLJIT_AND | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ctype_word);
   add_jump(compiler, backtracks, JUMP(type == OP_WORDCHAR ? SLJIT_ZERO : SLJIT_NOT_ZERO));
   return cc;
 
@@ -6320,8 +6335,8 @@ switch(type)
 #elif defined COMPILE_PCRE16
     jump[0] = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xd800);
     OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
-    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
-    OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
+    OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
+    OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_EQUAL);
     OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
     OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
 #endif
@@ -6383,6 +6398,7 @@ switch(type)
     detect_partial_match(common, backtracks);
   read_char_range(common, 0x9, 0x3000, type == OP_NOT_HSPACE);
   add_jump(compiler, &common->hspace, JUMP(SLJIT_FAST_CALL));
+  sljit_set_current_flags(compiler, SLJIT_SET_Z);
   add_jump(compiler, backtracks, JUMP(type == OP_NOT_HSPACE ? SLJIT_NOT_ZERO : SLJIT_ZERO));
   return cc;
 
@@ -6392,6 +6408,7 @@ switch(type)
     detect_partial_match(common, backtracks);
   read_char_range(common, 0xa, 0x2029, type == OP_NOT_VSPACE);
   add_jump(compiler, &common->vspace, JUMP(SLJIT_FAST_CALL));
+  sljit_set_current_flags(compiler, SLJIT_SET_Z);
   add_jump(compiler, backtracks, JUMP(type == OP_NOT_VSPACE ? SLJIT_NOT_ZERO : SLJIT_ZERO));
   return cc;
 
@@ -6418,7 +6435,7 @@ switch(type)
   OP1(SLJIT_MOV_U32, TMP1, 0, SLJIT_MEM1(STACK_TOP), (sljit_sw)PRIV(ucp_gbtable));
   OP1(SLJIT_MOV, STACK_TOP, 0, TMP2, 0);
   OP2(SLJIT_SHL, TMP2, 0, SLJIT_IMM, 1, TMP2, 0);
-  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);
+  OP2(SLJIT_AND | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);
   JUMPTO(SLJIT_NOT_ZERO, label);
 
   OP1(SLJIT_MOV, STR_PTR, 0, TMP3, 0);
@@ -6587,7 +6604,7 @@ switch(type)
   OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 3);
   OP1(SLJIT_MOV_U8, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)cc);
   OP2(SLJIT_SHL, TMP2, 0, SLJIT_IMM, 1, TMP2, 0);
-  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);
+  OP2(SLJIT_AND | SLJIT_SET_Z, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);
   add_jump(compiler, backtracks, JUMP(SLJIT_ZERO));
 
 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
@@ -6604,7 +6621,7 @@ switch(type)
   return cc + GET(cc, 0) - 1;
 #endif
   }
-SLJIT_ASSERT_STOP();
+SLJIT_UNREACHABLE();
 return cc;
 }
 
@@ -6790,9 +6807,9 @@ else
 #endif /* SUPPORT_UTF && SUPPORT_UCP */
   {
   if (ref)
-    OP2(SLJIT_SUB | SLJIT_SET_E, TMP2, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(offset + 1), TMP1, 0);
+    OP2(SLJIT_SUB | SLJIT_SET_Z, TMP2, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(offset + 1), TMP1, 0);
   else
-    OP2(SLJIT_SUB | SLJIT_SET_E, TMP2, 0, SLJIT_MEM1(TMP2), sizeof(sljit_sw), TMP1, 0);
+    OP2(SLJIT_SUB | SLJIT_SET_Z, TMP2, 0, SLJIT_MEM1(TMP2), sizeof(sljit_sw), TMP1, 0);
 
   if (withchecks)
     jump = JUMP(SLJIT_ZERO);
@@ -6883,7 +6900,7 @@ switch(type)
   cc += 1 + IMM2_SIZE + 1 + 2 * IMM2_SIZE;
   break;
   default:
-  SLJIT_ASSERT_STOP();
+  SLJIT_UNREACHABLE();
   break;
   }
 
@@ -6897,7 +6914,7 @@ if (!minimize)
     OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), STACK(0), STR_PTR, 0);
     OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), STACK(1), SLJIT_IMM, 0);
     /* Temporary release of STR_PTR. */
-    OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, sizeof(sljit_sw));
+    OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, sizeof(sljit_sw));
     /* Handles both invalid and empty cases. Since the minimum repeat,
     is zero the invalid case is basically the same as an empty case. */
     if (ref)
@@ -6910,7 +6927,7 @@ if (!minimize)
       zerolength = CMP(SLJIT_EQUAL, TMP1, 0, SLJIT_MEM1(TMP2), sizeof(sljit_sw));
       }
     /* Restore if not zero length. */
-    OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, sizeof(sljit_sw));
+    OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, sizeof(sljit_sw));
     }
   else
     {
@@ -7157,7 +7174,7 @@ return (*PUBL(callout))(callout_block);
     (((int)sizeof(PUBL(callout_block)) + 7) & ~7)
 
 #define CALLOUT_ARG_OFFSET(arg) \
-    (-CALLOUT_ARG_SIZE + SLJIT_OFFSETOF(PUBL(callout_block), arg))
+    SLJIT_OFFSETOF(PUBL(callout_block), arg)
 
 static SLJIT_INLINE pcre_uchar *compile_callout_matchingpath(compiler_common *common, pcre_uchar *cc, backtrack_common *parent)
 {
@@ -7187,7 +7204,8 @@ OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), CALLOUT_ARG_OFFSET(mark), (common->mark_pt
 
 /* Needed to save important temporary registers. */
 OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), LOCALS0, STACK_TOP, 0);
-OP2(SLJIT_SUB, SLJIT_R1, 0, STACK_TOP, 0, SLJIT_IMM, CALLOUT_ARG_SIZE);
+/* SLJIT_R0 = arguments */
+OP1(SLJIT_MOV, SLJIT_R1, 0, STACK_TOP, 0);
 GET_LOCAL_BASE(SLJIT_R2, 0, OVECTOR_START);
 sljit_emit_ijump(compiler, SLJIT_CALL3, SLJIT_IMM, SLJIT_FUNC_OFFSET(do_callout));
 OP1(SLJIT_MOV_S32, SLJIT_RETURN_REG, 0, SLJIT_RETURN_REG, 0);
@@ -7195,12 +7213,12 @@ OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(SLJIT_SP), LOCALS0);
 free_stack(common, CALLOUT_ARG_SIZE / sizeof(sljit_sw));
 
 /* Check return value. */
-OP2(SLJIT_SUB | SLJIT_SET_S, SLJIT_UNUSED, 0, SLJIT_RETURN_REG, 0, SLJIT_IMM, 0);
+OP2(SLJIT_SUB | SLJIT_SET_Z | SLJIT_SET_SIG_GREATER, SLJIT_UNUSED, 0, SLJIT_RETURN_REG, 0, SLJIT_IMM, 0);
 add_jump(compiler, &backtrack->topbacktracks, JUMP(SLJIT_SIG_GREATER));
 if (common->forced_quit_label == NULL)
-  add_jump(compiler, &common->forced_quit, JUMP(SLJIT_SIG_LESS));
+  add_jump(compiler, &common->forced_quit, JUMP(SLJIT_NOT_EQUAL) /* SIG_LESS */);
 else
-  JUMPTO(SLJIT_SIG_LESS, common->forced_quit_label);
+  JUMPTO(SLJIT_NOT_EQUAL /* SIG_LESS */, common->forced_quit_label);
 return cc + 2 + 2 * LINK_SIZE;
 }
 
@@ -7321,7 +7339,7 @@ else
   allocate_stack(common, framesize + extrasize);
 
   OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr);
-  OP2(SLJIT_SUB, TMP2, 0, STACK_TOP, 0, SLJIT_IMM, (framesize + extrasize) * sizeof(sljit_sw));
+  OP2(SLJIT_ADD, TMP2, 0, STACK_TOP, 0, SLJIT_IMM, (framesize + extrasize) * sizeof(sljit_sw));
   OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), private_data_ptr, TMP2, 0);
   if (needs_control_head)
     OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_SP), common->control_head_ptr);
@@ -7392,22 +7410,22 @@ while (1)
       free_stack(common, extrasize);
 
     if (needs_control_head)
-      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->control_head_ptr, SLJIT_MEM1(STACK_TOP), 0);
+      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->control_head_ptr, SLJIT_MEM1(STACK_TOP), STACK(-1));
     }
   else
     {
     if ((opcode != OP_ASSERT_NOT && opcode != OP_ASSERTBACK_NOT) || conditional)
       {
       /* We don't need to keep the STR_PTR, only the previous private_data_ptr. */
-      OP2(SLJIT_ADD, STACK_TOP, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr, SLJIT_IMM, (framesize + 1) * sizeof(sljit_sw));
+      OP2(SLJIT_SUB, STACK_TOP, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr, SLJIT_IMM, (framesize + 1) * sizeof(sljit_sw));
       if (needs_control_head)
-        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->control_head_ptr, SLJIT_MEM1(STACK_TOP), 0);
+        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->control_head_ptr, SLJIT_MEM1(STACK_TOP), STACK(-1));
       }
     else
       {
       OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr);
       if (needs_control_head)
-        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->control_head_ptr, SLJIT_MEM1(STACK_TOP), (framesize + 1) * sizeof(sljit_sw));
+        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->control_head_ptr, SLJIT_MEM1(STACK_TOP), STACK(-framesize - 2));
       add_jump(compiler, &common->revertframes, JUMP(SLJIT_FAST_CALL));
       }
     }
@@ -7418,25 +7436,25 @@ while (1)
     if (conditional)
       {
       if (extrasize > 0)
-        OP1(SLJIT_MOV, STR_PTR, 0, SLJIT_MEM1(STACK_TOP), needs_control_head ? sizeof(sljit_sw) : 0);
+        OP1(SLJIT_MOV, STR_PTR, 0, SLJIT_MEM1(STACK_TOP), needs_control_head ? STACK(-2) : STACK(-1));
       }
     else if (bra == OP_BRAZERO)
       {
       if (framesize < 0)
-        OP1(SLJIT_MOV, STR_PTR, 0, SLJIT_MEM1(STACK_TOP), (extrasize - 1) * sizeof(sljit_sw));
+        OP1(SLJIT_MOV, STR_PTR, 0, SLJIT_MEM1(STACK_TOP), STACK(-extrasize));
       else
         {
-        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), framesize * sizeof(sljit_sw));
-        OP1(SLJIT_MOV, STR_PTR, 0, SLJIT_MEM1(STACK_TOP), (framesize + extrasize - 1) * sizeof(sljit_sw));
+        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), STACK(-framesize - 1));
+        OP1(SLJIT_MOV, STR_PTR, 0, SLJIT_MEM1(STACK_TOP), STACK(-framesize - extrasize));
         OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), private_data_ptr, TMP1, 0);
         }
-      OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, sizeof(sljit_sw));
+      OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, sizeof(sljit_sw));
       OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), STACK(0), SLJIT_IMM, 0);
       }
     else if (framesize >= 0)
       {
       /* For OP_BRA and OP_BRAMINZERO. */
-      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), private_data_ptr, SLJIT_MEM1(STACK_TOP), framesize * sizeof(sljit_sw));
+      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), private_data_ptr, SLJIT_MEM1(STACK_TOP), STACK(-framesize - 1));
       }
     }
   add_jump(compiler, found, JUMP(SLJIT_JUMP));
@@ -7480,12 +7498,12 @@ if (common->positive_assert_quit != NULL)
   set_jumps(common->positive_assert_quit, LABEL());
   SLJIT_ASSERT(framesize != no_stack);
   if (framesize < 0)
-    OP2(SLJIT_ADD, STACK_TOP, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr, SLJIT_IMM, extrasize * sizeof(sljit_sw));
+    OP2(SLJIT_SUB, STACK_TOP, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr, SLJIT_IMM, extrasize * sizeof(sljit_sw));
   else
     {
     OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr);
     add_jump(compiler, &common->revertframes, JUMP(SLJIT_FAST_CALL));
-    OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, (framesize + extrasize) * sizeof(sljit_sw));
+    OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, (framesize + extrasize) * sizeof(sljit_sw));
     }
   JUMPHERE(jump);
   }
@@ -7534,18 +7552,18 @@ if (opcode == OP_ASSERT || opcode == OP_ASSERTBACK)
     {
     /* We know that STR_PTR was stored on the top of the stack. */
     if (extrasize > 0)
-      OP1(SLJIT_MOV, STR_PTR, 0, SLJIT_MEM1(STACK_TOP), (extrasize - 1) * sizeof(sljit_sw));
+      OP1(SLJIT_MOV, STR_PTR, 0, SLJIT_MEM1(STACK_TOP), STACK(-extrasize));
 
     /* Keep the STR_PTR on the top of the stack. */
     if (bra == OP_BRAZERO)
       {
-      OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, sizeof(sljit_sw));
+      OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, sizeof(sljit_sw));
       if (extrasize == 2)
         OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), STACK(0), STR_PTR, 0);
       }
     else if (bra == OP_BRAMINZERO)
       {
-      OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, sizeof(sljit_sw));
+      OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, sizeof(sljit_sw));
       OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), STACK(0), SLJIT_IMM, 0);
       }
     }
@@ -7554,13 +7572,13 @@ if (opcode == OP_ASSERT || opcode == OP_ASSERTBACK)
     if (bra == OP_BRA)
       {
       /* We don't need to keep the STR_PTR, only the previous private_data_ptr. */
-      OP2(SLJIT_ADD, STACK_TOP, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr, SLJIT_IMM, (framesize + 1) * sizeof(sljit_sw));
-      OP1(SLJIT_MOV, STR_PTR, 0, SLJIT_MEM1(STACK_TOP), (extrasize - 2) * sizeof(sljit_sw));
+      OP2(SLJIT_SUB, STACK_TOP, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr, SLJIT_IMM, (framesize + 1) * sizeof(sljit_sw));
+      OP1(SLJIT_MOV, STR_PTR, 0, SLJIT_MEM1(STACK_TOP), STACK(-extrasize + 1));
       }
     else
       {
       /* We don't need to keep the STR_PTR, only the previous private_data_ptr. */
-      OP2(SLJIT_ADD, STACK_TOP, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr, SLJIT_IMM, (framesize + 2) * sizeof(sljit_sw));
+      OP2(SLJIT_SUB, STACK_TOP, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr, SLJIT_IMM, (framesize + 2) * sizeof(sljit_sw));
       if (extrasize == 2)
         {
         OP1(SLJIT_MOV, STR_PTR, 0, SLJIT_MEM1(STACK_TOP), STACK(0));
@@ -7588,7 +7606,7 @@ if (opcode == OP_ASSERT || opcode == OP_ASSERTBACK)
       {
       OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr);
       add_jump(compiler, &common->revertframes, JUMP(SLJIT_FAST_CALL));
-      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), private_data_ptr, SLJIT_MEM1(STACK_TOP), framesize * sizeof(sljit_sw));
+      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), private_data_ptr, SLJIT_MEM1(STACK_TOP), STACK(-framesize - 1));
       }
     set_jumps(backtrack->common.topbacktracks, LABEL());
     }
@@ -7675,23 +7693,23 @@ if (framesize < 0)
     }
 
   if (needs_control_head)
-    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), (ket != OP_KET || has_alternatives) ? sizeof(sljit_sw) : 0);
+    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), (ket != OP_KET || has_alternatives) ? STACK(-2) : STACK(-1));
 
   /* TMP2 which is set here used by OP_KETRMAX below. */
   if (ket == OP_KETRMAX)
-    OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), 0);
+    OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), STACK(-1));
   else if (ket == OP_KETRMIN)
     {
     /* Move the STR_PTR to the private_data_ptr. */
-    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), private_data_ptr, SLJIT_MEM1(STACK_TOP), 0);
+    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), private_data_ptr, SLJIT_MEM1(STACK_TOP), STACK(-1));
     }
   }
 else
   {
   stacksize = (ket != OP_KET || has_alternatives) ? 2 : 1;
-  OP2(SLJIT_ADD, STACK_TOP, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr, SLJIT_IMM, (framesize + stacksize) * sizeof(sljit_sw));
+  OP2(SLJIT_SUB, STACK_TOP, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr, SLJIT_IMM, (framesize + stacksize) * sizeof(sljit_sw));
   if (needs_control_head)
-    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), 0);
+    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), STACK(-1));
 
   if (ket == OP_KETRMAX)
     {
@@ -7927,7 +7945,7 @@ if (bra == OP_BRAMINZERO)
         {
         /* Except when the whole stack frame must be saved. */
         OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr);
-        braminzero = CMP(SLJIT_EQUAL, STR_PTR, 0, SLJIT_MEM1(TMP1), (BACKTRACK_AS(bracket_backtrack)->u.framesize + 1) * sizeof(sljit_sw));
+        braminzero = CMP(SLJIT_EQUAL, STR_PTR, 0, SLJIT_MEM1(TMP1), STACK(-BACKTRACK_AS(bracket_backtrack)->u.framesize - 2));
         }
       JUMPHERE(skip);
       }
@@ -8000,7 +8018,7 @@ if (opcode == OP_ONCE)
         OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr);
       OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), STACK(stacksize), STR_PTR, 0);
       if (BACKTRACK_AS(bracket_backtrack)->u.framesize == no_frame)
-        OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_SP), private_data_ptr, STACK_TOP, 0, SLJIT_IMM, needs_control_head ? (2 * sizeof(sljit_sw)) : sizeof(sljit_sw));
+        OP2(SLJIT_ADD, SLJIT_MEM1(SLJIT_SP), private_data_ptr, STACK_TOP, 0, SLJIT_IMM, needs_control_head ? (2 * sizeof(sljit_sw)) : sizeof(sljit_sw));
       OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), STACK(stacksize + 1), TMP2, 0);
       }
     else if (ket == OP_KETRMAX || has_alternatives)
@@ -8018,7 +8036,7 @@ if (opcode == OP_ONCE)
       OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), STACK(0), TMP2, 0);
 
     OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr);
-    OP2(SLJIT_SUB, TMP2, 0, STACK_TOP, 0, SLJIT_IMM, stacksize * sizeof(sljit_sw));
+    OP2(SLJIT_ADD, TMP2, 0, STACK_TOP, 0, SLJIT_IMM, stacksize * sizeof(sljit_sw));
 
     stacksize = needs_control_head ? 1 : 0;
     if (ket != OP_KET || has_alternatives)
@@ -8090,13 +8108,13 @@ if (opcode == OP_COND || opcode == OP_SCOND)
     slot = common->name_table + GET2(matchingpath, 1) * common->name_entry_size;
     OP1(SLJIT_MOV, TMP3, 0, STR_PTR, 0);
     OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(1));
-    OP2(SLJIT_SUB | SLJIT_SET_E, TMP2, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(GET2(slot, 0) << 1), TMP1, 0);
+    OP2(SLJIT_SUB | SLJIT_SET_Z, TMP2, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(GET2(slot, 0) << 1), TMP1, 0);
     slot += common->name_entry_size;
     i--;
     while (i-- > 0)
       {
       OP2(SLJIT_SUB, STR_PTR, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(GET2(slot, 0) << 1), TMP1, 0);
-      OP2(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, STR_PTR, 0);
+      OP2(SLJIT_OR | SLJIT_SET_Z, TMP2, 0, TMP2, 0, STR_PTR, 0);
       slot += common->name_entry_size;
       }
     OP1(SLJIT_MOV, STR_PTR, 0, TMP3, 0);
@@ -8111,7 +8129,7 @@ if (opcode == OP_COND || opcode == OP_SCOND)
 
     if (*matchingpath == OP_FAIL)
       stacksize = 0;
-    if (*matchingpath == OP_RREF)
+    else if (*matchingpath == OP_RREF)
       {
       stacksize = GET2(matchingpath, 1);
       if (common->currententry == NULL)
@@ -8244,7 +8262,7 @@ if (ket == OP_KETRMAX)
     {
     if (has_alternatives)
       BACKTRACK_AS(bracket_backtrack)->alternative_matchingpath = LABEL();
-    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_MEM1(SLJIT_SP), repeat_ptr, SLJIT_MEM1(SLJIT_SP), repeat_ptr, SLJIT_IMM, 1);
+    OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_MEM1(SLJIT_SP), repeat_ptr, SLJIT_MEM1(SLJIT_SP), repeat_ptr, SLJIT_IMM, 1);
     JUMPTO(SLJIT_NOT_ZERO, rmax_label);
     /* Drop STR_PTR for greedy plus quantifier. */
     if (opcode != OP_ONCE)
@@ -8274,7 +8292,7 @@ if (ket == OP_KETRMAX)
 if (repeat_type == OP_EXACT)
   {
   count_match(common);
-  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_MEM1(SLJIT_SP), repeat_ptr, SLJIT_MEM1(SLJIT_SP), repeat_ptr, SLJIT_IMM, 1);
+  OP2(SLJIT_SUB | SLJIT_SET_Z, SLJIT_MEM1(SLJIT_SP), repeat_ptr, SLJIT_MEM1(SLJIT_SP), repeat_ptr, SLJIT_IMM, 1);
   JUMPTO(SLJIT_NOT_ZERO, rmax_label);
   }
 else if (repeat_type == OP_UPTO)
@@ -8374,7 +8392,7 @@ switch(opcode)
   break;
 
   default:
-  SLJIT_ASSERT_STOP();
+  SLJIT_UNREACHABLE();
   break;
   }
 
@@ -8452,7 +8470,7 @@ else
   OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr);
   if (needs_control_head)
     OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_SP), common->control_head_ptr);
-  OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_SP), private_data_ptr, STACK_TOP, 0, SLJIT_IMM, -STACK(stacksize - 1));
+  OP2(SLJIT_ADD, SLJIT_MEM1(SLJIT_SP), private_data_ptr, STACK_TOP, 0, SLJIT_IMM, stacksize * sizeof(sljit_sw));
 
   stack = 0;
   if (!zero)
@@ -8524,7 +8542,7 @@ while (*cc != OP_KETRPOS)
     {
     if (offset != 0)
       {
-      OP2(SLJIT_ADD, STACK_TOP, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr, SLJIT_IMM, stacksize * sizeof(sljit_sw));
+      OP2(SLJIT_SUB, STACK_TOP, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr, SLJIT_IMM, stacksize * sizeof(sljit_sw));
       OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), cbraprivptr);
       OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), OVECTOR(offset + 1), STR_PTR, 0);
       OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), cbraprivptr, STR_PTR, 0);
@@ -8535,10 +8553,10 @@ while (*cc != OP_KETRPOS)
     else
       {
       OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr);
-      OP2(SLJIT_ADD, STACK_TOP, 0, TMP2, 0, SLJIT_IMM, stacksize * sizeof(sljit_sw));
+      OP2(SLJIT_SUB, STACK_TOP, 0, TMP2, 0, SLJIT_IMM, stacksize * sizeof(sljit_sw));
       if (opcode == OP_SBRAPOS)
-        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(TMP2), (framesize + 1) * sizeof(sljit_sw));
-      OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), (framesize + 1) * sizeof(sljit_sw), STR_PTR, 0);
+        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(TMP2), STACK(-framesize - 2));
+      OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), STACK(-framesize - 2), STR_PTR, 0);
       }
 
     /* Even if the match is empty, we need to reset the control head. */
@@ -8584,7 +8602,7 @@ while (*cc != OP_KETRPOS)
     else
       {
       OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr);
-      OP1(SLJIT_MOV, STR_PTR, 0, SLJIT_MEM1(TMP2), (framesize + 1) * sizeof(sljit_sw));
+      OP1(SLJIT_MOV, STR_PTR, 0, SLJIT_MEM1(TMP2), STACK(-framesize - 2));
       }
     }
 
@@ -8601,7 +8619,7 @@ if (!zero)
   if (framesize < 0)
     add_jump(compiler, &backtrack->topbacktracks, CMP(SLJIT_NOT_EQUAL, SLJIT_MEM1(STACK_TOP), STACK(stacksize - 1), SLJIT_IMM, 0));
   else /* TMP2 is set to [private_data_ptr] above. */
-    add_jump(compiler, &backtrack->topbacktracks, CMP(SLJIT_NOT_EQUAL, SLJIT_MEM1(TMP2), (stacksize - 1) * sizeof(sljit_sw), SLJIT_IMM, 0));
+    add_jump(compiler, &backtrack->topbacktracks, CMP(SLJIT_NOT_EQUAL, SLJIT_MEM1(TMP2), STACK(-stacksize), SLJIT_IMM, 0));
   }
 
 /* None of them matched. */
@@ -8824,7 +8842,7 @@ if (exact > 1)
     OP1(SLJIT_MOV, tmp_base, tmp_offset, SLJIT_IMM, exact);
     label = LABEL();
     compile_char1_matchingpath(common, type, cc, &backtrack->topbacktracks, FALSE);
-    OP2(SLJIT_SUB | SLJIT_SET_E, tmp_base, tmp_offset, tmp_base, tmp_offset, SLJIT_IMM, 1);
+    OP2(SLJIT_SUB | SLJIT_SET_Z, tmp_base, tmp_offset, tmp_base, tmp_offset, SLJIT_IMM, 1);
     JUMPTO(SLJIT_NOT_ZERO, label);
     }
   else
@@ -8832,7 +8850,7 @@ if (exact > 1)
     OP1(SLJIT_MOV, tmp_base, tmp_offset, SLJIT_IMM, exact);
     label = LABEL();
     compile_char1_matchingpath(common, type, cc, &backtrack->topbacktracks, TRUE);
-    OP2(SLJIT_SUB | SLJIT_SET_E, tmp_base, tmp_offset, tmp_base, tmp_offset, SLJIT_IMM, 1);
+    OP2(SLJIT_SUB | SLJIT_SET_Z, tmp_base, tmp_offset, tmp_base, tmp_offset, SLJIT_IMM, 1);
     JUMPTO(SLJIT_NOT_ZERO, label);
     }
   }
@@ -8862,7 +8880,7 @@ switch(opcode)
     if (opcode == OP_UPTO)
       {
       OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), POSSESSIVE0);
-      OP2(SLJIT_SUB | SLJIT_SET_E, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
+      OP2(SLJIT_SUB | SLJIT_SET_Z, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
       jump = JUMP(SLJIT_ZERO);
       OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), POSSESSIVE0, TMP1, 0);
       }
@@ -8924,7 +8942,7 @@ switch(opcode)
       label = LABEL();
       if (opcode == OP_UPTO)
         {
-        OP2(SLJIT_SUB | SLJIT_SET_E, tmp_base, tmp_offset, tmp_base, tmp_offset, SLJIT_IMM, 1);
+        OP2(SLJIT_SUB | SLJIT_SET_Z, tmp_base, tmp_offset, tmp_base, tmp_offset, SLJIT_IMM, 1);
         add_jump(compiler, &backtrack->topbacktracks, JUMP(SLJIT_ZERO));
         }
       compile_char1_matchingpath(common, type, cc, &backtrack->topbacktracks, FALSE);
@@ -8944,7 +8962,7 @@ switch(opcode)
       OP1(SLJIT_MOV, base, offset1, STR_PTR, 0);
       if (opcode == OP_UPTO)
         {
-        OP2(SLJIT_SUB | SLJIT_SET_E, tmp_base, tmp_offset, tmp_base, tmp_offset, SLJIT_IMM, 1);
+        OP2(SLJIT_SUB | SLJIT_SET_Z, tmp_base, tmp_offset, tmp_base, tmp_offset, SLJIT_IMM, 1);
         add_jump(compiler, &no_match, JUMP(SLJIT_ZERO));
         }
 
@@ -8971,7 +8989,7 @@ switch(opcode)
 
       if (opcode == OP_UPTO)
         {
-        OP2(SLJIT_SUB | SLJIT_SET_E, tmp_base, tmp_offset, tmp_base, tmp_offset, SLJIT_IMM, 1);
+        OP2(SLJIT_SUB | SLJIT_SET_Z, tmp_base, tmp_offset, tmp_base, tmp_offset, SLJIT_IMM, 1);
         JUMPTO(SLJIT_NOT_ZERO, label);
         }
       else
@@ -9000,7 +9018,7 @@ switch(opcode)
 
       if (opcode == OP_UPTO)
         {
-        OP2(SLJIT_SUB | SLJIT_SET_E, tmp_base, tmp_offset, tmp_base, tmp_offset, SLJIT_IMM, 1);
+        OP2(SLJIT_SUB | SLJIT_SET_Z, tmp_base, tmp_offset, tmp_base, tmp_offset, SLJIT_IMM, 1);
         JUMPTO(SLJIT_NOT_ZERO, label);
         }
       else
@@ -9026,7 +9044,7 @@ switch(opcode)
       compile_char1_matchingpath(common, type, cc, &no_char1_match, FALSE);
       if (opcode == OP_UPTO)
         {
-        OP2(SLJIT_SUB | SLJIT_SET_E, tmp_base, tmp_offset, tmp_base, tmp_offset, SLJIT_IMM, 1);
+        OP2(SLJIT_SUB | SLJIT_SET_Z, tmp_base, tmp_offset, tmp_base, tmp_offset, SLJIT_IMM, 1);
         JUMPTO(SLJIT_NOT_ZERO, label);
         OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
         }
@@ -9113,7 +9131,7 @@ switch(opcode)
     label = LABEL();
     compile_char1_matchingpath(common, type, cc, &no_match, TRUE);
     OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), POSSESSIVE1, STR_PTR, 0);
-    OP2(SLJIT_SUB | SLJIT_SET_E, tmp_base, tmp_offset, tmp_base, tmp_offset, SLJIT_IMM, 1);
+    OP2(SLJIT_SUB | SLJIT_SET_Z, tmp_base, tmp_offset, tmp_base, tmp_offset, SLJIT_IMM, 1);
     JUMPTO(SLJIT_NOT_ZERO, label);
     set_jumps(no_match, LABEL());
     OP1(SLJIT_MOV, STR_PTR, 0, SLJIT_MEM1(SLJIT_SP), POSSESSIVE1);
@@ -9124,7 +9142,7 @@ switch(opcode)
   label = LABEL();
   detect_partial_match(common, &no_match);
   compile_char1_matchingpath(common, type, cc, &no_char1_match, FALSE);
-  OP2(SLJIT_SUB | SLJIT_SET_E, tmp_base, tmp_offset, tmp_base, tmp_offset, SLJIT_IMM, 1);
+  OP2(SLJIT_SUB | SLJIT_SET_Z, tmp_base, tmp_offset, tmp_base, tmp_offset, SLJIT_IMM, 1);
   JUMPTO(SLJIT_NOT_ZERO, label);
   OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
   set_jumps(no_char1_match, LABEL());
@@ -9142,7 +9160,7 @@ switch(opcode)
   break;
 
   default:
-  SLJIT_ASSERT_STOP();
+  SLJIT_UNREACHABLE();
   break;
   }
 
@@ -9264,7 +9282,7 @@ size = 3 + (size < 0 ? 0 : size);
 OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_SP), common->control_head_ptr);
 allocate_stack(common, size);
 if (size > 3)
-  OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_SP), common->control_head_ptr, STACK_TOP, 0, SLJIT_IMM, (size - 3) * sizeof(sljit_sw));
+  OP2(SLJIT_ADD, SLJIT_MEM1(SLJIT_SP), common->control_head_ptr, STACK_TOP, 0, SLJIT_IMM, (size - 3) * sizeof(sljit_sw));
 else
   OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->control_head_ptr, STACK_TOP, 0);
 OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), STACK(size - 1), SLJIT_IMM, BACKTRACK_AS(then_trap_backtrack)->start);
@@ -9569,7 +9587,7 @@ while (cc < ccend)
     break;
 
     default:
-    SLJIT_ASSERT_STOP();
+    SLJIT_UNREACHABLE();
     return;
     }
   if (cc == NULL)
@@ -9677,7 +9695,7 @@ switch(opcode)
   case OP_MINUPTO:
   OP1(SLJIT_MOV, TMP1, 0, base, offset1);
   OP1(SLJIT_MOV, STR_PTR, 0, base, offset0);
-  OP2(SLJIT_SUB | SLJIT_SET_E, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
+  OP2(SLJIT_SUB | SLJIT_SET_Z, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
   add_jump(compiler, &jumplist, JUMP(SLJIT_ZERO));
 
   OP1(SLJIT_MOV, base, offset1, TMP1, 0);
@@ -9723,7 +9741,7 @@ switch(opcode)
   break;
 
   default:
-  SLJIT_ASSERT_STOP();
+  SLJIT_UNREACHABLE();
   break;
   }
 
@@ -9831,7 +9849,7 @@ if (*cc == OP_ASSERT || *cc == OP_ASSERTBACK)
   {
   OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(SLJIT_SP), CURRENT_AS(assert_backtrack)->private_data_ptr);
   add_jump(compiler, &common->revertframes, JUMP(SLJIT_FAST_CALL));
-  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), CURRENT_AS(assert_backtrack)->private_data_ptr, SLJIT_MEM1(STACK_TOP), CURRENT_AS(assert_backtrack)->framesize * sizeof(sljit_sw));
+  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), CURRENT_AS(assert_backtrack)->private_data_ptr, SLJIT_MEM1(STACK_TOP), STACK(-CURRENT_AS(assert_backtrack)->framesize - 1));
 
   set_jumps(current->topbacktracks, LABEL());
   }
@@ -9841,7 +9859,7 @@ else
 if (bra == OP_BRAZERO)
   {
   /* We know there is enough place on the stack. */
-  OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, sizeof(sljit_sw));
+  OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, sizeof(sljit_sw));
   OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), STACK(0), SLJIT_IMM, 0);
   JUMPTO(SLJIT_JUMP, CURRENT_AS(assert_backtrack)->matchingpath);
   JUMPHERE(brajump);
@@ -9954,7 +9972,7 @@ else if (ket == OP_KETRMIN)
       else
         {
         OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), private_data_ptr);
-        CMPTO(SLJIT_NOT_EQUAL, STR_PTR, 0, SLJIT_MEM1(TMP1), (CURRENT_AS(bracket_backtrack)->u.framesize + 1) * sizeof(sljit_sw), CURRENT_AS(bracket_backtrack)->recursive_matchingpath);
+        CMPTO(SLJIT_NOT_EQUAL, STR_PTR, 0, SLJIT_MEM1(TMP1), STACK(-CURRENT_AS(bracket_backtrack)->u.framesize - 2), CURRENT_AS(bracket_backtrack)->recursive_matchingpath);
         }
       /* Drop STR_PTR for non-greedy plus quantifier. */
       if (opcode != OP_ONCE)
@@ -10060,7 +10078,7 @@ if (SLJIT_UNLIKELY(opcode == OP_COND) || SLJIT_UNLIKELY(opcode == OP_SCOND))
       {
       OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(SLJIT_SP), assert->private_data_ptr);
       add_jump(compiler, &common->revertframes, JUMP(SLJIT_FAST_CALL));
-      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), assert->private_data_ptr, SLJIT_MEM1(STACK_TOP), assert->framesize * sizeof(sljit_sw));
+      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), assert->private_data_ptr, SLJIT_MEM1(STACK_TOP), STACK(-assert->framesize - 1));
       }
     cond = JUMP(SLJIT_JUMP);
     set_jumps(CURRENT_AS(bracket_backtrack)->u.assert->condfailed, LABEL());
@@ -10201,7 +10219,7 @@ if (has_alternatives)
       {
       OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(SLJIT_SP), assert->private_data_ptr);
       add_jump(compiler, &common->revertframes, JUMP(SLJIT_FAST_CALL));
-      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), assert->private_data_ptr, SLJIT_MEM1(STACK_TOP), assert->framesize * sizeof(sljit_sw));
+      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), assert->private_data_ptr, SLJIT_MEM1(STACK_TOP), STACK(-assert->framesize - 1));
       }
     JUMPHERE(cond);
     }
@@ -10256,7 +10274,7 @@ else if (opcode == OP_ONCE)
   JUMPHERE(once);
   /* Restore previous private_data_ptr */
   if (CURRENT_AS(bracket_backtrack)->u.framesize >= 0)
-    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), private_data_ptr, SLJIT_MEM1(STACK_TOP), CURRENT_AS(bracket_backtrack)->u.framesize * sizeof(sljit_sw));
+    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), private_data_ptr, SLJIT_MEM1(STACK_TOP), STACK(-CURRENT_AS(bracket_backtrack)->u.framesize - 1));
   else if (ket == OP_KETRMIN)
     {
     OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), STACK(1));
@@ -10346,7 +10364,7 @@ if (current->topbacktracks)
   free_stack(common, CURRENT_AS(bracketpos_backtrack)->stacksize);
   JUMPHERE(jump);
   }
-OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), CURRENT_AS(bracketpos_backtrack)->private_data_ptr, SLJIT_MEM1(STACK_TOP), CURRENT_AS(bracketpos_backtrack)->framesize * sizeof(sljit_sw));
+OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), CURRENT_AS(bracketpos_backtrack)->private_data_ptr, SLJIT_MEM1(STACK_TOP), STACK(-CURRENT_AS(bracketpos_backtrack)->framesize - 1));
 }
 
 static SLJIT_INLINE void compile_braminzero_backtrackingpath(compiler_common *common, struct backtrack_common *current)
@@ -10392,10 +10410,10 @@ if (opcode == OP_THEN || opcode == OP_THEN_ARG)
     jump = JUMP(SLJIT_JUMP);
 
     loop = LABEL();
-    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), -(int)sizeof(sljit_sw));
+    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), STACK(0));
     JUMPHERE(jump);
-    CMPTO(SLJIT_NOT_EQUAL, SLJIT_MEM1(STACK_TOP), -(int)(2 * sizeof(sljit_sw)), TMP1, 0, loop);
-    CMPTO(SLJIT_NOT_EQUAL, SLJIT_MEM1(STACK_TOP), -(int)(3 * sizeof(sljit_sw)), TMP2, 0, loop);
+    CMPTO(SLJIT_NOT_EQUAL, SLJIT_MEM1(STACK_TOP), STACK(1), TMP1, 0, loop);
+    CMPTO(SLJIT_NOT_EQUAL, SLJIT_MEM1(STACK_TOP), STACK(2), TMP2, 0, loop);
     add_jump(compiler, &common->then_trap->quit, JUMP(SLJIT_JUMP));
     return;
     }
@@ -10645,7 +10663,7 @@ while (current)
     break;
 
     default:
-    SLJIT_ASSERT_STOP();
+    SLJIT_UNREACHABLE();
     break;
     }
   current = current->prev;
@@ -10684,7 +10702,7 @@ sljit_emit_fast_enter(compiler, TMP2, 0);
 count_match(common);
 allocate_stack(common, private_data_size + framesize + alternativesize);
 OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), STACK(private_data_size + framesize + alternativesize - 1), TMP2, 0);
-copy_private_data(common, ccbegin, ccend, TRUE, private_data_size + framesize + alternativesize, framesize + alternativesize, needs_control_head);
+copy_private_data(common, ccbegin, ccend, TRUE, framesize + alternativesize, private_data_size + framesize + alternativesize, needs_control_head);
 if (needs_control_head)
   OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->control_head_ptr, SLJIT_IMM, 0);
 OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->recursive_head_ptr, STACK_TOP, 0);
@@ -10737,9 +10755,9 @@ if (common->quit != NULL)
   OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(SLJIT_SP), common->recursive_head_ptr);
   if (needs_frame)
     {
-    OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, (framesize + alternativesize) * sizeof(sljit_sw));
-    add_jump(compiler, &common->revertframes, JUMP(SLJIT_FAST_CALL));
     OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, (framesize + alternativesize) * sizeof(sljit_sw));
+    add_jump(compiler, &common->revertframes, JUMP(SLJIT_FAST_CALL));
+    OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, (framesize + alternativesize) * sizeof(sljit_sw));
     }
   OP1(SLJIT_MOV, TMP3, 0, SLJIT_IMM, 0);
   common->quit = NULL;
@@ -10750,32 +10768,32 @@ set_jumps(common->accept, LABEL());
 OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(SLJIT_SP), common->recursive_head_ptr);
 if (needs_frame)
   {
-  OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, (framesize + alternativesize) * sizeof(sljit_sw));
-  add_jump(compiler, &common->revertframes, JUMP(SLJIT_FAST_CALL));
   OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, (framesize + alternativesize) * sizeof(sljit_sw));
+  add_jump(compiler, &common->revertframes, JUMP(SLJIT_FAST_CALL));
+  OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, (framesize + alternativesize) * sizeof(sljit_sw));
   }
 OP1(SLJIT_MOV, TMP3, 0, SLJIT_IMM, 1);
 
 JUMPHERE(jump);
 if (common->quit != NULL)
   set_jumps(common->quit, LABEL());
-copy_private_data(common, ccbegin, ccend, FALSE, private_data_size + framesize + alternativesize, framesize + alternativesize, needs_control_head);
+copy_private_data(common, ccbegin, ccend, FALSE, framesize + alternativesize, private_data_size + framesize + alternativesize, needs_control_head);
 free_stack(common, private_data_size + framesize + alternativesize);
 if (needs_control_head)
   {
-  OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), 2 * sizeof(sljit_sw));
-  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), sizeof(sljit_sw));
+  OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), STACK(-3));
+  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), STACK(-2));
   OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->recursive_head_ptr, TMP1, 0);
   OP1(SLJIT_MOV, TMP1, 0, TMP3, 0);
   OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->control_head_ptr, TMP2, 0);
   }
 else
   {
-  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), sizeof(sljit_sw));
+  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), STACK(-2));
   OP1(SLJIT_MOV, TMP1, 0, TMP3, 0);
   OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->recursive_head_ptr, TMP2, 0);
   }
-sljit_emit_fast_return(compiler, SLJIT_MEM1(STACK_TOP), 0);
+sljit_emit_fast_return(compiler, SLJIT_MEM1(STACK_TOP), STACK(-1));
 }
 
 #undef COMPILE_BACKTRACKINGPATH
@@ -11237,7 +11255,7 @@ OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), LOCALS1, TMP2, 0);
 OP1(SLJIT_MOV, TMP1, 0, ARGUMENTS, 0);
 OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(TMP1), SLJIT_OFFSETOF(jit_arguments, stack));
 OP1(SLJIT_MOV, SLJIT_MEM1(TMP1), SLJIT_OFFSETOF(struct sljit_stack, top), STACK_TOP, 0);
-OP2(SLJIT_ADD, TMP2, 0, SLJIT_MEM1(TMP1), SLJIT_OFFSETOF(struct sljit_stack, limit), SLJIT_IMM, STACK_GROWTH_RATE);
+OP2(SLJIT_SUB, TMP2, 0, SLJIT_MEM1(TMP1), SLJIT_OFFSETOF(struct sljit_stack, limit), SLJIT_IMM, STACK_GROWTH_RATE);
 
 sljit_emit_ijump(compiler, SLJIT_CALL2, SLJIT_IMM, SLJIT_FUNC_OFFSET(sljit_stack_resize));
 jump = CMP(SLJIT_NOT_EQUAL, SLJIT_RETURN_REG, 0, SLJIT_IMM, 0);
@@ -11391,10 +11409,10 @@ union {
 sljit_u8 local_space[MACHINE_STACK_SIZE];
 struct sljit_stack local_stack;
 
-local_stack.top = (sljit_sw)&local_space;
-local_stack.base = local_stack.top;
-local_stack.limit = local_stack.base + MACHINE_STACK_SIZE;
-local_stack.max_limit = local_stack.limit;
+local_stack.max_limit = local_space;
+local_stack.limit = local_space;
+local_stack.base = local_space + MACHINE_STACK_SIZE;
+local_stack.top = local_space + MACHINE_STACK_SIZE;
 arguments->stack = &local_stack;
 convert_executable_func.executable_func = executable_func;
 return convert_executable_func.call_executable_func(arguments);
diff --git a/src/third_party/pcre-8.39/pcre_jit_test.c b/src/third_party/pcre-8.41/pcre_jit_test.c
index 9b61ec000fa..034cb52697f 100644
--- a/src/third_party/pcre-8.39/pcre_jit_test.c
+++ b/src/third_party/pcre-8.41/pcre_jit_test.c
@@ -687,6 +687,7 @@ static struct regression_test_case regression_test_cases[] = {
 	{ PCRE_FIRSTLINE | PCRE_NEWLINE_LF | PCRE_DOTALL, 0 | F_NOMATCH, "ab.", "ab" },
 	{ MUA | PCRE_FIRSTLINE, 1 | F_NOMATCH, "^[a-d0-9]", "\nxx\nd" },
 	{ PCRE_NEWLINE_ANY | PCRE_FIRSTLINE | PCRE_DOTALL, 0, "....a", "012\n0a" },
+	{ MUA | PCRE_FIRSTLINE, 0, "[aC]", "a" },
 
 	/* Recurse. */
 	{ MUA, 0, "(a)(?1)", "aa" },
diff --git a/src/third_party/pcre-8.39/pcre_maketables.c b/src/third_party/pcre-8.41/pcre_maketables.c
index a44a6eaa905..a44a6eaa905 100644
--- a/src/third_party/pcre-8.39/pcre_maketables.c
+++ b/src/third_party/pcre-8.41/pcre_maketables.c
diff --git a/src/third_party/pcre-8.39/pcre_newline.c b/src/third_party/pcre-8.41/pcre_newline.c
index b8f5a4de19c..b8f5a4de19c 100644
--- a/src/third_party/pcre-8.39/pcre_newline.c
+++ b/src/third_party/pcre-8.41/pcre_newline.c
diff --git a/src/third_party/pcre-8.39/pcre_ord2utf8.c b/src/third_party/pcre-8.41/pcre_ord2utf8.c
index 95f1beb963e..95f1beb963e 100644
--- a/src/third_party/pcre-8.39/pcre_ord2utf8.c
+++ b/src/third_party/pcre-8.41/pcre_ord2utf8.c
diff --git a/src/third_party/pcre-8.39/pcre_printint.c b/src/third_party/pcre-8.41/pcre_printint.c
index 60dcb55efbf..60dcb55efbf 100644
--- a/src/third_party/pcre-8.39/pcre_printint.c
+++ b/src/third_party/pcre-8.41/pcre_printint.c
diff --git a/src/third_party/pcre-8.39/pcre_refcount.c b/src/third_party/pcre-8.41/pcre_refcount.c
index 79efa90f216..79efa90f216 100644
--- a/src/third_party/pcre-8.39/pcre_refcount.c
+++ b/src/third_party/pcre-8.41/pcre_refcount.c
diff --git a/src/third_party/pcre-8.39/pcre_scanner.cc b/src/third_party/pcre-8.41/pcre_scanner.cc
index 6be2be6829b..6be2be6829b 100644
--- a/src/third_party/pcre-8.39/pcre_scanner.cc
+++ b/src/third_party/pcre-8.41/pcre_scanner.cc
diff --git a/src/third_party/pcre-8.39/pcre_scanner.h b/src/third_party/pcre-8.41/pcre_scanner.h
index 5617e4515cb..5617e4515cb 100644
--- a/src/third_party/pcre-8.39/pcre_scanner.h
+++ b/src/third_party/pcre-8.41/pcre_scanner.h
diff --git a/src/third_party/pcre-8.39/pcre_scanner_unittest.cc b/src/third_party/pcre-8.41/pcre_scanner_unittest.cc
index c00312c4f63..623e2afda80 100644
--- a/src/third_party/pcre-8.39/pcre_scanner_unittest.cc
+++ b/src/third_party/pcre-8.41/pcre_scanner_unittest.cc
@@ -57,6 +57,7 @@
 } while (0)
 
 using std::vector;
+using std::string;
 using pcrecpp::StringPiece;
 using pcrecpp::Scanner;
 
diff --git a/src/third_party/pcre-8.39/pcre_string_utils.c b/src/third_party/pcre-8.41/pcre_string_utils.c
index 25eacc85073..25eacc85073 100644
--- a/src/third_party/pcre-8.39/pcre_string_utils.c
+++ b/src/third_party/pcre-8.41/pcre_string_utils.c
diff --git a/src/third_party/pcre-8.39/pcre_stringpiece.cc b/src/third_party/pcre-8.41/pcre_stringpiece.cc
index 67c0f1fc0e5..67c0f1fc0e5 100644
--- a/src/third_party/pcre-8.39/pcre_stringpiece.cc
+++ b/src/third_party/pcre-8.41/pcre_stringpiece.cc
diff --git a/src/third_party/pcre-8.39/pcre_stringpiece.h b/src/third_party/pcre-8.41/pcre_stringpiece.h
index cc3dc42963a..cb94f52a010 100644
--- a/src/third_party/pcre-8.39/pcre_stringpiece.h
+++ b/src/third_party/pcre-8.41/pcre_stringpiece.h
@@ -52,12 +52,12 @@
 
 #include <pcre.h>
 
+namespace pcrecpp {
+
 using std::memcmp;
 using std::strlen;
 using std::string;
 
-namespace pcrecpp {
-
 class PCRECPP_EXP_DEFN StringPiece {
  private:
   const char*   ptr_;
diff --git a/src/third_party/pcre-8.39/pcre_stringpiece.h.in b/src/third_party/pcre-8.41/pcre_stringpiece.h.in
index eb25826b453..f54f3f3b31b 100644
--- a/src/third_party/pcre-8.39/pcre_stringpiece.h.in
+++ b/src/third_party/pcre-8.41/pcre_stringpiece.h.in
@@ -52,12 +52,12 @@
 
 #include <pcre.h>
 
+namespace pcrecpp {
+
 using std::memcmp;
 using std::strlen;
 using std::string;
 
-namespace pcrecpp {
-
 class PCRECPP_EXP_DEFN StringPiece {
  private:
   const char*   ptr_;
diff --git a/src/third_party/pcre-8.39/pcre_stringpiece_unittest.cc b/src/third_party/pcre-8.41/pcre_stringpiece_unittest.cc
index 1c4759da3b0..88e73a1f976 100644
--- a/src/third_party/pcre-8.39/pcre_stringpiece_unittest.cc
+++ b/src/third_party/pcre-8.41/pcre_stringpiece_unittest.cc
@@ -24,6 +24,7 @@
   }                                                     \
 } while (0)
 
+using std::string;
 using pcrecpp::StringPiece;
 
 static void CheckSTLComparator() {
diff --git a/src/third_party/pcre-8.39/pcre_study.c b/src/third_party/pcre-8.41/pcre_study.c
index d9d4960d84e..d9d4960d84e 100644
--- a/src/third_party/pcre-8.39/pcre_study.c
+++ b/src/third_party/pcre-8.41/pcre_study.c
diff --git a/src/third_party/pcre-8.39/pcre_tables.c b/src/third_party/pcre-8.41/pcre_tables.c
index 4960af57c4d..5e18e8cf904 100644
--- a/src/third_party/pcre-8.39/pcre_tables.c
+++ b/src/third_party/pcre-8.41/pcre_tables.c
@@ -6,7 +6,7 @@
 and semantics are as close as possible to those of the Perl 5 language.
 
                        Written by Philip Hazel
-           Copyright (c) 1997-2012 University of Cambridge
+           Copyright (c) 1997-2017 University of Cambridge
 
 -----------------------------------------------------------------------------
 Redistribution and use in source and binary forms, with or without
@@ -161,7 +161,7 @@ const pcre_uint32 PRIV(ucp_gbtable[]) = {
 
    (1<<ucp_gbExtend)|(1<<ucp_gbSpacingMark),                /*  5 SpacingMark */
    (1<<ucp_gbExtend)|(1<<ucp_gbSpacingMark)|(1<<ucp_gbL)|   /*  6 L */
-     (1<<ucp_gbL)|(1<<ucp_gbV)|(1<<ucp_gbLV)|(1<<ucp_gbLVT),
+     (1<<ucp_gbV)|(1<<ucp_gbLV)|(1<<ucp_gbLVT),
 
    (1<<ucp_gbExtend)|(1<<ucp_gbSpacingMark)|(1<<ucp_gbV)|   /*  7 V */
      (1<<ucp_gbT),
diff --git a/src/third_party/pcre-8.39/pcre_ucd.c b/src/third_party/pcre-8.41/pcre_ucd.c
index 69c4fd42c34..f22f826c4c2 100644
--- a/src/third_party/pcre-8.39/pcre_ucd.c
+++ b/src/third_party/pcre-8.41/pcre_ucd.c
@@ -38,6 +38,20 @@ const pcre_uint16 PRIV(ucd_stage2)[] = {0};
 const pcre_uint32 PRIV(ucd_caseless_sets)[] = {0};
 #else
 
+/* If the 32-bit library is run in non-32-bit mode, character values
+greater than 0x10ffff may be encountered. For these we set up a
+special record. */
+
+#ifdef COMPILE_PCRE32
+const ucd_record PRIV(dummy_ucd_record)[] = {{
+  ucp_Common,    /* script */
+  ucp_Cn,        /* type unassigned */
+  ucp_gbOther,   /* grapheme break property */
+  0,             /* case set */
+  0,             /* other case */
+  }};
+#endif
+
 /* When recompiling tables with a new Unicode version, please check the
 types in this structure definition from pcre_internal.h (the actual
 field names will be different):
diff --git a/src/third_party/pcre-8.39/pcre_valid_utf8.c b/src/third_party/pcre-8.41/pcre_valid_utf8.c
index 3b0f6464a35..3b0f6464a35 100644
--- a/src/third_party/pcre-8.39/pcre_valid_utf8.c
+++ b/src/third_party/pcre-8.41/pcre_valid_utf8.c
diff --git a/src/third_party/pcre-8.39/pcre_version.c b/src/third_party/pcre-8.41/pcre_version.c
index ae86ff28bc8..ae86ff28bc8 100644
--- a/src/third_party/pcre-8.39/pcre_version.c
+++ b/src/third_party/pcre-8.41/pcre_version.c
diff --git a/src/third_party/pcre-8.39/pcre_xclass.c b/src/third_party/pcre-8.41/pcre_xclass.c
index ef759a589a6..ef759a589a6 100644
--- a/src/third_party/pcre-8.39/pcre_xclass.c
+++ b/src/third_party/pcre-8.41/pcre_xclass.c
diff --git a/src/third_party/pcre-8.39/pcrecpp.cc b/src/third_party/pcre-8.41/pcrecpp.cc
index d09c9abc516..d09c9abc516 100644
--- a/src/third_party/pcre-8.39/pcrecpp.cc
+++ b/src/third_party/pcre-8.41/pcrecpp.cc
diff --git a/src/third_party/pcre-8.39/pcrecpp.h b/src/third_party/pcre-8.41/pcrecpp.h
index 3e594b0d439..3e594b0d439 100644
--- a/src/third_party/pcre-8.39/pcrecpp.h
+++ b/src/third_party/pcre-8.41/pcrecpp.h
diff --git a/src/third_party/pcre-8.39/pcrecpp_internal.h b/src/third_party/pcre-8.41/pcrecpp_internal.h
index 827f9e04e2a..827f9e04e2a 100644
--- a/src/third_party/pcre-8.39/pcrecpp_internal.h
+++ b/src/third_party/pcre-8.41/pcrecpp_internal.h
diff --git a/src/third_party/pcre-8.39/pcrecpp_unittest.cc b/src/third_party/pcre-8.41/pcrecpp_unittest.cc
index 92cae8fbea5..4b15fbef1c3 100644
--- a/src/third_party/pcre-8.39/pcrecpp_unittest.cc
+++ b/src/third_party/pcre-8.41/pcrecpp_unittest.cc
@@ -43,6 +43,7 @@
 #include <vector>
 #include "pcrecpp.h"
 
+using std::string;
 using pcrecpp::StringPiece;
 using pcrecpp::RE;
 using pcrecpp::RE_Options;
diff --git a/src/third_party/pcre-8.39/pcrecpparg.h b/src/third_party/pcre-8.41/pcrecpparg.h
index b4f9c3f4989..b4f9c3f4989 100644
--- a/src/third_party/pcre-8.39/pcrecpparg.h
+++ b/src/third_party/pcre-8.41/pcrecpparg.h
diff --git a/src/third_party/pcre-8.39/pcrecpparg.h.in b/src/third_party/pcre-8.41/pcrecpparg.h.in
index 61bcab5402c..61bcab5402c 100644
--- a/src/third_party/pcre-8.39/pcrecpparg.h.in
+++ b/src/third_party/pcre-8.41/pcrecpparg.h.in
diff --git a/src/third_party/pcre-8.39/pcredemo.c b/src/third_party/pcre-8.41/pcredemo.c
index 946aba45cdc..946aba45cdc 100644
--- a/src/third_party/pcre-8.39/pcredemo.c
+++ b/src/third_party/pcre-8.41/pcredemo.c
diff --git a/src/third_party/pcre-8.39/pcregexp.pas b/src/third_party/pcre-8.41/pcregexp.pas
index bb2b3da8f3d..bb2b3da8f3d 100644
--- a/src/third_party/pcre-8.39/pcregexp.pas
+++ b/src/third_party/pcre-8.41/pcregexp.pas
diff --git a/src/third_party/pcre-8.39/pcregrep.c b/src/third_party/pcre-8.41/pcregrep.c
index cd53c648da2..317f7454e13 100644
--- a/src/third_party/pcre-8.39/pcregrep.c
+++ b/src/third_party/pcre-8.41/pcregrep.c
@@ -1803,6 +1803,7 @@ while (ptr < endptr)
         match = FALSE;
         if (line_buffered) fflush(stdout);
         rc = 0;                      /* Had some success */
+
         startoffset = offsets[1];    /* Restart after the match */
         if (startoffset <= oldstartoffset)
           {
@@ -1812,6 +1813,22 @@ while (ptr < endptr)
           if (utf8)
             while ((matchptr[startoffset] & 0xc0) == 0x80) startoffset++;
           }
+
+        /* If the current match ended past the end of the line (only possible
+        in multiline mode), we must move on to the line in which it did end
+        before searching for more matches. */
+
+        while (startoffset > (int)linelength)
+          {
+          matchptr = ptr += linelength + endlinelength;
+          filepos += (int)(linelength + endlinelength);
+          linenumber++;
+          startoffset -= (int)(linelength + endlinelength);
+          t = end_of_line(ptr, endptr, &endlinelength);
+          linelength = t - ptr - endlinelength;
+          length = (size_t)(endptr - ptr);
+          }
+
         goto ONLY_MATCHING_RESTART;
         }
       }
@@ -3173,9 +3190,11 @@ for (j = 1, cp = patterns; cp != NULL; j++, cp = cp->next)
   cp->hint = pcre_study(cp->compiled, study_options, &error);
   if (error != NULL)
     {
-    char s[16];
-    if (patterns->next == NULL) s[0] = 0; else sprintf(s, " number %d", j);
-    fprintf(stderr, "pcregrep: Error while studying regex%s: %s\n", s, error);
+    if (patterns->next == NULL)
+      fprintf(stderr, "pcregrep: Error while studying regex: %s\n", error);
+    else
+      fprintf(stderr, "pcregrep: Error while studying regex number %d: %s\n",
+        j, error);
     goto EXIT2;
     }
 #ifdef SUPPORT_PCREGREP_JIT
diff --git a/src/third_party/pcre-8.39/pcreposix.c b/src/third_party/pcre-8.41/pcreposix.c
index cf75588c40d..7b404a71100 100644
--- a/src/third_party/pcre-8.39/pcreposix.c
+++ b/src/third_party/pcre-8.41/pcreposix.c
@@ -6,7 +6,7 @@
 and semantics are as close as possible to those of the Perl 5 language.
 
                        Written by Philip Hazel
-           Copyright (c) 1997-2016 University of Cambridge
+           Copyright (c) 1997-2017 University of Cambridge
 
 -----------------------------------------------------------------------------
 Redistribution and use in source and binary forms, with or without
@@ -389,8 +389,8 @@ if (rc >= 0)
     {
     for (i = 0; i < (size_t)rc; i++)
       {
-      pmatch[i].rm_so = ovector[i*2];
-      pmatch[i].rm_eo = ovector[i*2+1];
+      pmatch[i].rm_so = ovector[i*2] + so;
+      pmatch[i].rm_eo = ovector[i*2+1] + so;
       }
     if (allocated_ovector) free(ovector);
     for (; i < nmatch; i++) pmatch[i].rm_so = pmatch[i].rm_eo = -1;
diff --git a/src/third_party/pcre-8.39/pcreposix.h b/src/third_party/pcre-8.41/pcreposix.h
index c77c0b0523c..c77c0b0523c 100644
--- a/src/third_party/pcre-8.39/pcreposix.h
+++ b/src/third_party/pcre-8.41/pcreposix.h
diff --git a/src/third_party/pcre-8.39/pcretest.c b/src/third_party/pcre-8.41/pcretest.c
index 78ef5177df7..f1303037281 100644
--- a/src/third_party/pcre-8.39/pcretest.c
+++ b/src/third_party/pcre-8.41/pcretest.c
@@ -177,7 +177,7 @@ that differ in their output from isprint() even in the "C" locale. */
 #define PRINTABLE(c) ((c) >= 32 && (c) < 127)
 #endif
 
-#define PRINTOK(c) (locale_set? isprint(c) : PRINTABLE(c))
+#define PRINTOK(c) (locale_set? (((c) < 256) && isprint(c)) : PRINTABLE(c))
 
 /* Posix support is disabled in 16 or 32 bit only mode. */
 #if !defined SUPPORT_PCRE8 && !defined NOPOSIX
@@ -426,11 +426,11 @@ argument, the casting might be incorrectly applied. */
 #define PCRE_COPY_NAMED_SUBSTRING32(rc, re, bptr, offsets, count, \
     namesptr, cbuffer, size) \
   rc = pcre32_copy_named_substring((pcre32 *)re, (PCRE_SPTR32)bptr, offsets, \
-    count, (PCRE_SPTR32)namesptr, (PCRE_UCHAR32 *)cbuffer, size/2)
+    count, (PCRE_SPTR32)namesptr, (PCRE_UCHAR32 *)cbuffer, size/4)
 
 #define PCRE_COPY_SUBSTRING32(rc, bptr, offsets, count, i, cbuffer, size) \
   rc = pcre32_copy_substring((PCRE_SPTR32)bptr, offsets, count, i, \
-    (PCRE_UCHAR32 *)cbuffer, size/2)
+    (PCRE_UCHAR32 *)cbuffer, size/4)
 
 #define PCRE_DFA_EXEC32(count, re, extra, bptr, len, start_offset, options, \
     offsets, size_offsets, workspace, size_workspace) \
@@ -1982,6 +1982,7 @@ return(result);
 static int pchar(pcre_uint32 c, FILE *f)
 {
 int n = 0;
+char tempbuffer[16];
 if (PRINTOK(c))
   {
   if (f != NULL) fprintf(f, "%c", c);
@@ -2003,6 +2004,8 @@ if (c < 0x100)
   }
 
 if (f != NULL) n = fprintf(f, "\\x{%02x}", c);
+  else n = sprintf(tempbuffer, "\\x{%02x}", c);
+
 return n >= 0 ? n : 0;
 }
 
@@ -4831,7 +4834,16 @@ while (!done)
         continue;
 
         case 'O':
-        while(isdigit(*p)) n = n * 10 + *p++ - '0';
+        while(isdigit(*p))
+          {
+          if (n > (INT_MAX-10)/10)   /* Hack to stop fuzzers */
+            {
+            printf("** \\O argument is too big\n");
+            yield = 1;
+            goto EXIT;
+            }
+          n = n * 10 + *p++ - '0';
+          }
         if (n > size_offsets_max)
           {
           size_offsets_max = n;
@@ -5042,7 +5054,7 @@ while (!done)
 
     if ((all_use_dfa || use_dfa) && find_match_limit)
       {
-      printf("**Match limit not relevant for DFA matching: ignored\n");
+      printf("** Match limit not relevant for DFA matching: ignored\n");
       find_match_limit = 0;
       }
 
@@ -5255,10 +5267,17 @@ while (!done)
 
         if (do_allcaps)
           {
-          if (new_info(re, NULL, PCRE_INFO_CAPTURECOUNT, &count) < 0)
-            goto SKIP_DATA;
-          count++;   /* Allow for full match */
-          if (count * 2 > use_size_offsets) count = use_size_offsets/2;
+          if (all_use_dfa || use_dfa)
+            {
+            fprintf(outfile, "** Show all captures ignored after DFA matching\n");
+            }
+          else
+           {
+            if (new_info(re, NULL, PCRE_INFO_CAPTURECOUNT, &count) < 0)
+              goto SKIP_DATA;
+            count++;   /* Allow for full match */
+            if (count * 2 > use_size_offsets) count = use_size_offsets/2;
+            }
           }
 
         /* Output the captured substrings. Note that, for the matched string,
diff --git a/src/third_party/pcre-8.39/perltest.pl b/src/third_party/pcre-8.41/perltest.pl
index 29b808b5293..29b808b5293 100755
--- a/src/third_party/pcre-8.39/perltest.pl
+++ b/src/third_party/pcre-8.41/perltest.pl
diff --git a/src/third_party/pcre-8.39/ucp.h b/src/third_party/pcre-8.41/ucp.h
index 2fa00296e42..2fa00296e42 100644
--- a/src/third_party/pcre-8.39/ucp.h
+++ b/src/third_party/pcre-8.41/ucp.h
diff --git a/src/third_party/scripts/pcre_get_sources.sh b/src/third_party/scripts/pcre_get_sources.sh
index da73652ef66..16c79a3e911 100644
--- a/src/third_party/scripts/pcre_get_sources.sh
+++ b/src/third_party/scripts/pcre_get_sources.sh
@@ -16,7 +16,7 @@ if [ "$#" -ne 0 ]; then
     exit 1
 fi
 
-VERSION=8.39
+VERSION=8.41
 NAME=pcre
 TARBALL=$NAME-$VERSION.tar.gz
 TARBALL_DIR=$NAME-$VERSION
diff --git a/src/third_party/wiredtiger/build_linux/wiredtiger_config.h b/src/third_party/wiredtiger/build_linux/wiredtiger_config.h
index 1122e1e319d..6fffea61ad1 100644
--- a/src/third_party/wiredtiger/build_linux/wiredtiger_config.h
+++ b/src/third_party/wiredtiger/build_linux/wiredtiger_config.h
@@ -82,6 +82,9 @@
 /* Define to 1 if you have the `posix_memalign' function. */
 #define HAVE_POSIX_MEMALIGN 1
 
+/* Define to 1 if pthread condition variables support monotonic clocks. */
+#define HAVE_PTHREAD_COND_MONOTONIC 1
+
 /* Define to 1 if you have the <pthread_np.h> header file. */
 /* #undef HAVE_PTHREAD_NP_H */
 
diff --git a/src/third_party/wiredtiger/build_posix/configure.ac.in b/src/third_party/wiredtiger/build_posix/configure.ac.in
index 0fef587b4b8..415545a0d56 100644
--- a/src/third_party/wiredtiger/build_posix/configure.ac.in
+++ b/src/third_party/wiredtiger/build_posix/configure.ac.in
@@ -160,6 +160,44 @@ AS_CASE([$host_os], [darwin*], [], [AC_CHECK_FUNCS([fdatasync])])
 # the generic declaration in AC_CHECK_FUNCS is incompatible.
 AX_FUNC_POSIX_MEMALIGN
 
+# Check for POSIX condition variables with monotonic clock support
+AC_CACHE_CHECK([for condition waits with monotonic clock support],
+  [wt_cv_pthread_cond_monotonic],
+  [AC_RUN_IFELSE([AC_LANG_SOURCE([[
+#include <errno.h>
+#include <pthread.h>
+#include <stdlib.h>
+#include <time.h>
+
+int main()
+{
+  int ret;
+  pthread_condattr_t condattr;
+  pthread_cond_t cond;
+  pthread_mutex_t mtx;
+  struct timespec ts;
+
+  if ((ret = pthread_condattr_init(&condattr)) != 0) exit(1);
+  if ((ret = pthread_condattr_setclock(&condattr, CLOCK_MONOTONIC)) != 0) exit(1);
+  if ((ret = pthread_cond_init(&cond, &condattr)) != 0) exit(1);
+  if ((ret = pthread_mutex_init(&mtx, NULL)) != 0) exit(1);
+  if ((ret = clock_gettime(CLOCK_MONOTONIC, &ts)) != 0) exit(1);
+  ts.tv_sec += 1;
+  if ((ret = pthread_mutex_lock(&mtx)) != 0) exit(1);
+  if ((ret = pthread_cond_timedwait(&cond, &mtx, &ts)) != 0 && ret != EINTR && ret != ETIMEDOUT) exit(1);
+
+  exit(0);
+}
+    ]])],
+    [wt_pthread_cond_monotonic=yes],
+    [wt_pthread_cond_monotonic=no],
+    [wt_pthread_cond_monotonic=no])])
+AC_MSG_RESULT($wt_pthread_cond_monotonic)
+if test "$wt_pthread_cond_monotonic" = "yes" ; then
+  AC_DEFINE([HAVE_PTHREAD_COND_MONOTONIC], [1],
+    [Define to 1 if pthread condition variables support monotonic clocks.])
+fi
+
 AC_SYS_LARGEFILE
 
 AC_C_BIGENDIAN
diff --git a/src/third_party/wiredtiger/build_win/wiredtiger_config.h b/src/third_party/wiredtiger/build_win/wiredtiger_config.h
index 78d2784cb70..8babdbfdc1b 100644
--- a/src/third_party/wiredtiger/build_win/wiredtiger_config.h
+++ b/src/third_party/wiredtiger/build_win/wiredtiger_config.h
@@ -79,6 +79,9 @@
 /* Define to 1 if you have the <memory.h> header file. */
 /* #undef HAVE_MEMORY_H */
 
+/* Define to 1 if pthread condition variables support monotonic clocks. */
+/* #undef HAVE_PTHREAD_COND_MONOTONIC */
+
 /* Define to 1 if you have the `posix_fadvise' function. */
 /* #undef HAVE_POSIX_FADVISE */
 
diff --git a/src/third_party/wiredtiger/dist/api_data.py b/src/third_party/wiredtiger/dist/api_data.py
index 22600dd5e29..596099647be 100644
--- a/src/third_party/wiredtiger/dist/api_data.py
+++ b/src/third_party/wiredtiger/dist/api_data.py
@@ -753,6 +753,9 @@ wiredtiger_open_common =\
     Config('session_scratch_max', '2MB', r'''
         maximum memory to cache in each session''',
         type='int', undoc=True),
+    Config('session_table_cache', 'true', r'''
+        Maintain a per-session cache of tables''',
+        type='boolean'),
     Config('transaction_sync', '', r'''
         how to sync log records when the transaction commits''',
         type='category', subconfig=[
diff --git a/src/third_party/wiredtiger/dist/filelist b/src/third_party/wiredtiger/dist/filelist
index 5a3348b940a..f53509e96ec 100644
--- a/src/third_party/wiredtiger/dist/filelist
+++ b/src/third_party/wiredtiger/dist/filelist
@@ -191,6 +191,7 @@ src/support/rand.c
 src/support/scratch.c
 src/support/stat.c
 src/support/thread_group.c
+src/support/time.c
 src/txn/txn.c
 src/txn/txn_ckpt.c
 src/txn/txn_ext.c
diff --git a/src/third_party/wiredtiger/dist/flags.py b/src/third_party/wiredtiger/dist/flags.py
index 64b5d789e72..1ce717f3586 100644
--- a/src/third_party/wiredtiger/dist/flags.py
+++ b/src/third_party/wiredtiger/dist/flags.py
@@ -32,7 +32,6 @@ flags = {
         'READ_PREV',
         'READ_RESTART_OK',
         'READ_SKIP_INTL',
-        'READ_SKIP_LEAF',
         'READ_TRUNCATE',
         'READ_WONT_NEED',
     ],
@@ -111,6 +110,7 @@ flags = {
         'CONN_SERVER_LSM',
         'CONN_SERVER_STATISTICS',
         'CONN_SERVER_SWEEP',
+        'CONN_TABLE_CACHE',
         'CONN_WAS_BACKUP',
     ],
     'session' : [
diff --git a/src/third_party/wiredtiger/dist/s_string.ok b/src/third_party/wiredtiger/dist/s_string.ok
index f3852d00ac8..fad8c2f021a 100644
--- a/src/third_party/wiredtiger/dist/s_string.ok
+++ b/src/third_party/wiredtiger/dist/s_string.ok
@@ -807,6 +807,7 @@ intl
 intnum
 intpack
 intptr
+intr
 intrin
 inuse
 io
@@ -1060,6 +1061,7 @@ rebalancing
 recno
 recnos
 reconfig
+reconfigures
 reconfiguring
 recsize
 rectype
diff --git a/src/third_party/wiredtiger/dist/stat_data.py b/src/third_party/wiredtiger/dist/stat_data.py
index 512892eb44d..b66e95ce49b 100644
--- a/src/third_party/wiredtiger/dist/stat_data.py
+++ b/src/third_party/wiredtiger/dist/stat_data.py
@@ -431,11 +431,19 @@ connection_stats = [
     ##########################################
     YieldStat('application_cache_time', 'application thread time waiting for cache (usecs)'),
     YieldStat('application_evict_time', 'application thread time evicting (usecs)'),
+    YieldStat('child_modify_blocked_page', 'page reconciliation yielded due to child modification'),
+    YieldStat('conn_close_blocked_lsm', 'connection close yielded for lsm manager shutdown'),
+    YieldStat('dhandle_lock_blocked', 'data handle lock yielded'),
+    YieldStat('log_server_sync_blocked', 'log server sync yielded for log write'),
     YieldStat('page_busy_blocked', 'page acquire busy blocked'),
+    YieldStat('page_del_rollback_blocked', 'page delete rollback yielded for instantiation'),
     YieldStat('page_forcible_evict_blocked', 'page acquire eviction blocked'),
+    YieldStat('page_index_slot_blocked', 'reference for page index and slot yielded'),
     YieldStat('page_locked_blocked', 'page acquire locked blocked'),
     YieldStat('page_read_blocked', 'page acquire read blocked'),
     YieldStat('page_sleep', 'page acquire time sleeping (usecs)'),
+    YieldStat('tree_descend_blocked', 'tree descend one level yielded for split page index update'),
+    YieldStat('txn_release_blocked', 'connection close blocked waiting for transaction state stabilization'),
 ]
 
 connection_stats = sorted(connection_stats, key=attrgetter('desc'))
diff --git a/src/third_party/wiredtiger/import.data b/src/third_party/wiredtiger/import.data
index 03d5746ffde..e867147e59e 100644
--- a/src/third_party/wiredtiger/import.data
+++ b/src/third_party/wiredtiger/import.data
@@ -1,5 +1,5 @@
 {
-    "commit": "d2f2eae6d7718a53ac5bacf7141773ee0696f3c6", 
+    "commit": "0b36171f4aa0bea6ab6118b0d3bcf6329a24939e", 
     "github": "wiredtiger/wiredtiger.git", 
     "vendor": "wiredtiger", 
     "branch": "mongodb-3.4"
diff --git a/src/third_party/wiredtiger/src/async/async_worker.c b/src/third_party/wiredtiger/src/async/async_worker.c
index 11f59ed14f1..89877652f1e 100644
--- a/src/third_party/wiredtiger/src/async/async_worker.c
+++ b/src/third_party/wiredtiger/src/async/async_worker.c
@@ -301,11 +301,10 @@ __wt_async_worker(void *arg)
 		WT_ERR(__async_op_dequeue(conn, session, &op));
 		if (op != NULL && op != &async->flush_op) {
 			/*
-			 * If an operation fails, we want the worker thread to
-			 * keep running, unless there is a panic.
+			 * Operation failure doesn't cause the worker thread to
+			 * exit.
 			 */
 			(void)__async_worker_op(session, op, &worker);
-			WT_ERR(WT_SESSION_CHECK_PANIC(session));
 		} else if (async->flush_state == WT_ASYNC_FLUSHING) {
 			/*
 			 * Worker flushing going on.  Last worker to the party
diff --git a/src/third_party/wiredtiger/src/btree/bt_cursor.c b/src/third_party/wiredtiger/src/btree/bt_cursor.c
index f0aa632551b..28977eee9f5 100644
--- a/src/third_party/wiredtiger/src/btree/bt_cursor.c
+++ b/src/third_party/wiredtiger/src/btree/bt_cursor.c
@@ -52,15 +52,18 @@ __cursor_state_restore(WT_CURSOR *cursor, WT_CURFILE_STATE *state)
 
 /*
  * __cursor_page_pinned --
- *	Return if we have a page pinned and it's not been flagged for forced
- * eviction (the forced eviction test is so we periodically release pages
- * grown too large).
+ *	Return if we have a page pinned.
  */
 static inline bool
-__cursor_page_pinned(WT_CURSOR_BTREE *cbt)
+__cursor_page_pinned(WT_CURSOR_BTREE *cbt, bool eviction_ok)
 {
+	/*
+	 * Optionally fail the page-pinned test when the page is flagged for
+	 * forced eviction (so we periodically release pages grown too large).
+	 * The test is optional as not all callers can release pinned pages.
+	 */
 	return (F_ISSET(cbt, WT_CBT_ACTIVE) &&
-	    cbt->ref->page->read_gen != WT_READGEN_OLDEST);
+	    (!eviction_ok || cbt->ref->page->read_gen != WT_READGEN_OLDEST));
 }
 
 /*
@@ -156,8 +159,10 @@ __cursor_disable_bulk(WT_SESSION_IMPL *session, WT_BTREE *btree)
 	 * into a tree.  Eviction is disabled when an empty tree is opened, and
 	 * it must only be enabled once.
 	 */
-	if (__wt_atomic_cas8(&btree->original, 1, 0))
+	if (__wt_atomic_cas8(&btree->original, 1, 0)) {
+		btree->evict_disabled_open = false;
 		__wt_evict_file_exclusive_off(session);
+	}
 }
 
 /*
@@ -443,7 +448,7 @@ __wt_btcur_search(WT_CURSOR_BTREE *cbt)
 	 * from the root.
 	 */
 	valid = false;
-	if (__cursor_page_pinned(cbt)) {
+	if (__cursor_page_pinned(cbt, true)) {
 		__wt_txn_cursor_op(session);
 
 		WT_ERR(btree->type == BTREE_ROW ?
@@ -535,7 +540,7 @@ __wt_btcur_search_near(WT_CURSOR_BTREE *cbt, int *exactp)
 	 * existing record.
 	 */
 	valid = false;
-	if (btree->type == BTREE_ROW && __cursor_page_pinned(cbt)) {
+	if (btree->type == BTREE_ROW && __cursor_page_pinned(cbt, true)) {
 		__wt_txn_cursor_op(session);
 
 		WT_ERR(__cursor_row_search(session, cbt, cbt->ref, true));
@@ -665,7 +670,7 @@ __wt_btcur_insert(WT_CURSOR_BTREE *cbt)
 	 * configured for append aren't included, regardless of whether or not
 	 * they meet all other criteria.
 	 */
-	if (__cursor_page_pinned(cbt) &&
+	if (__cursor_page_pinned(cbt, true) &&
 	    F_ISSET_ALL(cursor, WT_CURSTD_KEY_INT | WT_CURSTD_OVERWRITE) &&
 	    !append_key) {
 		WT_ERR(__wt_txn_autocommit_check(session));
@@ -880,8 +885,22 @@ __wt_btcur_remove(WT_CURSOR_BTREE *cbt)
 	 * removed, and the record must exist with a positioned cursor. The
 	 * cursor won't be positioned on a page with an external key set, but
 	 * be sure.
+	 *
+	 * There's trickiness in the page-pinned check. By definition a remove
+	 * operation leaves a cursor positioned if it's initially positioned.
+	 * However, if every item on the page is deleted and we unpin the page,
+	 * eviction might delete the page and our search will re-instantiate an
+	 * empty page for us. Cursor remove returns not-found whether or not
+	 * that eviction/deletion happens and it's OK unless cursor-overwrite
+	 * is configured (which means we return success even if there's no item
+	 * to delete). In that case, we'll fail when we try to point the cursor
+	 * at the key on the page to satisfy the positioned requirement. It's
+	 * arguably safe to simply leave the key initialized in the cursor (as
+	 * that's all a positioned cursor implies), but it's probably safer to
+	 * avoid page eviction entirely in the positioned case.
 	 */
-	if (__cursor_page_pinned(cbt) && F_ISSET(cursor, WT_CURSTD_KEY_INT)) {
+	if (__cursor_page_pinned(cbt, !positioned) &&
+	    F_ISSET(cursor, WT_CURSTD_KEY_INT)) {
 		WT_ERR(__wt_txn_autocommit_check(session));
 
 		/*
@@ -1024,7 +1043,8 @@ __wt_btcur_update(WT_CURSOR_BTREE *cbt)
 	 * cursor won't be positioned on a page with an external key set, but
 	 * be sure.
 	 */
-	if (__cursor_page_pinned(cbt) && F_ISSET(cursor, WT_CURSTD_KEY_INT)) {
+	if (__cursor_page_pinned(cbt, true) &&
+	    F_ISSET(cursor, WT_CURSTD_KEY_INT)) {
 		WT_ERR(__wt_txn_autocommit_check(session));
 		/*
 		 * The cursor position may not be exact (the cursor's comparison
diff --git a/src/third_party/wiredtiger/src/btree/bt_delete.c b/src/third_party/wiredtiger/src/btree/bt_delete.c
index b55ad291c5e..5c4625044d3 100644
--- a/src/third_party/wiredtiger/src/btree/bt_delete.c
+++ b/src/third_party/wiredtiger/src/btree/bt_delete.c
@@ -153,6 +153,7 @@ void
 __wt_delete_page_rollback(WT_SESSION_IMPL *session, WT_REF *ref)
 {
 	WT_UPDATE **upd;
+	uint64_t yield_count;
 
 	/*
 	 * If the page is still "deleted", it's as we left it, reset the state
@@ -160,7 +161,7 @@ __wt_delete_page_rollback(WT_SESSION_IMPL *session, WT_REF *ref)
 	 * instantiated or being instantiated.  Loop because it's possible for
 	 * the page to return to the deleted state if instantiation fails.
 	 */
-	for (;; __wt_yield())
+	for (yield_count = 0;; yield_count++, __wt_yield())
 		switch (ref->state) {
 		case WT_REF_DISK:
 		case WT_REF_READING:
@@ -173,7 +174,7 @@ __wt_delete_page_rollback(WT_SESSION_IMPL *session, WT_REF *ref)
 			 */
 			if (__wt_atomic_casv32(
 			    &ref->state, WT_REF_DELETED, WT_REF_DISK))
-				return;
+				goto done;
 			break;
 		case WT_REF_LOCKED:
 			/*
@@ -203,8 +204,10 @@ __wt_delete_page_rollback(WT_SESSION_IMPL *session, WT_REF *ref)
 			 */
 			__wt_free(session, ref->page_del->update_list);
 			__wt_free(session, ref->page_del);
-			return;
+			goto done;
 		}
+
+done:	WT_STAT_CONN_INCRV(session, page_del_rollback_blocked, yield_count);
 }
 
 /*
diff --git a/src/third_party/wiredtiger/src/btree/bt_handle.c b/src/third_party/wiredtiger/src/btree/bt_handle.c
index a0da7df0998..8637a8cd751 100644
--- a/src/third_party/wiredtiger/src/btree/bt_handle.c
+++ b/src/third_party/wiredtiger/src/btree/bt_handle.c
@@ -66,7 +66,6 @@ __wt_btree_open(WT_SESSION_IMPL *session, const char *op_cfg[])
 	WT_DATA_HANDLE *dhandle;
 	WT_DECL_RET;
 	size_t root_addr_size;
-	uint32_t mask;
 	uint8_t root_addr[WT_BTREE_MAX_ADDR_COOKIE];
 	const char *filename;
 	bool creation, forced_salvage, readonly;
@@ -75,15 +74,14 @@ __wt_btree_open(WT_SESSION_IMPL *session, const char *op_cfg[])
 	dhandle = session->dhandle;
 
 	/*
-	 * This may be a re-open of an underlying object and we have to clean
-	 * up. We can't clear the operation flags, however, they're set by the
-	 * connection handle software that called us.
+	 * This may be a re-open, clean up the btree structure.
+	 * Clear the fields that don't persist across a re-open.
+	 * Clear all flags other than the operation flags (which are set by the
+	 * connection handle software that called us).
 	 */
 	WT_RET(__btree_clear(session));
-
-	mask = F_MASK(btree, WT_BTREE_SPECIAL_FLAGS);
-	memset(btree, 0, sizeof(*btree));
-	btree->flags = mask;
+	memset(btree, 0, WT_BTREE_CLEAR_SIZE);
+	F_CLR(btree, ~WT_BTREE_SPECIAL_FLAGS);
 
 	/* Set the data handle first, our called functions reasonably use it. */
 	btree->dhandle = dhandle;
@@ -185,13 +183,19 @@ __wt_btree_open(WT_SESSION_IMPL *session, const char *op_cfg[])
 	 *
 	 * Files that can still be bulk-loaded cannot be evicted.
 	 * Permanently cache-resident files can never be evicted.
-	 * Special operations don't enable eviction. (The underlying commands
-	 * may turn on eviction, but it's their decision.)
+	 * Special operations don't enable eviction. The underlying commands may
+	 * turn on eviction (for example, verify turns on eviction while working
+	 * a file to keep from consuming the cache), but it's their decision. If
+	 * an underlying command reconfigures eviction, it must either clear the
+	 * evict-disabled-open flag or restore the eviction configuration when
+	 * finished so that handle close behaves correctly.
 	 */
 	if (btree->original ||
 	    F_ISSET(btree, WT_BTREE_IN_MEMORY | WT_BTREE_REBALANCE |
-	    WT_BTREE_SALVAGE | WT_BTREE_UPGRADE | WT_BTREE_VERIFY))
+	    WT_BTREE_SALVAGE | WT_BTREE_UPGRADE | WT_BTREE_VERIFY)) {
 		WT_ERR(__wt_evict_file_exclusive_on(session));
+		btree->evict_disabled_open = true;
+	}
 
 	if (0) {
 err:		WT_TRET(__wt_btree_close(session));
@@ -228,6 +232,15 @@ __wt_btree_close(WT_SESSION_IMPL *session)
 		return (0);
 	F_SET(btree, WT_BTREE_CLOSED);
 
+	/*
+	 * If we turned eviction off and never turned it back on, do that now,
+	 * otherwise the counter will be off.
+	 */
+	if (btree->evict_disabled_open) {
+		btree->evict_disabled_open = false;
+		__wt_evict_file_exclusive_off(session);
+	}
+
 	/* Discard any underlying block manager resources. */
 	if ((bm = btree->bm) != NULL) {
 		btree->bm = NULL;
@@ -447,9 +460,11 @@ __btree_conf(WT_SESSION_IMPL *session, WT_CKPT *ckpt)
 	WT_RET(__wt_rwlock_init(session, &btree->ovfl_lock));
 	WT_RET(__wt_spin_init(session, &btree->flush_lock, "btree flush"));
 
-	btree->checkpointing = WT_CKPT_OFF;		/* Not checkpointing */
 	btree->modified = false;			/* Clean */
-	btree->write_gen = ckpt->write_gen;		/* Write generation */
+
+	btree->checkpointing = WT_CKPT_OFF;	/* Not checkpointing */
+	btree->write_gen = ckpt->write_gen;	/* Write generation */
+	btree->checkpoint_gen = S2C(session)->txn_global.checkpoint_gen;
 
 	return (0);
 }
diff --git a/src/third_party/wiredtiger/src/btree/bt_random.c b/src/third_party/wiredtiger/src/btree/bt_random.c
index c5948ec4ab5..b4f05c440ba 100644
--- a/src/third_party/wiredtiger/src/btree/bt_random.c
+++ b/src/third_party/wiredtiger/src/btree/bt_random.c
@@ -395,8 +395,7 @@ __wt_btcur_next_random(WT_CURSOR_BTREE *cbt)
 	 */
 	for (skip = cbt->next_random_leaf_skip; cbt->ref == NULL || skip > 0;) {
 		n = skip;
-		WT_ERR(__wt_tree_walk_skip(session, &cbt->ref, &skip,
-		    WT_READ_NO_GEN | WT_READ_SKIP_INTL | WT_READ_WONT_NEED));
+		WT_ERR(__wt_tree_walk_skip(session, &cbt->ref, &skip));
 		if (n == skip) {
 			if (skip == 0)
 				break;
diff --git a/src/third_party/wiredtiger/src/btree/bt_walk.c b/src/third_party/wiredtiger/src/btree/bt_walk.c
index 86484feb7c9..c22b99c55d0 100644
--- a/src/third_party/wiredtiger/src/btree/bt_walk.c
+++ b/src/third_party/wiredtiger/src/btree/bt_walk.c
@@ -18,9 +18,16 @@ __ref_index_slot(WT_SESSION_IMPL *session,
 {
 	WT_PAGE_INDEX *pindex;
 	WT_REF **start, **stop, **p, **t;
+	uint64_t yield_count;
 	uint32_t entries, slot;
 
-	for (;;) {
+	/*
+	 * If we don't find our reference, the page split and our home
+	 * pointer references the wrong page. When internal pages
+	 * split, their WT_REF structure home values are updated; yield
+	 * and wait for that to happen.
+	 */
+	for (yield_count = 0;; yield_count++, __wt_yield()) {
 		/*
 		 * Copy the parent page's index value: the page can split at
 		 * any time, but the index's value is always valid, even if
@@ -59,18 +66,13 @@ __ref_index_slot(WT_SESSION_IMPL *session,
 			}
 		}
 
-		/*
-		 * If we don't find our reference, the page split and our home
-		 * pointer references the wrong page. When internal pages
-		 * split, their WT_REF structure home values are updated; yield
-		 * and wait for that to happen.
-		 */
-		__wt_yield();
 	}
 
 found:	WT_ASSERT(session, pindex->index[slot] == ref);
 	*pindexp = pindex;
 	*slotp = slot;
+
+	WT_STAT_CONN_INCRV(session, page_index_slot_blocked, yield_count);
 }
 
 /*
@@ -177,12 +179,13 @@ __ref_descend_prev(
     WT_SESSION_IMPL *session, WT_REF *ref, WT_PAGE_INDEX **pindexp)
 {
 	WT_PAGE_INDEX *pindex;
+	uint64_t yield_count;
 
 	/*
 	 * We're passed a child page into which we're descending, and on which
 	 * we have a hazard pointer.
 	 */
-	for (;; __wt_yield()) {
+	for (yield_count = 0;; yield_count++, __wt_yield()) {
 		/*
 		 * There's a split race when a cursor moving backwards through
 		 * the tree descends the tree. If we're splitting an internal
@@ -242,6 +245,7 @@ __ref_descend_prev(
 			break;
 	}
 	*pindexp = pindex;
+	WT_STAT_CONN_INCRV(session, tree_descend_blocked, yield_count);
 }
 
 /*
@@ -497,29 +501,21 @@ restart:	/*
 			}
 
 			/*
-			 * Optionally skip leaf pages: skip all leaf pages if
-			 * WT_READ_SKIP_LEAF is set, when the skip-leaf-count
-			 * variable is non-zero, skip some count of leaf pages.
-			 * If this page is disk-based, crack the cell to figure
-			 * out it's a leaf page without reading it.
+			 * Optionally skip leaf pages: when the skip-leaf-count
+			 * variable is non-zero, skip some count of leaf pages,
+			 * then take the next leaf page we can.
 			 *
-			 * If skipping some number of leaf pages, decrement the
-			 * count of pages to zero, and then take the next leaf
-			 * page we can. Be cautious around the page decrement,
-			 * if for some reason don't take this particular page,
-			 * we can take the next one, and, there are additional
-			 * tests/decrements when we're about to return a leaf
-			 * page.
+			 * The reason to do some of this work here (rather than
+			 * in our caller), is because we can look at the cell
+			 * and know it's a leaf page without reading it into
+			 * memory. If this page is disk-based, crack the cell
+			 * to figure out it's a leaf page without reading it.
 			 */
-			if (skipleafcntp != NULL || LF_ISSET(WT_READ_SKIP_LEAF))
-				if (__ref_is_leaf(ref)) {
-					if (LF_ISSET(WT_READ_SKIP_LEAF))
-						break;
-					if (*skipleafcntp > 0) {
-						--*skipleafcntp;
-						break;
-					}
-				}
+			if (skipleafcntp != NULL &&
+			    *skipleafcntp > 0 && __ref_is_leaf(ref)) {
+				--*skipleafcntp;
+				break;
+			}
 
 			ret = __wt_page_swap(session, couple, ref,
 			    WT_READ_NOTFOUND_OK | WT_READ_RESTART_OK | flags);
@@ -626,34 +622,18 @@ descend:			empty_internal = true;
 					    session, ref, &pindex);
 					slot = pindex->entries - 1;
 				}
-			} else {
-				/*
-				 * At the lowest tree level (considering a leaf
-				 * page), turn off the initial-descent state.
-				 * Descent race tests are different when moving
-				 * through the tree vs. the initial descent.
-				 */
-				initial_descent = false;
-
-				/*
-				 * Optionally skip leaf pages, the second half.
-				 * We didn't have an on-page cell to figure out
-				 * if it was a leaf page, we had to acquire the
-				 * hazard pointer and look at the page.
-				 */
-				if (skipleafcntp != NULL ||
-				    LF_ISSET(WT_READ_SKIP_LEAF)) {
-					if (LF_ISSET(WT_READ_SKIP_LEAF))
-						break;
-					if (*skipleafcntp > 0) {
-						--*skipleafcntp;
-						break;
-					}
-				}
-
-				*refp = ref;
-				goto done;
+				continue;
 			}
+
+			/*
+			 * The tree-walk restart code knows we return any leaf
+			 * page we acquire (never hazard-pointer coupling on
+			 * after acquiring a leaf page), and asserts no restart
+			 * happens while holding a leaf page. This page must be
+			 * returned to our caller.
+			 */
+			*refp = ref;
+			goto done;
 		}
 	}
 
@@ -690,8 +670,29 @@ __wt_tree_walk_count(WT_SESSION_IMPL *session,
  *	of leaf pages before returning.
  */
 int
-__wt_tree_walk_skip(WT_SESSION_IMPL *session,
-    WT_REF **refp, uint64_t *skipleafcntp, uint32_t flags)
+__wt_tree_walk_skip(
+    WT_SESSION_IMPL *session, WT_REF **refp, uint64_t *skipleafcntp)
 {
-	return (__tree_walk_internal(session, refp, NULL, skipleafcntp, flags));
+	/*
+	 * Optionally skip leaf pages, the second half. The tree-walk function
+	 * didn't have an on-page cell it could use to figure out if the page
+	 * was a leaf page or not, it had to acquire the hazard pointer and look
+	 * at the page. The tree-walk code never acquires a hazard pointer on a
+	 * leaf page without returning it, and it's not trivial to change that.
+	 * So, the tree-walk code returns all leaf pages here and we deal with
+	 * decrementing the count.
+	 */
+	do {
+		WT_RET(__tree_walk_internal(session, refp, NULL, skipleafcntp,
+		    WT_READ_NO_GEN | WT_READ_SKIP_INTL | WT_READ_WONT_NEED));
+
+		/*
+		 * The walk skipped internal pages, any page returned must be a
+		 * leaf page.
+		 */
+		if (*skipleafcntp > 0)
+			--*skipleafcntp;
+	} while (*skipleafcntp > 0);
+
+	return (0);
 }
diff --git a/src/third_party/wiredtiger/src/config/config_def.c b/src/third_party/wiredtiger/src/config/config_def.c
index f152fbacad4..c8a8c525751 100644
--- a/src/third_party/wiredtiger/src/config/config_def.c
+++ b/src/third_party/wiredtiger/src/config/config_def.c
@@ -733,6 +733,7 @@ static const WT_CONFIG_CHECK confchk_wiredtiger_open[] = {
 	{ "readonly", "boolean", NULL, NULL, NULL, 0 },
 	{ "session_max", "int", NULL, "min=1", NULL, 0 },
 	{ "session_scratch_max", "int", NULL, NULL, NULL, 0 },
+	{ "session_table_cache", "boolean", NULL, NULL, NULL, 0 },
 	{ "shared_cache", "category",
 	    NULL, NULL,
 	    confchk_wiredtiger_open_shared_cache_subconfigs, 5 },
@@ -820,6 +821,7 @@ static const WT_CONFIG_CHECK confchk_wiredtiger_open_all[] = {
 	{ "readonly", "boolean", NULL, NULL, NULL, 0 },
 	{ "session_max", "int", NULL, "min=1", NULL, 0 },
 	{ "session_scratch_max", "int", NULL, NULL, NULL, 0 },
+	{ "session_table_cache", "boolean", NULL, NULL, NULL, 0 },
 	{ "shared_cache", "category",
 	    NULL, NULL,
 	    confchk_wiredtiger_open_shared_cache_subconfigs, 5 },
@@ -904,6 +906,7 @@ static const WT_CONFIG_CHECK confchk_wiredtiger_open_basecfg[] = {
 	{ "readonly", "boolean", NULL, NULL, NULL, 0 },
 	{ "session_max", "int", NULL, "min=1", NULL, 0 },
 	{ "session_scratch_max", "int", NULL, NULL, NULL, 0 },
+	{ "session_table_cache", "boolean", NULL, NULL, NULL, 0 },
 	{ "shared_cache", "category",
 	    NULL, NULL,
 	    confchk_wiredtiger_open_shared_cache_subconfigs, 5 },
@@ -986,6 +989,7 @@ static const WT_CONFIG_CHECK confchk_wiredtiger_open_usercfg[] = {
 	{ "readonly", "boolean", NULL, NULL, NULL, 0 },
 	{ "session_max", "int", NULL, "min=1", NULL, 0 },
 	{ "session_scratch_max", "int", NULL, NULL, NULL, 0 },
+	{ "session_table_cache", "boolean", NULL, NULL, NULL, 0 },
 	{ "shared_cache", "category",
 	    NULL, NULL,
 	    confchk_wiredtiger_open_shared_cache_subconfigs, 5 },
@@ -1276,14 +1280,14 @@ static const WT_CONFIG_ENTRY config_entries[] = {
 	  "file_max=100MB,path=\".\",prealloc=true,recover=on,"
 	  "zero_fill=false),lsm_manager=(merge=true,worker_thread_max=4),"
 	  "lsm_merge=true,mmap=true,multiprocess=false,readonly=false,"
-	  "session_max=100,session_scratch_max=2MB,shared_cache=(chunk=10MB"
-	  ",name=,quota=0,reserve=0,size=500MB),statistics=none,"
-	  "statistics_log=(json=false,on_close=false,path=\".\",sources=,"
-	  "timestamp=\"%b %d %H:%M:%S\",wait=0),"
+	  "session_max=100,session_scratch_max=2MB,session_table_cache=true"
+	  ",shared_cache=(chunk=10MB,name=,quota=0,reserve=0,size=500MB),"
+	  "statistics=none,statistics_log=(json=false,on_close=false,"
+	  "path=\".\",sources=,timestamp=\"%b %d %H:%M:%S\",wait=0),"
 	  "transaction_sync=(enabled=false,method=fsync),"
 	  "use_environment=true,use_environment_priv=false,verbose=,"
 	  "write_through=",
-	  confchk_wiredtiger_open, 40
+	  confchk_wiredtiger_open, 41
 	},
 	{ "wiredtiger_open_all",
 	  "async=(enabled=false,ops_max=1024,threads=2),buffer_alignment=-1"
@@ -1300,14 +1304,14 @@ static const WT_CONFIG_ENTRY config_entries[] = {
 	  "file_max=100MB,path=\".\",prealloc=true,recover=on,"
 	  "zero_fill=false),lsm_manager=(merge=true,worker_thread_max=4),"
 	  "lsm_merge=true,mmap=true,multiprocess=false,readonly=false,"
-	  "session_max=100,session_scratch_max=2MB,shared_cache=(chunk=10MB"
-	  ",name=,quota=0,reserve=0,size=500MB),statistics=none,"
-	  "statistics_log=(json=false,on_close=false,path=\".\",sources=,"
-	  "timestamp=\"%b %d %H:%M:%S\",wait=0),"
+	  "session_max=100,session_scratch_max=2MB,session_table_cache=true"
+	  ",shared_cache=(chunk=10MB,name=,quota=0,reserve=0,size=500MB),"
+	  "statistics=none,statistics_log=(json=false,on_close=false,"
+	  "path=\".\",sources=,timestamp=\"%b %d %H:%M:%S\",wait=0),"
 	  "transaction_sync=(enabled=false,method=fsync),"
 	  "use_environment=true,use_environment_priv=false,verbose=,"
 	  "version=(major=0,minor=0),write_through=",
-	  confchk_wiredtiger_open_all, 41
+	  confchk_wiredtiger_open_all, 42
 	},
 	{ "wiredtiger_open_basecfg",
 	  "async=(enabled=false,ops_max=1024,threads=2),buffer_alignment=-1"
@@ -1323,12 +1327,13 @@ static const WT_CONFIG_ENTRY config_entries[] = {
 	  "path=\".\",prealloc=true,recover=on,zero_fill=false),"
 	  "lsm_manager=(merge=true,worker_thread_max=4),lsm_merge=true,"
 	  "mmap=true,multiprocess=false,readonly=false,session_max=100,"
-	  "session_scratch_max=2MB,shared_cache=(chunk=10MB,name=,quota=0,"
-	  "reserve=0,size=500MB),statistics=none,statistics_log=(json=false"
-	  ",on_close=false,path=\".\",sources=,timestamp=\"%b %d %H:%M:%S\""
-	  ",wait=0),transaction_sync=(enabled=false,method=fsync),verbose=,"
+	  "session_scratch_max=2MB,session_table_cache=true,"
+	  "shared_cache=(chunk=10MB,name=,quota=0,reserve=0,size=500MB),"
+	  "statistics=none,statistics_log=(json=false,on_close=false,"
+	  "path=\".\",sources=,timestamp=\"%b %d %H:%M:%S\",wait=0),"
+	  "transaction_sync=(enabled=false,method=fsync),verbose=,"
 	  "version=(major=0,minor=0),write_through=",
-	  confchk_wiredtiger_open_basecfg, 35
+	  confchk_wiredtiger_open_basecfg, 36
 	},
 	{ "wiredtiger_open_usercfg",
 	  "async=(enabled=false,ops_max=1024,threads=2),buffer_alignment=-1"
@@ -1344,12 +1349,13 @@ static const WT_CONFIG_ENTRY config_entries[] = {
 	  "path=\".\",prealloc=true,recover=on,zero_fill=false),"
 	  "lsm_manager=(merge=true,worker_thread_max=4),lsm_merge=true,"
 	  "mmap=true,multiprocess=false,readonly=false,session_max=100,"
-	  "session_scratch_max=2MB,shared_cache=(chunk=10MB,name=,quota=0,"
-	  "reserve=0,size=500MB),statistics=none,statistics_log=(json=false"
-	  ",on_close=false,path=\".\",sources=,timestamp=\"%b %d %H:%M:%S\""
-	  ",wait=0),transaction_sync=(enabled=false,method=fsync),verbose=,"
+	  "session_scratch_max=2MB,session_table_cache=true,"
+	  "shared_cache=(chunk=10MB,name=,quota=0,reserve=0,size=500MB),"
+	  "statistics=none,statistics_log=(json=false,on_close=false,"
+	  "path=\".\",sources=,timestamp=\"%b %d %H:%M:%S\",wait=0),"
+	  "transaction_sync=(enabled=false,method=fsync),verbose=,"
 	  "write_through=",
-	  confchk_wiredtiger_open_usercfg, 34
+	  confchk_wiredtiger_open_usercfg, 35
 	},
 	{ NULL, NULL, NULL, 0 }
 };
diff --git a/src/third_party/wiredtiger/src/conn/conn_api.c b/src/third_party/wiredtiger/src/conn/conn_api.c
index 68d45678965..103e4a68f04 100644
--- a/src/third_party/wiredtiger/src/conn/conn_api.c
+++ b/src/third_party/wiredtiger/src/conn/conn_api.c
@@ -1086,6 +1086,41 @@ err:	/*
 			WT_TRET(wt_session->close(wt_session, config));
 		}
 
+	/*
+	 * Perform a system-wide checkpoint so that all tables are consistent
+	 * with each other.  Do this before shutting down all the subsystems.
+	 * We have shut down all user sessions, but send in true for waiting
+	 * for internal races.
+	 */
+	if (!F_ISSET(conn, WT_CONN_IN_MEMORY | WT_CONN_READONLY)) {
+		s = NULL;
+		WT_TRET(__wt_open_internal_session(
+		    conn, "close_ckpt", true, 0, &s));
+		if (s != NULL) {
+			const char *checkpoint_cfg[] = {
+			    WT_CONFIG_BASE(session, WT_SESSION_checkpoint),
+			    NULL
+			};
+			wt_session = &s->iface;
+			WT_TRET(__wt_txn_checkpoint(s, checkpoint_cfg, true));
+
+			/*
+			 * Mark the metadata dirty so we flush it on close,
+			 * allowing recovery to be skipped.
+			 */
+			WT_WITH_DHANDLE(s, WT_SESSION_META_DHANDLE(s),
+			    __wt_tree_modify_set(s));
+
+			WT_TRET(wt_session->close(wt_session, config));
+		}
+	}
+
+	if (ret != 0) {
+		__wt_err(session, ret,
+		    "failure during close, disabling further writes");
+		F_SET(conn, WT_CONN_PANIC);
+	}
+
 	WT_TRET(__wt_connection_close(conn));
 
 	/* We no longer have a session, don't try to update it. */
@@ -2185,6 +2220,9 @@ wiredtiger_open(const char *home, WT_EVENT_HANDLER *event_handler,
 	WT_ERR(__wt_config_gets(session, cfg, "readonly", &cval));
 	if (cval.val)
 		F_SET(conn, WT_CONN_READONLY);
+	WT_ERR(__wt_config_gets(session, cfg, "session_table_cache", &cval));
+	if (cval.val)
+		F_SET(conn, WT_CONN_TABLE_CACHE);
 
 	/* Configure error messages so we get them right early. */
 	WT_ERR(__wt_config_gets(session, cfg, "error_prefix", &cval));
@@ -2475,8 +2513,15 @@ err:	/* Discard the scratch buffers. */
 		__wt_scr_discard(session);
 	__wt_scr_discard(&conn->dummy_session);
 
-	if (ret != 0)
+	if (ret != 0) {
+		/*
+		 * Set panic if we're returning the run recovery error so that
+		 * we don't try to checkpoint data handles.
+		 */
+		if (ret == WT_RUN_RECOVERY)
+			F_SET(conn, WT_CONN_PANIC);
 		WT_TRET(__wt_connection_close(conn));
+	}
 
 	return (ret);
 }
diff --git a/src/third_party/wiredtiger/src/conn/conn_dhandle.c b/src/third_party/wiredtiger/src/conn/conn_dhandle.c
index 1816e66b0b7..181f12ab2dd 100644
--- a/src/third_party/wiredtiger/src/conn/conn_dhandle.c
+++ b/src/third_party/wiredtiger/src/conn/conn_dhandle.c
@@ -317,6 +317,9 @@ __wt_conn_btree_open(
 	WT_ASSERT(session,
 	     !F_ISSET(S2C(session), WT_CONN_CLOSING_NO_MORE_OPENS));
 
+	/* Turn off eviction. */
+	WT_RET(__wt_evict_file_exclusive_on(session));
+
 	/*
 	 * If the handle is already open, it has to be closed so it can be
 	 * reopened with a new configuration.
@@ -330,11 +333,11 @@ __wt_conn_btree_open(
 	 * in the tree that can block the close.
 	 */
 	if (F_ISSET(dhandle, WT_DHANDLE_OPEN))
-		WT_RET(__wt_conn_btree_sync_and_close(session, false, false));
+		WT_ERR(__wt_conn_btree_sync_and_close(session, false, false));
 
 	/* Discard any previous configuration, set up the new configuration. */
 	__conn_btree_config_clear(session);
-	WT_RET(__conn_btree_config_set(session));
+	WT_ERR(__conn_btree_config_set(session));
 
 	/* Set any special flags on the handle. */
 	F_SET(btree, LF_MASK(WT_BTREE_SPECIAL_FLAGS));
@@ -374,6 +377,8 @@ __wt_conn_btree_open(
 err:		F_CLR(btree, WT_BTREE_SPECIAL_FLAGS);
 	}
 
+	__wt_evict_file_exclusive_off(session);
+
 	return (ret);
 }
 
@@ -577,7 +582,7 @@ __conn_dhandle_remove(WT_SESSION_IMPL *session, bool final)
 
 	WT_ASSERT(session,
 	    F_ISSET(session, WT_SESSION_LOCKED_HANDLE_LIST_WRITE));
-	WT_ASSERT(session, dhandle != conn->cache->evict_file_next);
+	WT_ASSERT(session, dhandle != conn->cache->walk_tree);
 
 	/* Check if the handle was reacquired by a session while we waited. */
 	if (!final &&
@@ -673,8 +678,8 @@ restart:
 			continue;
 
 		WT_WITH_DHANDLE(session, dhandle,
-		    WT_TRET(__wt_conn_dhandle_discard_single(
-		    session, true, F_ISSET(conn, WT_CONN_IN_MEMORY))));
+		    WT_TRET(__wt_conn_dhandle_discard_single(session, true,
+		    F_ISSET(conn, WT_CONN_IN_MEMORY | WT_CONN_PANIC))));
 		goto restart;
 	}
 
@@ -699,8 +704,8 @@ restart:
 	/* Close the metadata file handle. */
 	while ((dhandle = TAILQ_FIRST(&conn->dhqh)) != NULL)
 		WT_WITH_DHANDLE(session, dhandle,
-		    WT_TRET(__wt_conn_dhandle_discard_single(
-		    session, true, F_ISSET(conn, WT_CONN_IN_MEMORY))));
+		    WT_TRET(__wt_conn_dhandle_discard_single(session, true,
+		    F_ISSET(conn, WT_CONN_IN_MEMORY | WT_CONN_PANIC))));
 
 	return (ret);
 }
diff --git a/src/third_party/wiredtiger/src/conn/conn_log.c b/src/third_party/wiredtiger/src/conn/conn_log.c
index d2ed314fd2e..8b47d3b04c0 100644
--- a/src/third_party/wiredtiger/src/conn/conn_log.c
+++ b/src/third_party/wiredtiger/src/conn/conn_log.c
@@ -375,6 +375,7 @@ __log_file_server(void *arg)
 	WT_LOG *log;
 	WT_LSN close_end_lsn, min_lsn;
 	WT_SESSION_IMPL *session;
+	uint64_t yield_count;
 	uint32_t filenum;
 	bool locked;
 
@@ -382,6 +383,7 @@ __log_file_server(void *arg)
 	conn = S2C(session);
 	log = conn->log;
 	locked = false;
+	yield_count = 0;
 	while (F_ISSET(conn, WT_CONN_SERVER_LOG)) {
 		/*
 		 * If there is a log file to close, make sure any outstanding
@@ -512,6 +514,7 @@ __log_file_server(void *arg)
 				 * thread a chance to run and try again in
 				 * this case.
 				 */
+				yield_count++;
 				__wt_yield();
 				continue;
 			}
@@ -522,8 +525,9 @@ __log_file_server(void *arg)
 	}
 
 	if (0) {
-err:		__wt_err(session, ret, "log close server error");
+err:		WT_PANIC_MSG(session, ret, "log close server error");
 	}
+	WT_STAT_CONN_INCRV(session, log_server_sync_blocked, yield_count);
 	if (locked)
 		__wt_spin_unlock(session, &log->log_sync_lock);
 	return (WT_THREAD_RET_VALUE);
@@ -740,7 +744,8 @@ __log_wrlsn_server(void *arg)
 	WT_ERR(__wt_log_force_write(session, 1, NULL));
 	__wt_log_wrlsn(session, NULL);
 	if (0) {
-err:		__wt_err(session, ret, "log wrlsn server error");
+err:		WT_PANIC_MSG(session, ret, "log wrlsn server error");
+
 	}
 	return (WT_THREAD_RET_VALUE);
 }
@@ -757,7 +762,7 @@ __log_server(void *arg)
 	WT_DECL_RET;
 	WT_LOG *log;
 	WT_SESSION_IMPL *session;
-	uint64_t timediff;
+	uint64_t retry, timediff;
 	bool did_work, signalled;
 
 	session = arg;
@@ -783,6 +788,7 @@ __log_server(void *arg)
 	 * takes to sync out an earlier file.
 	 */
 	did_work = true;
+	retry = 0;
 	while (F_ISSET(conn, WT_CONN_SERVER_LOG)) {
 		/*
 		 * Slots depend on future activity.  Force out buffered
@@ -827,7 +833,24 @@ __log_server(void *arg)
 					ret = __log_archive_once(session, 0);
 					__wt_writeunlock(
 					    session, &log->log_archive_lock);
-					WT_ERR(ret);
+					/*
+					 * It is possible that an external
+					 * process on some systems may prevent
+					 * removal. If we get a permission
+					 * error, retry a few times.
+					 */
+					if (ret == EACCES &&
+					    retry < WT_RETRY_MAX) {
+						retry++;
+						ret = 0;
+					} else {
+						/*
+						 * Return the error if there is
+						 * one or reset on success.
+						 */
+						WT_ERR(ret);
+						retry = 0;
+					}
 				} else
 					__wt_verbose(session, WT_VERB_LOG,
 					    "log_archive: Blocked due to open "
@@ -844,7 +867,7 @@ __log_server(void *arg)
 	}
 
 	if (0) {
-err:		__wt_err(session, ret, "log server error");
+err:		WT_PANIC_MSG(session, ret, "log server error");
 	}
 	return (WT_THREAD_RET_VALUE);
 }
@@ -902,7 +925,7 @@ __wt_logmgr_create(WT_SESSION_IMPL *session, const char *cfg[])
 	WT_RET(__wt_cond_alloc(session, "log sync", &log->log_sync_cond));
 	WT_RET(__wt_cond_alloc(session, "log write", &log->log_write_cond));
 	WT_RET(__wt_log_open(session));
-	WT_RET(__wt_log_slot_init(session));
+	WT_RET(__wt_log_slot_init(session, true));
 
 	return (0);
 }
diff --git a/src/third_party/wiredtiger/src/conn/conn_open.c b/src/third_party/wiredtiger/src/conn/conn_open.c
index eb3c79422a0..649bfa7c81f 100644
--- a/src/third_party/wiredtiger/src/conn/conn_open.c
+++ b/src/third_party/wiredtiger/src/conn/conn_open.c
@@ -91,6 +91,7 @@ __wt_connection_close(WT_CONNECTION_IMPL *conn)
 		if (txn_global->oldest_id == txn_global->current &&
 		    txn_global->metadata_pinned == txn_global->current)
 			break;
+		WT_STAT_CONN_INCR(session, txn_release_blocked);
 		__wt_yield();
 	}
 
@@ -143,7 +144,7 @@ __wt_connection_close(WT_CONNECTION_IMPL *conn)
 	 * conditional because we allocate the log path so that printlog can
 	 * run without running logging or recovery.
 	 */
-	if (FLD_ISSET(conn->log_flags, WT_CONN_LOG_ENABLED) &&
+	if (ret == 0 && FLD_ISSET(conn->log_flags, WT_CONN_LOG_ENABLED) &&
 	    FLD_ISSET(conn->log_flags, WT_CONN_LOG_RECOVER_DONE))
 		WT_TRET(__wt_txn_checkpoint_log(
 		    session, true, WT_TXN_LOG_CKPT_STOP, NULL));
diff --git a/src/third_party/wiredtiger/src/cursor/cur_index.c b/src/third_party/wiredtiger/src/cursor/cur_index.c
index 6fc01c0421f..15c1271c992 100644
--- a/src/third_party/wiredtiger/src/cursor/cur_index.c
+++ b/src/third_party/wiredtiger/src/cursor/cur_index.c
@@ -382,7 +382,7 @@ __curindex_close(WT_CURSOR *cursor)
 	if (cindex->child != NULL)
 		WT_TRET(cindex->child->close(cindex->child));
 
-	__wt_schema_release_table(session, cindex->table);
+	WT_TRET(__wt_schema_release_table(session, cindex->table));
 	/* The URI is owned by the index. */
 	cursor->internal_uri = NULL;
 	WT_TRET(__wt_cursor_close(cursor));
@@ -485,7 +485,7 @@ __wt_curindex_open(WT_SESSION_IMPL *session,
 
 	if ((ret = __wt_schema_open_index(
 	    session, table, idxname, namesize, &idx)) != 0) {
-		__wt_schema_release_table(session, table);
+		WT_TRET(__wt_schema_release_table(session, table));
 		return (ret);
 	}
 	WT_RET(__wt_calloc_one(session, &cindex));
diff --git a/src/third_party/wiredtiger/src/cursor/cur_join.c b/src/third_party/wiredtiger/src/cursor/cur_join.c
index 80afaf798dc..3681d36b452 100644
--- a/src/third_party/wiredtiger/src/cursor/cur_join.c
+++ b/src/third_party/wiredtiger/src/cursor/cur_join.c
@@ -325,7 +325,7 @@ __curjoin_close(WT_CURSOR *cursor)
 
 	JOINABLE_CURSOR_API_CALL(cursor, session, close, NULL);
 
-	__wt_schema_release_table(session, cjoin->table);
+	WT_TRET(__wt_schema_release_table(session, cjoin->table));
 	/* This is owned by the table */
 	cursor->key_format = NULL;
 	if (cjoin->projection != NULL) {
diff --git a/src/third_party/wiredtiger/src/cursor/cur_table.c b/src/third_party/wiredtiger/src/cursor/cur_table.c
index f6855172e90..ffa9ca35926 100644
--- a/src/third_party/wiredtiger/src/cursor/cur_table.c
+++ b/src/third_party/wiredtiger/src/cursor/cur_table.c
@@ -786,7 +786,7 @@ __curtable_close(WT_CURSOR *cursor)
 	__wt_free(session, ctable->cg_cursors);
 	__wt_free(session, ctable->cg_valcopy);
 	__wt_free(session, ctable->idx_cursors);
-	__wt_schema_release_table(session, ctable->table);
+	WT_TRET(__wt_schema_release_table(session, ctable->table));
 	/* The URI is owned by the table. */
 	cursor->internal_uri = NULL;
 	WT_TRET(__wt_cursor_close(cursor));
@@ -942,7 +942,7 @@ __wt_curtable_open(WT_SESSION_IMPL *session,
 		ret = __wt_open_cursor(session,
 		    table->cgroups[0]->source, NULL, cfg, cursorp);
 
-		__wt_schema_release_table(session, table);
+		WT_TRET(__wt_schema_release_table(session, table));
 		if (ret == 0) {
 			/* Fix up the public URI to match what was passed in. */
 			cursor = *cursorp;
diff --git a/src/third_party/wiredtiger/src/evict/evict_lru.c b/src/third_party/wiredtiger/src/evict/evict_lru.c
index 26bbf9f679b..6e850f67b3e 100644
--- a/src/third_party/wiredtiger/src/evict/evict_lru.c
+++ b/src/third_party/wiredtiger/src/evict/evict_lru.c
@@ -17,7 +17,7 @@ static int  __evict_pass(WT_SESSION_IMPL *);
 static int  __evict_server(WT_SESSION_IMPL *, bool *);
 static int  __evict_tune_workers(WT_SESSION_IMPL *session);
 static int  __evict_walk(WT_SESSION_IMPL *, WT_EVICT_QUEUE *);
-static int  __evict_walk_file(
+static int  __evict_walk_tree(
     WT_SESSION_IMPL *, WT_EVICT_QUEUE *, u_int, u_int *);
 
 #define	WT_EVICT_HAS_WORKERS(s)				\
@@ -767,8 +767,10 @@ __evict_clear_walk(WT_SESSION_IMPL *session)
 	cache = S2C(session)->cache;
 
 	WT_ASSERT(session, F_ISSET(session, WT_SESSION_LOCKED_PASS));
-	if (session->dhandle == cache->evict_file_next)
-		cache->evict_file_next = NULL;
+	if (session->dhandle == cache->walk_tree) {
+		cache->walk_tree = NULL;
+		cache->walk_target = 0;
+	}
 
 	if ((ref = btree->evict_ref) == NULL)
 		return (0);
@@ -880,10 +882,8 @@ void
 __wt_evict_file_exclusive_off(WT_SESSION_IMPL *session)
 {
 	WT_BTREE *btree;
-	WT_CACHE *cache;
 
 	btree = S2BT(session);
-	cache = S2C(session)->cache;
 
 	/*
 	 * We have seen subtle bugs with multiple threads racing to turn
@@ -891,12 +891,26 @@ __wt_evict_file_exclusive_off(WT_SESSION_IMPL *session)
 	 */
 	WT_DIAGNOSTIC_YIELD;
 
-	/* Hold the walk lock to turn on eviction. */
-	__wt_spin_lock(session, &cache->evict_walk_lock);
-	WT_ASSERT(session,
-	    btree->evict_ref == NULL && btree->evict_disabled > 0);
-	--btree->evict_disabled;
-	__wt_spin_unlock(session, &cache->evict_walk_lock);
+	/*
+	 * Atomically decrement the evict-disabled count, without acquiring the
+	 * eviction walk-lock. We can't acquire that lock here because there's
+	 * a potential deadlock. When acquiring exclusive eviction access, we
+	 * acquire the eviction walk-lock and then the cache's pass-intr lock.
+	 * The current eviction implementation can hold the pass-intr lock and
+	 * call into this function (see WT-3303 for the details), which might
+	 * deadlock with another thread trying to get exclusive eviction access.
+	 */
+#if defined(HAVE_DIAGNOSTIC)
+	{
+	int32_t v;
+
+	WT_ASSERT(session, btree->evict_ref == NULL);
+	v = __wt_atomic_subi32(&btree->evict_disabled, 1);
+	WT_ASSERT(session, v >= 0);
+	}
+#else
+	(void)__wt_atomic_subi32(&btree->evict_disabled, 1);
+#endif
 }
 
 #define	EVICT_TUNE_BATCH	1	/* Max workers to add each period */
@@ -941,6 +955,13 @@ __evict_tune_workers(WT_SESSION_IMPL *session)
 	conn = S2C(session);
 	cache = conn->cache;
 
+	/*
+	 * If we have a fixed number of eviction threads, there is no value in
+	 * calculating if we should do any tuning.
+	 */
+       if (conn->evict_threads_max == conn->evict_threads_min)
+		return (0);
+
 	WT_ASSERT(session, conn->evict_threads.threads[0]->session == session);
 	pgs_evicted_cur = pgs_evicted_persec_cur = 0;
 
@@ -1371,19 +1392,22 @@ retry:	while (slot < max_entries) {
 			 * scan last time through.  If we don't have a saved
 			 * handle, start from the beginning of the list.
 			 */
-			if ((dhandle = cache->evict_file_next) != NULL)
-				cache->evict_file_next = NULL;
-			else
+			if ((dhandle = cache->walk_tree) != NULL)
+				cache->walk_tree = NULL;
+			else {
 				dhandle = TAILQ_FIRST(&conn->dhqh);
+				cache->walk_target = 0;
+			}
 		} else {
 			if (incr) {
 				WT_ASSERT(session, dhandle->session_inuse > 0);
 				(void)__wt_atomic_subi32(
 				    &dhandle->session_inuse, 1);
 				incr = false;
-				cache->evict_file_next = NULL;
+				cache->walk_tree = NULL;
 			}
 			dhandle = TAILQ_NEXT(dhandle, q);
+			cache->walk_target = 0;
 		}
 
 		/* If we reach the end of the list, we're done. */
@@ -1445,29 +1469,26 @@ retry:	while (slot < max_entries) {
 
 		/*
 		 * Re-check the "no eviction" flag, used to enforce exclusive
-		 * access when a handle is being closed. If not set, remember
-		 * the file to visit first, next loop.
+		 * access when a handle is being closed.
 		 *
 		 * Only try to acquire the lock and simply continue if we fail;
 		 * the lock is held while the thread turning off eviction clears
 		 * the tree's current eviction point, and part of the process is
 		 * waiting on this thread to acknowledge that action.
+		 *
+		 * If a handle is being discarded, it will still be marked open,
+		 * but won't have a root page.
 		 */
 		if (btree->evict_disabled == 0 &&
 		    !__wt_spin_trylock(session, &cache->evict_walk_lock)) {
-			if (btree->evict_disabled == 0) {
+			if (btree->evict_disabled == 0 &&
+			    btree->root.page != NULL) {
 				/*
-				 * Assert the handle has a root page: eviction
-				 * should have been locked out if the tree is
-				 * being discarded or the root page is changing.
-				 * As this has not always been the case, assert
-				 * to debug that change.
+				 * Remember the file to visit first, next loop.
 				 */
-				WT_ASSERT(session, btree->root.page != NULL);
-
-				cache->evict_file_next = dhandle;
+				cache->walk_tree = dhandle;
 				WT_WITH_DHANDLE(session, dhandle,
-				    ret = __evict_walk_file(
+				    ret = __evict_walk_tree(
 				    session, queue, max_entries, &slot));
 
 				WT_ASSERT(session, session->split_gen == 0);
@@ -1489,7 +1510,7 @@ retry:	while (slot < max_entries) {
 	 * candidates and we aren't finding more.
 	 */
 	if (slot < max_entries && (retries < 2 ||
-	    (retries < 10 &&
+	    (retries < WT_RETRY_MAX &&
 	    (slot == queue->evict_entries || slot > start_slot)))) {
 		start_slot = slot;
 		++retries;
@@ -1556,45 +1577,21 @@ __evict_push_candidate(WT_SESSION_IMPL *session,
 }
 
 /*
- * __evict_walk_file --
- *	Get a few page eviction candidates from a single underlying file.
+ * __evict_walk_target --
+ *	Calculate how many pages to queue for a given tree.
  */
-static int
-__evict_walk_file(WT_SESSION_IMPL *session,
-    WT_EVICT_QUEUE *queue, u_int max_entries, u_int *slotp)
+static uint32_t
+__evict_walk_target(
+    WT_SESSION_IMPL *session, WT_EVICT_QUEUE *queue, u_int max_entries)
 {
-	WT_BTREE *btree;
 	WT_CACHE *cache;
-	WT_CONNECTION_IMPL *conn;
-	WT_DECL_RET;
-	WT_EVICT_ENTRY *end, *evict, *start;
-	WT_PAGE *page;
-	WT_PAGE_MODIFY *mod;
-	WT_REF *ref;
-	WT_TXN_GLOBAL *txn_global;
-	uint64_t btree_inuse, bytes_per_slot, cache_inuse, min_pages;
-	uint64_t pages_seen, pages_queued, refs_walked;
-	uint32_t remaining_slots, total_slots, walk_flags;
+	uint64_t btree_inuse, bytes_per_slot, cache_inuse;
 	uint32_t target_pages_clean, target_pages_dirty, target_pages;
-	int internal_pages, restarts;
-	bool give_up, modified, urgent_queued;
+	uint32_t total_slots;
 
-	conn = S2C(session);
-	btree = S2BT(session);
-	cache = conn->cache;
-	txn_global = &conn->txn_global;
-	internal_pages = restarts = 0;
-	give_up = urgent_queued = false;
-
-	/*
-	 * Figure out how many slots to fill from this tree.
-	 * Note that some care is taken in the calculation to avoid overflow.
-	 */
-	start = queue->evict_queue + *slotp;
-	remaining_slots = max_entries - *slotp;
-	total_slots = max_entries - queue->evict_entries;
-	btree_inuse = cache_inuse = 0;
+	cache = S2C(session)->cache;
 	target_pages_clean = target_pages_dirty = 0;
+	total_slots = max_entries - queue->evict_entries;
 
 	/*
 	 * The number of times we should fill the queue by the end of
@@ -1640,26 +1637,16 @@ __evict_walk_file(WT_SESSION_IMPL *session,
 	    QUEUE_FILLS_PER_PASS;
 
 	/*
-	 * Randomly walk trees with a small fraction of the cache in case there
-	 * are so many trees that none of them use enough of the cache to be
-	 * allocated slots.
-	 *
-	 * The chance of walking a tree is equal to the chance that a random
-	 * byte in cache belongs to the tree, weighted by how many times we
-	 * want to fill queues during a pass through all the trees in cache.
+	 * Walk trees with a small fraction of the cache in case there are so
+	 * many trees that none of them use enough of the cache to be allocated
+	 * slots.  Only skip a tree if it has no bytes of interest.
 	 */
 	if (target_pages == 0) {
-		if (F_ISSET(cache, WT_CACHE_EVICT_CLEAN)) {
-			btree_inuse = __wt_btree_bytes_evictable(session);
-			cache_inuse = __wt_cache_bytes_inuse(cache);
-		} else {
-			btree_inuse = __wt_btree_dirty_leaf_inuse(session);
-			cache_inuse = __wt_cache_dirty_leaf_inuse(cache);
-		}
-		if (btree_inuse == 0 || cache_inuse == 0)
-			return (0);
-		if (__wt_random64(&session->rnd) % cache_inuse >
-		    btree_inuse * QUEUE_FILLS_PER_PASS)
+		btree_inuse = F_ISSET(cache, WT_CACHE_EVICT_CLEAN) ?
+		    __wt_btree_bytes_evictable(session) :
+		    __wt_btree_dirty_leaf_inuse(session);
+
+		if (btree_inuse == 0)
 			return (0);
 	}
 
@@ -1670,13 +1657,64 @@ __evict_walk_file(WT_SESSION_IMPL *session,
 	if (target_pages < MIN_PAGES_PER_TREE)
 		target_pages = MIN_PAGES_PER_TREE;
 
+	/* If the tree is dead, take a lot of pages.  */
+	if (F_ISSET(session->dhandle, WT_DHANDLE_DEAD))
+		target_pages *= 10;
+
+	return (target_pages);
+}
+
+/*
+ * __evict_walk_tree --
+ *	Get a few page eviction candidates from a single underlying file.
+ */
+static int
+__evict_walk_tree(WT_SESSION_IMPL *session,
+    WT_EVICT_QUEUE *queue, u_int max_entries, u_int *slotp)
+{
+	WT_BTREE *btree;
+	WT_CACHE *cache;
+	WT_CONNECTION_IMPL *conn;
+	WT_DECL_RET;
+	WT_EVICT_ENTRY *end, *evict, *start;
+	WT_PAGE *page;
+	WT_PAGE_MODIFY *mod;
+	WT_REF *ref;
+	WT_TXN_GLOBAL *txn_global;
+	uint64_t min_pages, pages_seen, pages_queued, refs_walked;
+	uint32_t remaining_slots, target_pages, walk_flags;
+	int internal_pages, restarts;
+	bool give_up, modified, urgent_queued;
+
+	conn = S2C(session);
+	btree = S2BT(session);
+	cache = conn->cache;
+	txn_global = &conn->txn_global;
+	internal_pages = restarts = 0;
+	give_up = urgent_queued = false;
+
 	/*
-	 * If the tree is dead or we're near the end of the queue, fill the
-	 * remaining slots.
+	 * Figure out how many slots to fill from this tree.
+	 * Note that some care is taken in the calculation to avoid overflow.
 	 */
-	if (F_ISSET(session->dhandle, WT_DHANDLE_DEAD) ||
-	    target_pages > remaining_slots)
+	start = queue->evict_queue + *slotp;
+	remaining_slots = max_entries - *slotp;
+	if (cache->walk_target != 0) {
+		WT_ASSERT(session, cache->walk_progress <= cache->walk_target);
+		target_pages = cache->walk_target - cache->walk_progress;
+	} else {
+		target_pages = cache->walk_target =
+		    __evict_walk_target(session, queue, max_entries);
+		cache->walk_progress = 0;
+	}
+
+	if (target_pages > remaining_slots)
 		target_pages = remaining_slots;
+
+	/* If we don't want any pages from this tree, move on. */
+	if (target_pages == 0)
+		return (0);
+
 	end = start + target_pages;
 
 	/*
@@ -1849,6 +1887,7 @@ fast:		/* If the page can't be evicted, give up. */
 			continue;
 		++evict;
 		++pages_queued;
+		++cache->walk_progress;
 
 		if (WT_PAGE_IS_INTERNAL(page))
 			++internal_pages;
@@ -1863,6 +1902,10 @@ fast:		/* If the page can't be evicted, give up. */
 	WT_STAT_CONN_INCRV(
 	    session, cache_eviction_pages_queued, (u_int)(evict - start));
 
+	__wt_verbose(session, WT_VERB_EVICTSERVER,
+	    "%s walk: seen %" PRIu64 ", queued %" PRIu64,
+	    session->dhandle->name, pages_seen, pages_queued);
+
 	/*
 	 * If we couldn't find the number of pages we were looking for, skip
 	 * the tree next time.
@@ -2445,14 +2488,23 @@ __wt_verbose_dump_cache(WT_SESSION_IMPL *session)
 	WT_CONNECTION_IMPL *conn;
 	WT_DATA_HANDLE *dhandle;
 	WT_DECL_RET;
+	u_int pct;
 	uint64_t total_bytes, total_dirty_bytes;
 
 	conn = S2C(session);
 	total_bytes = total_dirty_bytes = 0;
+	pct = 0;				/* [-Werror=uninitialized] */
 
 	WT_RET(__wt_msg(session, "%s", WT_DIVIDER));
 	WT_RET(__wt_msg(session, "cache dump"));
 
+	WT_RET(__wt_msg(session,
+	    "cache full: %s", __wt_cache_full(session) ? "yes" : "no"));
+	WT_RET(__wt_msg(session, "cache clean check: %s (%u%%)",
+	    __wt_eviction_clean_needed(session, &pct) ? "yes" : "no", pct));
+	WT_RET(__wt_msg(session, "cache dirty check: %s (%u%%)",
+	    __wt_eviction_dirty_needed(session, &pct) ? "yes" : "no", pct));
+
 	for (dhandle = NULL;;) {
 		WT_WITH_HANDLE_LIST_READ_LOCK(session,
 		    WT_DHANDLE_NEXT(session, dhandle, &conn->dhqh, q));
diff --git a/src/third_party/wiredtiger/src/evict/evict_page.c b/src/third_party/wiredtiger/src/evict/evict_page.c
index 85689efd0b1..70a16442dc4 100644
--- a/src/third_party/wiredtiger/src/evict/evict_page.c
+++ b/src/third_party/wiredtiger/src/evict/evict_page.c
@@ -156,7 +156,7 @@ __wt_evict(WT_SESSION_IMPL *session, WT_REF *ref, bool closing)
 	/* Update the reference and discard the page. */
 	if (__wt_ref_is_root(ref))
 		__wt_ref_out(session, ref);
-	else if ((clean_page && !LF_ISSET(WT_EVICT_IN_MEMORY)) || tree_dead)
+	else if ((clean_page && !F_ISSET(conn, WT_CONN_IN_MEMORY)) || tree_dead)
 		/*
 		 * Pages that belong to dead trees never write back to disk
 		 * and can't support page splits.
@@ -202,8 +202,8 @@ __evict_delete_ref(WT_SESSION_IMPL *session, WT_REF *ref, bool closing)
 		return (0);
 
 	/*
-	 * Avoid doing reverse splits when closing the file, it is
-	 * wasted work and some structure may already have been freed.
+	 * Avoid doing reverse splits when closing the file, it is wasted work
+	 * and some structures may have already been freed.
 	 */
 	if (!closing) {
 		parent = ref->home;
@@ -393,11 +393,13 @@ __evict_review(
     WT_SESSION_IMPL *session, WT_REF *ref, uint32_t *flagsp, bool closing)
 {
 	WT_CACHE *cache;
+	WT_CONNECTION_IMPL *conn;
 	WT_DECL_RET;
 	WT_PAGE *page;
 	uint32_t flags;
-	bool lookaside_retry, modified;
+	bool lookaside_retry, *lookaside_retryp, modified;
 
+	conn = S2C(session);
 	flags = WT_EVICTING;
 	*flagsp = flags;
 
@@ -453,7 +455,7 @@ __evict_review(
 	 * Clean pages can't be evicted when running in memory only. This
 	 * should be uncommon - we don't add clean pages to the queue.
 	 */
-	if (F_ISSET(S2C(session), WT_CONN_IN_MEMORY) && !modified && !closing)
+	if (F_ISSET(conn, WT_CONN_IN_MEMORY) && !modified && !closing)
 		return (EBUSY);
 
 	/* Check if the page can be evicted. */
@@ -519,11 +521,14 @@ __evict_review(
 	 * Additionally, if we aren't trying to free space in the cache, scrub
 	 * the page and keep it in memory.
 	 */
-	cache = S2C(session)->cache;
+	cache = conn->cache;
+	lookaside_retry = false;
+	lookaside_retryp = NULL;
+
 	if (closing)
 		LF_SET(WT_VISIBILITY_ERR);
 	else if (!WT_PAGE_IS_INTERNAL(page)) {
-		if (F_ISSET(S2C(session), WT_CONN_IN_MEMORY))
+		if (F_ISSET(conn, WT_CONN_IN_MEMORY))
 			LF_SET(WT_EVICT_IN_MEMORY |
 			    WT_EVICT_SCRUB | WT_EVICT_UPDATE_RESTORE);
 		else {
@@ -531,21 +536,26 @@ __evict_review(
 
 			if (F_ISSET(cache, WT_CACHE_EVICT_SCRUB))
 				LF_SET(WT_EVICT_SCRUB);
+
+			/*
+			 * Check if reconciliation suggests trying the
+			 * lookaside table.
+			 */
+			lookaside_retryp = &lookaside_retry;
 		}
 	}
 
 	/* Reconcile the page. */
-	ret = __wt_reconcile(session, ref, NULL, flags, &lookaside_retry);
+	ret = __wt_reconcile(session, ref, NULL, flags, lookaside_retryp);
 
 	/*
-	 * If reconciliation fails, eviction is stuck and reconciliation reports
-	 * it might succeed if we use the lookaside table (the page didn't have
-	 * uncommitted updates, it was not-yet-globally visible updates causing
-	 * the problem), configure reconciliation to write those updates to the
-	 * lookaside table, allowing the eviction of pages we'd otherwise have
-	 * to retain in cache to support older readers.
+	 * If reconciliation fails, eviction is stuck and reconciliation
+	 * reports it might succeed if we use the lookaside table, then
+	 * configure reconciliation to write those updates to the lookaside
+	 * table, allowing the eviction of pages we'd otherwise have to retain
+	 * in cache to support older readers.
 	 */
-	if (ret == EBUSY && __wt_cache_stuck(session) && lookaside_retry) {
+	if (ret == EBUSY && lookaside_retry && __wt_cache_stuck(session)) {
 		LF_CLR(WT_EVICT_SCRUB | WT_EVICT_UPDATE_RESTORE);
 		LF_SET(WT_EVICT_LOOKASIDE);
 		ret = __wt_reconcile(session, ref, NULL, flags, NULL);
diff --git a/src/third_party/wiredtiger/src/include/btmem.h b/src/third_party/wiredtiger/src/include/btmem.h
index d0b21b17965..e965724dffe 100644
--- a/src/third_party/wiredtiger/src/include/btmem.h
+++ b/src/third_party/wiredtiger/src/include/btmem.h
@@ -714,7 +714,7 @@ struct __wt_page {
  *	Related information for fast-delete, on-disk pages.
  */
 struct __wt_page_deleted {
-	uint64_t txnid;			/* Transaction ID */
+	volatile uint64_t txnid;	/* Transaction ID */
 
 	WT_UPDATE **update_list;	/* List of updates for abort */
 };
@@ -904,7 +904,7 @@ struct __wt_ikey {
  * list.
  */
 WT_PACKED_STRUCT_BEGIN(__wt_update)
-	uint64_t txnid;			/* update transaction */
+	volatile uint64_t txnid;	/* Transaction ID */
 
 	WT_UPDATE *next;		/* forward-linked list */
 
diff --git a/src/third_party/wiredtiger/src/include/btree.h b/src/third_party/wiredtiger/src/include/btree.h
index 28fe1b94b23..5b1b4d68976 100644
--- a/src/third_party/wiredtiger/src/include/btree.h
+++ b/src/third_party/wiredtiger/src/include/btree.h
@@ -134,34 +134,44 @@ struct __wt_btree {
 	WT_BM	*bm;			/* Block manager reference */
 	u_int	 block_header;		/* WT_PAGE_HEADER_BYTE_SIZE */
 
-	uint64_t checkpoint_gen;	/* Checkpoint generation */
-	uint64_t rec_max_txn;		/* Maximum txn seen (clean trees) */
 	uint64_t write_gen;		/* Write generation */
+	uint64_t rec_max_txn;		/* Maximum txn seen (clean trees) */
+	uint64_t checkpoint_gen;	/* Checkpoint generation */
+	volatile enum {
+		WT_CKPT_OFF, WT_CKPT_PREPARE, WT_CKPT_RUNNING
+	} checkpointing;		/* Checkpoint in progress */
 
 	uint64_t    bytes_inmem;	/* Cache bytes in memory. */
 	uint64_t    bytes_dirty_intl;	/* Bytes in dirty internal pages. */
 	uint64_t    bytes_dirty_leaf;	/* Bytes in dirty leaf pages. */
 
+	/*
+	 * We flush pages from the tree (in order to make checkpoint faster),
+	 * without a high-level lock.  To avoid multiple threads flushing at
+	 * the same time, lock the tree.
+	 */
+	WT_SPINLOCK	flush_lock;	/* Lock to flush the tree's pages */
+
+	/*
+	 * All of the following fields live at the end of the structure so it's
+	 * easier to clear everything but the fields that persist.
+	 */
+#define	WT_BTREE_CLEAR_SIZE	(offsetof(WT_BTREE, evict_ref))
+
+	/*
+	 * Eviction information is maintained in the btree handle, but owned by
+	 * eviction, not the btree code.
+	 */
 	WT_REF	   *evict_ref;		/* Eviction thread's location */
 	uint64_t    evict_priority;	/* Relative priority of cached pages */
 	u_int	    evict_walk_period;	/* Skip this many LRU walks */
 	u_int	    evict_walk_saved;	/* Saved walk skips for checkpoints */
 	u_int	    evict_walk_skips;	/* Number of walks skipped */
-	int	    evict_disabled;	/* Eviction disabled count */
+	int32_t	    evict_disabled;	/* Eviction disabled count */
+	bool	    evict_disabled_open;/* Eviction disabled on open */
 	volatile uint32_t evict_busy;	/* Count of threads in eviction */
 	int	    evict_start_type;	/* Start position for eviction walk
 					   (see WT_EVICT_WALK_START). */
-	enum {
-		WT_CKPT_OFF, WT_CKPT_PREPARE, WT_CKPT_RUNNING
-	} checkpointing;		/* Checkpoint in progress */
-
-	/*
-	 * We flush pages from the tree (in order to make checkpoint faster),
-	 * without a high-level lock.  To avoid multiple threads flushing at
-	 * the same time, lock the tree.
-	 */
-	WT_SPINLOCK	flush_lock;	/* Lock to flush the tree's pages */
-
 	/* Flags values up to 0xff are reserved for WT_DHANDLE_* */
 #define	WT_BTREE_ALLOW_SPLITS	0x000100 /* Allow splits, even with no evict */
 #define	WT_BTREE_BULK		0x000200 /* Bulk-load handle */
diff --git a/src/third_party/wiredtiger/src/include/cache.h b/src/third_party/wiredtiger/src/include/cache.h
index 04920c3585a..42a152e5f10 100644
--- a/src/third_party/wiredtiger/src/include/cache.h
+++ b/src/third_party/wiredtiger/src/include/cache.h
@@ -133,7 +133,8 @@ struct __wt_cache {
 	 */
 	WT_SPINLOCK evict_pass_lock;	/* Eviction pass lock */
 	WT_SESSION_IMPL *walk_session;	/* Eviction pass session */
-	WT_DATA_HANDLE *evict_file_next;/* LRU next file to search */
+	WT_DATA_HANDLE *walk_tree;	/* LRU walk current tree */
+	uint32_t walk_progress, walk_target;/* Progress in current tree */
 
 	WT_SPINLOCK evict_queue_lock;	/* Eviction current queue lock */
 	WT_EVICT_QUEUE evict_queues[WT_EVICT_QUEUE_MAX];
diff --git a/src/third_party/wiredtiger/src/include/connection.h b/src/third_party/wiredtiger/src/include/connection.h
index f74732684f5..74611de1131 100644
--- a/src/third_party/wiredtiger/src/include/connection.h
+++ b/src/third_party/wiredtiger/src/include/connection.h
@@ -314,9 +314,10 @@ struct __wt_connection_impl {
 #define	WT_CONN_LOG_ARCHIVE		0x01	/* Archive is enabled */
 #define	WT_CONN_LOG_ENABLED		0x02	/* Logging is enabled */
 #define	WT_CONN_LOG_EXISTED		0x04	/* Log files found */
-#define	WT_CONN_LOG_RECOVER_DONE	0x08	/* Recovery completed */
-#define	WT_CONN_LOG_RECOVER_ERR		0x10	/* Error if recovery required */
-#define	WT_CONN_LOG_ZERO_FILL		0x20	/* Manually zero files */
+#define	WT_CONN_LOG_RECOVER_DIRTY	0x08	/* Recovering unclean */
+#define	WT_CONN_LOG_RECOVER_DONE	0x10	/* Recovery completed */
+#define	WT_CONN_LOG_RECOVER_ERR		0x20	/* Error if recovery required */
+#define	WT_CONN_LOG_ZERO_FILL		0x40	/* Manually zero files */
 	uint32_t	 log_flags;	/* Global logging configuration */
 	WT_CONDVAR	*log_cond;	/* Log server wait mutex */
 	WT_SESSION_IMPL *log_session;	/* Log server session */
diff --git a/src/third_party/wiredtiger/src/include/extern.h b/src/third_party/wiredtiger/src/include/extern.h
index bf3279d0f94..3c45b290dd8 100644
--- a/src/third_party/wiredtiger/src/include/extern.h
+++ b/src/third_party/wiredtiger/src/include/extern.h
@@ -181,7 +181,7 @@ extern int __wt_verify_dsk_image(WT_SESSION_IMPL *session, const char *tag, cons
 extern int __wt_verify_dsk(WT_SESSION_IMPL *session, const char *tag, WT_ITEM *buf) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_tree_walk(WT_SESSION_IMPL *session, WT_REF **refp, uint32_t flags) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_tree_walk_count(WT_SESSION_IMPL *session, WT_REF **refp, uint64_t *walkcntp, uint32_t flags) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
-extern int __wt_tree_walk_skip(WT_SESSION_IMPL *session, WT_REF **refp, uint64_t *skipleafcntp, uint32_t flags) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
+extern int __wt_tree_walk_skip( WT_SESSION_IMPL *session, WT_REF **refp, uint64_t *skipleafcntp) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_col_modify(WT_SESSION_IMPL *session, WT_CURSOR_BTREE *cbt, uint64_t recno, WT_ITEM *value, WT_UPDATE *upd_arg, bool is_remove, bool exclusive) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_col_search(WT_SESSION_IMPL *session, uint64_t search_recno, WT_REF *leaf, WT_CURSOR_BTREE *cbt) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_row_leaf_keys(WT_SESSION_IMPL *session, WT_PAGE *page) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
@@ -369,6 +369,7 @@ extern int __wt_log_needs_recovery(WT_SESSION_IMPL *session, WT_LSN *ckp_lsn, bo
 extern void __wt_log_written_reset(WT_SESSION_IMPL *session);
 extern int __wt_log_get_all_files(WT_SESSION_IMPL *session, char ***filesp, u_int *countp, uint32_t *maxid, bool active_only) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_log_extract_lognum( WT_SESSION_IMPL *session, const char *name, uint32_t *id) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
+extern int __wt_log_reset(WT_SESSION_IMPL *session, uint32_t lognum) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_log_acquire(WT_SESSION_IMPL *session, uint64_t recsize, WT_LOGSLOT *slot) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_log_allocfile( WT_SESSION_IMPL *session, uint32_t lognum, const char *dest) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_log_remove(WT_SESSION_IMPL *session, const char *file_prefix, uint32_t lognum) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
@@ -405,7 +406,7 @@ extern int __wt_logop_row_truncate_print(WT_SESSION_IMPL *session, const uint8_t
 extern int __wt_txn_op_printlog(WT_SESSION_IMPL *session, const uint8_t **pp, const uint8_t *end, uint32_t flags) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern void __wt_log_slot_activate(WT_SESSION_IMPL *session, WT_LOGSLOT *slot);
 extern int __wt_log_slot_switch(WT_SESSION_IMPL *session, WT_MYSLOT *myslot, bool retry, bool forced, bool *did_work) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
-extern int __wt_log_slot_init(WT_SESSION_IMPL *session) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
+extern int __wt_log_slot_init(WT_SESSION_IMPL *session, bool alloc) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_log_slot_destroy(WT_SESSION_IMPL *session) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_log_slot_join(WT_SESSION_IMPL *session, uint64_t mysize, uint32_t flags, WT_MYSLOT *myslot) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int64_t __wt_log_slot_release(WT_SESSION_IMPL *session, WT_MYSLOT *myslot, int64_t size);
@@ -564,12 +565,13 @@ extern int __wt_schema_index_source(WT_SESSION_IMPL *session, WT_TABLE *table, c
 extern int __wt_schema_create( WT_SESSION_IMPL *session, const char *uri, const char *config) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_schema_drop(WT_SESSION_IMPL *session, const char *uri, const char *cfg[]) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_schema_get_table(WT_SESSION_IMPL *session, const char *name, size_t namelen, bool ok_incomplete, WT_TABLE **tablep) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
-extern void __wt_schema_release_table(WT_SESSION_IMPL *session, WT_TABLE *table);
+extern int __wt_schema_release_table(WT_SESSION_IMPL *session, WT_TABLE *table) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern void __wt_schema_destroy_colgroup(WT_SESSION_IMPL *session, WT_COLGROUP **colgroupp);
 extern int __wt_schema_destroy_index(WT_SESSION_IMPL *session, WT_INDEX **idxp) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_schema_destroy_table(WT_SESSION_IMPL *session, WT_TABLE **tablep) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_schema_remove_table(WT_SESSION_IMPL *session, WT_TABLE *table) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_schema_close_tables(WT_SESSION_IMPL *session) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
+extern int __wt_schema_sweep_tables(WT_SESSION_IMPL *session) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_schema_colgroup_name(WT_SESSION_IMPL *session, WT_TABLE *table, const char *cgname, size_t len, WT_ITEM *buf) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_schema_open_colgroups(WT_SESSION_IMPL *session, WT_TABLE *table) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_schema_open_index(WT_SESSION_IMPL *session, WT_TABLE *table, const char *idxname, size_t len, WT_INDEX **indexp) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
@@ -691,7 +693,6 @@ extern uint32_t __wt_rduppo2(uint32_t n, uint32_t po2);
 extern void __wt_random_init(WT_RAND_STATE volatile *rnd_state) WT_GCC_FUNC_DECL_ATTRIBUTE((visibility("default")));
 extern void __wt_random_init_seed( WT_SESSION_IMPL *session, WT_RAND_STATE volatile *rnd_state) WT_GCC_FUNC_DECL_ATTRIBUTE((visibility("default")));
 extern uint32_t __wt_random(WT_RAND_STATE volatile *rnd_state) WT_GCC_FUNC_DECL_ATTRIBUTE((visibility("default")));
-extern uint64_t __wt_random64(WT_RAND_STATE volatile *rnd_state) WT_GCC_FUNC_DECL_ATTRIBUTE((visibility("default")));
 extern int __wt_buf_grow_worker(WT_SESSION_IMPL *session, WT_ITEM *buf, size_t size) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_buf_fmt(WT_SESSION_IMPL *session, WT_ITEM *buf, const char *fmt, ...) WT_GCC_FUNC_DECL_ATTRIBUTE((format (printf, 3, 4))) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_buf_catfmt(WT_SESSION_IMPL *session, WT_ITEM *buf, const char *fmt, ...) WT_GCC_FUNC_DECL_ATTRIBUTE((format (printf, 3, 4))) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
@@ -732,6 +733,8 @@ extern int __wt_thread_group_create( WT_SESSION_IMPL *session, WT_THREAD_GROUP *
 extern int __wt_thread_group_destroy(WT_SESSION_IMPL *session, WT_THREAD_GROUP *group) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_thread_group_start_one( WT_SESSION_IMPL *session, WT_THREAD_GROUP *group, bool wait) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_thread_group_stop_one( WT_SESSION_IMPL *session, WT_THREAD_GROUP *group, bool wait) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
+extern void __wt_epoch(WT_SESSION_IMPL *session, struct timespec *tsp) WT_GCC_FUNC_DECL_ATTRIBUTE((visibility("default")));
+extern void __wt_seconds(WT_SESSION_IMPL *session, time_t *timep);
 extern void __wt_txn_release_snapshot(WT_SESSION_IMPL *session);
 extern void __wt_txn_get_snapshot(WT_SESSION_IMPL *session);
 extern int __wt_txn_update_oldest(WT_SESSION_IMPL *session, uint32_t flags) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
diff --git a/src/third_party/wiredtiger/src/include/extern_posix.h b/src/third_party/wiredtiger/src/include/extern_posix.h
index c0ed056c7b6..9e32e86e64c 100644
--- a/src/third_party/wiredtiger/src/include/extern_posix.h
+++ b/src/third_party/wiredtiger/src/include/extern_posix.h
@@ -28,5 +28,5 @@ extern int __wt_vsnprintf_len_incr( char *buf, size_t size, size_t *retsizep, co
 extern int __wt_thread_create(WT_SESSION_IMPL *session, wt_thread_t *tidret, WT_THREAD_CALLBACK(*func)(void *), void *arg) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_thread_join(WT_SESSION_IMPL *session, wt_thread_t tid) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_thread_id(char *buf, size_t buflen) WT_GCC_FUNC_DECL_ATTRIBUTE((visibility("default"))) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
-extern void __wt_epoch(WT_SESSION_IMPL *session, struct timespec *tsp) WT_GCC_FUNC_DECL_ATTRIBUTE((visibility("default")));
+extern void __wt_epoch_raw(WT_SESSION_IMPL *session, struct timespec *tsp);
 extern void __wt_yield(void) WT_GCC_FUNC_DECL_ATTRIBUTE((visibility("default")));
diff --git a/src/third_party/wiredtiger/src/include/extern_win.h b/src/third_party/wiredtiger/src/include/extern_win.h
index d548ee0b2ec..85db8175615 100644
--- a/src/third_party/wiredtiger/src/include/extern_win.h
+++ b/src/third_party/wiredtiger/src/include/extern_win.h
@@ -26,7 +26,7 @@ extern int __wt_vsnprintf_len_incr( char *buf, size_t size, size_t *retsizep, co
 extern int __wt_thread_create(WT_SESSION_IMPL *session, wt_thread_t *tidret, WT_THREAD_CALLBACK(*func)(void *), void *arg) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_thread_join(WT_SESSION_IMPL *session, wt_thread_t tid) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_thread_id(char *buf, size_t buflen) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
-extern void __wt_epoch(WT_SESSION_IMPL *session, struct timespec *tsp);
+extern void __wt_epoch_raw(WT_SESSION_IMPL *session, struct timespec *tsp);
 extern int __wt_to_utf16_string( WT_SESSION_IMPL *session, const char*utf8, WT_ITEM **outbuf) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern int __wt_to_utf8_string( WT_SESSION_IMPL *session, const wchar_t*wide, WT_ITEM **outbuf) WT_GCC_FUNC_DECL_ATTRIBUTE((warn_unused_result));
 extern DWORD __wt_getlasterror(void);
diff --git a/src/third_party/wiredtiger/src/include/flags.h b/src/third_party/wiredtiger/src/include/flags.h
index f26a45c68f5..0e6cf7657ed 100644
--- a/src/third_party/wiredtiger/src/include/flags.h
+++ b/src/third_party/wiredtiger/src/include/flags.h
@@ -21,7 +21,8 @@
 #define	WT_CONN_SERVER_LSM				0x00008000
 #define	WT_CONN_SERVER_STATISTICS			0x00010000
 #define	WT_CONN_SERVER_SWEEP				0x00020000
-#define	WT_CONN_WAS_BACKUP				0x00040000
+#define	WT_CONN_TABLE_CACHE				0x00040000
+#define	WT_CONN_WAS_BACKUP				0x00080000
 #define	WT_EVICTING					0x00000002
 #define	WT_EVICT_INMEM_SPLIT				0x00000004
 #define	WT_EVICT_IN_MEMORY				0x00000008
@@ -47,9 +48,8 @@
 #define	WT_READ_PREV					0x00000080
 #define	WT_READ_RESTART_OK				0x00000100
 #define	WT_READ_SKIP_INTL				0x00000200
-#define	WT_READ_SKIP_LEAF				0x00000400
-#define	WT_READ_TRUNCATE				0x00000800
-#define	WT_READ_WONT_NEED				0x00001000
+#define	WT_READ_TRUNCATE				0x00000400
+#define	WT_READ_WONT_NEED				0x00000800
 #define	WT_SESSION_CAN_WAIT				0x00000001
 #define	WT_SESSION_INTERNAL				0x00000002
 #define	WT_SESSION_LOCKED_CHECKPOINT			0x00000004
diff --git a/src/third_party/wiredtiger/src/include/misc.i b/src/third_party/wiredtiger/src/include/misc.i
index fad10f01103..eb99de3dcab 100644
--- a/src/third_party/wiredtiger/src/include/misc.i
+++ b/src/third_party/wiredtiger/src/include/misc.i
@@ -41,45 +41,6 @@ __wt_strdup(WT_SESSION_IMPL *session, const char *str, void *retp)
 }
 
 /*
- * __wt_seconds --
- *	Return the seconds since the Epoch.
- */
-static inline void
-__wt_seconds(WT_SESSION_IMPL *session, time_t *timep)
-{
-	struct timespec t;
-
-	__wt_epoch(session, &t);
-
-	*timep = t.tv_sec;
-}
-
-/*
- * __wt_time_check_monotonic --
- *	Check and prevent time running backward.  If we detect that it has, we
- *	set the time structure to the previous values, making time stand still
- *	until we see a time in the future of the highest value seen so far.
- */
-static inline void
-__wt_time_check_monotonic(WT_SESSION_IMPL *session, struct timespec *tsp)
-{
-	/*
-	 * Detect time going backward.  If so, use the last
-	 * saved timestamp.
-	 */
-	if (session == NULL)
-		return;
-
-	if (tsp->tv_sec < session->last_epoch.tv_sec ||
-	     (tsp->tv_sec == session->last_epoch.tv_sec &&
-	     tsp->tv_nsec < session->last_epoch.tv_nsec)) {
-		WT_STAT_CONN_INCR(session, time_travel);
-		*tsp = session->last_epoch;
-	} else
-		session->last_epoch = *tsp;
-}
-
-/*
  * __wt_verbose --
  * 	Verbose message.
  *
diff --git a/src/third_party/wiredtiger/src/include/os.h b/src/third_party/wiredtiger/src/include/os.h
index 73d89268392..2c03b115e7c 100644
--- a/src/third_party/wiredtiger/src/include/os.h
+++ b/src/third_party/wiredtiger/src/include/os.h
@@ -34,9 +34,11 @@
 		(ret) = __wt_errno();					\
 } while (0)
 
+#define	WT_RETRY_MAX	10
+
 #define	WT_SYSCALL_RETRY(call, ret) do {				\
 	int __retry;							\
-	for (__retry = 0; __retry < 10; ++__retry) {			\
+	for (__retry = 0; __retry < WT_RETRY_MAX; ++__retry) {		\
 		WT_SYSCALL(call, ret);					\
 		switch (ret) {						\
 		case EAGAIN:						\
diff --git a/src/third_party/wiredtiger/src/include/session.h b/src/third_party/wiredtiger/src/include/session.h
index 0074b4a9d43..d05dee68641 100644
--- a/src/third_party/wiredtiger/src/include/session.h
+++ b/src/third_party/wiredtiger/src/include/session.h
@@ -98,6 +98,12 @@ struct __wt_session_impl {
 	 */
 	TAILQ_HEAD(__tables, __wt_table) tables;
 
+	/*
+	 * Updated when the table cache is swept of all tables older than the
+	 * current schema generation.
+	 */
+	uint64_t table_sweep_gen;
+
 	/* Current rwlock for callback. */
 	WT_RWLOCK *current_rwlock;
 	uint8_t current_rwticket;
diff --git a/src/third_party/wiredtiger/src/include/stat.h b/src/third_party/wiredtiger/src/include/stat.h
index db48a841571..01e622a5695 100644
--- a/src/third_party/wiredtiger/src/include/stat.h
+++ b/src/third_party/wiredtiger/src/include/stat.h
@@ -475,11 +475,19 @@ struct __wt_connection_stats {
 	int64_t thread_write_active;
 	int64_t application_evict_time;
 	int64_t application_cache_time;
+	int64_t txn_release_blocked;
+	int64_t conn_close_blocked_lsm;
+	int64_t dhandle_lock_blocked;
+	int64_t log_server_sync_blocked;
 	int64_t page_busy_blocked;
 	int64_t page_forcible_evict_blocked;
 	int64_t page_locked_blocked;
 	int64_t page_read_blocked;
 	int64_t page_sleep;
+	int64_t page_del_rollback_blocked;
+	int64_t child_modify_blocked_page;
+	int64_t page_index_slot_blocked;
+	int64_t tree_descend_blocked;
 	int64_t txn_snapshots_created;
 	int64_t txn_snapshots_dropped;
 	int64_t txn_begin;
diff --git a/src/third_party/wiredtiger/src/include/txn.h b/src/third_party/wiredtiger/src/include/txn.h
index 7e802c188ab..fdf9c714afa 100644
--- a/src/third_party/wiredtiger/src/include/txn.h
+++ b/src/third_party/wiredtiger/src/include/txn.h
@@ -93,6 +93,8 @@ struct __wt_txn_global {
 	 * the global transaction state.
 	 */
 	WT_RWLOCK scan_rwlock;
+	/* Protects logging, checkpoints and transaction visibility. */
+	WT_RWLOCK visibility_rwlock;
 
 	/*
 	 * Track information about the running checkpoint. The transaction
diff --git a/src/third_party/wiredtiger/src/include/txn.i b/src/third_party/wiredtiger/src/include/txn.i
index 314c948e4d1..39273a1995c 100644
--- a/src/third_party/wiredtiger/src/include/txn.i
+++ b/src/third_party/wiredtiger/src/include/txn.i
@@ -148,16 +148,6 @@ __wt_txn_oldest_id(WT_SESSION_IMPL *session)
 }
 
 /*
- * __wt_txn_committed --
- *	Return if a transaction has been committed.
- */
-static inline bool
-__wt_txn_committed(WT_SESSION_IMPL *session, uint64_t id)
-{
-	return (WT_TXNID_LT(id, S2C(session)->txn_global.last_running));
-}
-
-/*
  * __wt_txn_visible_all --
  *	Check if a given transaction ID is "globally visible".	This is, if
  *	all sessions in the system will see the transaction ID including the
diff --git a/src/third_party/wiredtiger/src/include/wiredtiger.in b/src/third_party/wiredtiger/src/include/wiredtiger.in
index 821efdf5fa1..30bff37017a 100644
--- a/src/third_party/wiredtiger/src/include/wiredtiger.in
+++ b/src/third_party/wiredtiger/src/include/wiredtiger.in
@@ -2458,6 +2458,8 @@ struct __wt_connection {
  * readonly for more information., a boolean flag; default \c false.}
  * @config{session_max, maximum expected number of sessions (including server
  * threads)., an integer greater than or equal to 1; default \c 100.}
+ * @config{session_table_cache, Maintain a per-session cache of tables., a
+ * boolean flag; default \c true.}
  * @config{shared_cache = (, shared cache configuration options.  A database
  * should configure either a cache_size or a shared_cache not both.  Enabling a
  * shared cache uses a session from the configured session_max., a set of
@@ -4806,72 +4808,94 @@ extern int wiredtiger_extension_terminate(WT_CONNECTION *connection);
 #define	WT_STAT_CONN_APPLICATION_EVICT_TIME		1216
 /*! thread-yield: application thread time waiting for cache (usecs) */
 #define	WT_STAT_CONN_APPLICATION_CACHE_TIME		1217
+/*!
+ * thread-yield: connection close blocked waiting for transaction state
+ * stabilization
+ */
+#define	WT_STAT_CONN_TXN_RELEASE_BLOCKED		1218
+/*! thread-yield: connection close yielded for lsm manager shutdown */
+#define	WT_STAT_CONN_CONN_CLOSE_BLOCKED_LSM		1219
+/*! thread-yield: data handle lock yielded */
+#define	WT_STAT_CONN_DHANDLE_LOCK_BLOCKED		1220
+/*! thread-yield: log server sync yielded for log write */
+#define	WT_STAT_CONN_LOG_SERVER_SYNC_BLOCKED		1221
 /*! thread-yield: page acquire busy blocked */
-#define	WT_STAT_CONN_PAGE_BUSY_BLOCKED			1218
+#define	WT_STAT_CONN_PAGE_BUSY_BLOCKED			1222
 /*! thread-yield: page acquire eviction blocked */
-#define	WT_STAT_CONN_PAGE_FORCIBLE_EVICT_BLOCKED	1219
+#define	WT_STAT_CONN_PAGE_FORCIBLE_EVICT_BLOCKED	1223
 /*! thread-yield: page acquire locked blocked */
-#define	WT_STAT_CONN_PAGE_LOCKED_BLOCKED		1220
+#define	WT_STAT_CONN_PAGE_LOCKED_BLOCKED		1224
 /*! thread-yield: page acquire read blocked */
-#define	WT_STAT_CONN_PAGE_READ_BLOCKED			1221
+#define	WT_STAT_CONN_PAGE_READ_BLOCKED			1225
 /*! thread-yield: page acquire time sleeping (usecs) */
-#define	WT_STAT_CONN_PAGE_SLEEP				1222
+#define	WT_STAT_CONN_PAGE_SLEEP				1226
+/*! thread-yield: page delete rollback yielded for instantiation */
+#define	WT_STAT_CONN_PAGE_DEL_ROLLBACK_BLOCKED		1227
+/*! thread-yield: page reconciliation yielded due to child modification */
+#define	WT_STAT_CONN_CHILD_MODIFY_BLOCKED_PAGE		1228
+/*! thread-yield: reference for page index and slot yielded */
+#define	WT_STAT_CONN_PAGE_INDEX_SLOT_BLOCKED		1229
+/*!
+ * thread-yield: tree descend one level yielded for split page index
+ * update
+ */
+#define	WT_STAT_CONN_TREE_DESCEND_BLOCKED		1230
 /*! transaction: number of named snapshots created */
-#define	WT_STAT_CONN_TXN_SNAPSHOTS_CREATED		1223
+#define	WT_STAT_CONN_TXN_SNAPSHOTS_CREATED		1231
 /*! transaction: number of named snapshots dropped */
-#define	WT_STAT_CONN_TXN_SNAPSHOTS_DROPPED		1224
+#define	WT_STAT_CONN_TXN_SNAPSHOTS_DROPPED		1232
 /*! transaction: transaction begins */
-#define	WT_STAT_CONN_TXN_BEGIN				1225
+#define	WT_STAT_CONN_TXN_BEGIN				1233
 /*! transaction: transaction checkpoint currently running */
-#define	WT_STAT_CONN_TXN_CHECKPOINT_RUNNING		1226
+#define	WT_STAT_CONN_TXN_CHECKPOINT_RUNNING		1234
 /*! transaction: transaction checkpoint generation */
-#define	WT_STAT_CONN_TXN_CHECKPOINT_GENERATION		1227
+#define	WT_STAT_CONN_TXN_CHECKPOINT_GENERATION		1235
 /*! transaction: transaction checkpoint max time (msecs) */
-#define	WT_STAT_CONN_TXN_CHECKPOINT_TIME_MAX		1228
+#define	WT_STAT_CONN_TXN_CHECKPOINT_TIME_MAX		1236
 /*! transaction: transaction checkpoint min time (msecs) */
-#define	WT_STAT_CONN_TXN_CHECKPOINT_TIME_MIN		1229
+#define	WT_STAT_CONN_TXN_CHECKPOINT_TIME_MIN		1237
 /*! transaction: transaction checkpoint most recent time (msecs) */
-#define	WT_STAT_CONN_TXN_CHECKPOINT_TIME_RECENT		1230
+#define	WT_STAT_CONN_TXN_CHECKPOINT_TIME_RECENT		1238
 /*! transaction: transaction checkpoint scrub dirty target */
-#define	WT_STAT_CONN_TXN_CHECKPOINT_SCRUB_TARGET	1231
+#define	WT_STAT_CONN_TXN_CHECKPOINT_SCRUB_TARGET	1239
 /*! transaction: transaction checkpoint scrub time (msecs) */
-#define	WT_STAT_CONN_TXN_CHECKPOINT_SCRUB_TIME		1232
+#define	WT_STAT_CONN_TXN_CHECKPOINT_SCRUB_TIME		1240
 /*! transaction: transaction checkpoint total time (msecs) */
-#define	WT_STAT_CONN_TXN_CHECKPOINT_TIME_TOTAL		1233
+#define	WT_STAT_CONN_TXN_CHECKPOINT_TIME_TOTAL		1241
 /*! transaction: transaction checkpoints */
-#define	WT_STAT_CONN_TXN_CHECKPOINT			1234
+#define	WT_STAT_CONN_TXN_CHECKPOINT			1242
 /*!
  * transaction: transaction checkpoints skipped because database was
  * clean
  */
-#define	WT_STAT_CONN_TXN_CHECKPOINT_SKIPPED		1235
+#define	WT_STAT_CONN_TXN_CHECKPOINT_SKIPPED		1243
 /*! transaction: transaction failures due to cache overflow */
-#define	WT_STAT_CONN_TXN_FAIL_CACHE			1236
+#define	WT_STAT_CONN_TXN_FAIL_CACHE			1244
 /*!
  * transaction: transaction fsync calls for checkpoint after allocating
  * the transaction ID
  */
-#define	WT_STAT_CONN_TXN_CHECKPOINT_FSYNC_POST		1237
+#define	WT_STAT_CONN_TXN_CHECKPOINT_FSYNC_POST		1245
 /*!
  * transaction: transaction fsync duration for checkpoint after
  * allocating the transaction ID (usecs)
  */
-#define	WT_STAT_CONN_TXN_CHECKPOINT_FSYNC_POST_DURATION	1238
+#define	WT_STAT_CONN_TXN_CHECKPOINT_FSYNC_POST_DURATION	1246
 /*! transaction: transaction range of IDs currently pinned */
-#define	WT_STAT_CONN_TXN_PINNED_RANGE			1239
+#define	WT_STAT_CONN_TXN_PINNED_RANGE			1247
 /*! transaction: transaction range of IDs currently pinned by a checkpoint */
-#define	WT_STAT_CONN_TXN_PINNED_CHECKPOINT_RANGE	1240
+#define	WT_STAT_CONN_TXN_PINNED_CHECKPOINT_RANGE	1248
 /*!
  * transaction: transaction range of IDs currently pinned by named
  * snapshots
  */
-#define	WT_STAT_CONN_TXN_PINNED_SNAPSHOT_RANGE		1241
+#define	WT_STAT_CONN_TXN_PINNED_SNAPSHOT_RANGE		1249
 /*! transaction: transaction sync calls */
-#define	WT_STAT_CONN_TXN_SYNC				1242
+#define	WT_STAT_CONN_TXN_SYNC				1250
 /*! transaction: transactions committed */
-#define	WT_STAT_CONN_TXN_COMMIT				1243
+#define	WT_STAT_CONN_TXN_COMMIT				1251
 /*! transaction: transactions rolled back */
-#define	WT_STAT_CONN_TXN_ROLLBACK			1244
+#define	WT_STAT_CONN_TXN_ROLLBACK			1252
 
 /*!
  * @}
diff --git a/src/third_party/wiredtiger/src/log/log.c b/src/third_party/wiredtiger/src/log/log.c
index 803d3e8dfab..da179400755 100644
--- a/src/third_party/wiredtiger/src/log/log.c
+++ b/src/third_party/wiredtiger/src/log/log.c
@@ -8,6 +8,7 @@
 
 #include "wt_internal.h"
 
+static int __log_newfile(WT_SESSION_IMPL *, bool, bool *);
 static int __log_openfile(
 	WT_SESSION_IMPL *, WT_FH **, const char *, uint32_t, uint32_t);
 static int __log_write_internal(
@@ -442,6 +443,59 @@ __wt_log_extract_lognum(
 }
 
 /*
+ * __wt_log_reset --
+ *	Reset the existing log file to after the given file number.
+ *	Called from recovery when toggling logging back on, it was off
+ *	the previous open but it was on earlier before that toggle.
+ */
+int
+__wt_log_reset(WT_SESSION_IMPL *session, uint32_t lognum)
+{
+	WT_CONNECTION_IMPL *conn;
+	WT_DECL_RET;
+	WT_LOG *log;
+	uint32_t old_lognum;
+	u_int i, logcount;
+	char **logfiles;
+
+	conn = S2C(session);
+	log = conn->log;
+
+	if (!FLD_ISSET(conn->log_flags, WT_CONN_LOG_ENABLED) ||
+	    log->fileid > lognum)
+		return (0);
+
+	WT_ASSERT(session, F_ISSET(conn, WT_CONN_RECOVERING));
+	WT_ASSERT(session, !F_ISSET(conn, WT_CONN_READONLY));
+	/*
+	 * We know we're single threaded and called from recovery only when
+	 * toggling logging back on.  Therefore the only log files we have are
+	 * old and outdated and the new one created when logging opened before
+	 * recovery.  We have to remove all old log files first and then create
+	 * the new one so that log file numbers are contiguous in the file
+	 * system.
+	 */
+	WT_RET(__wt_close(session, &log->log_fh));
+	WT_RET(__log_get_files(session,
+	    WT_LOG_FILENAME, &logfiles, &logcount));
+	for (i = 0; i < logcount; i++) {
+		WT_ERR(__wt_log_extract_lognum(
+		    session, logfiles[i], &old_lognum));
+		WT_ASSERT(session, old_lognum < lognum || lognum == 1);
+		WT_ERR(__wt_log_remove(session, WT_LOG_FILENAME, old_lognum));
+	}
+	log->fileid = lognum;
+
+	/* Send in true to update connection creation LSNs. */
+	WT_WITH_SLOT_LOCK(session, log,
+	    ret = __log_newfile(session, true, NULL));
+	WT_ERR(__wt_log_slot_init(session, false));
+err:	WT_TRET(
+	    __wt_fs_directory_list_free(session, &logfiles, logcount));
+	return (ret);
+}
+
+/*
  * __log_zero --
  *	Zero a log file.
  */
@@ -1611,11 +1665,6 @@ __wt_log_scan(WT_SESSION_IMPL *session, WT_LSN *lsnp, uint32_t flags,
 	if (func == NULL)
 		return (0);
 
-	if (LF_ISSET(WT_LOGSCAN_RECOVER))
-		__wt_verbose(session, WT_VERB_LOG,
-		    "__wt_log_scan truncating to %" PRIu32 "/%" PRIu32,
-		    log->trunc_lsn.l.file, log->trunc_lsn.l.offset);
-
 	if (log != NULL) {
 		allocsize = log->allocsize;
 
@@ -1709,9 +1758,14 @@ advance:
 			/*
 			 * Truncate this log file before we move to the next.
 			 */
-			if (LF_ISSET(WT_LOGSCAN_RECOVER))
+			if (LF_ISSET(WT_LOGSCAN_RECOVER) &&
+			    __wt_log_cmp(&rd_lsn, &log->trunc_lsn) < 0) {
+				__wt_verbose(session, WT_VERB_LOG,
+				    "Truncate end of log %" PRIu32 "/%" PRIu32,
+				    rd_lsn.l.file, rd_lsn.l.offset);
 				WT_ERR(__log_truncate(session,
 				    &rd_lsn, WT_LOG_FILENAME, 1));
+			}
 			/*
 			 * If we had a partial record, we'll want to break
 			 * now after closing and truncating.  Although for now
diff --git a/src/third_party/wiredtiger/src/log/log_slot.c b/src/third_party/wiredtiger/src/log/log_slot.c
index 97e317ce68c..b23d589c8e2 100644
--- a/src/third_party/wiredtiger/src/log/log_slot.c
+++ b/src/third_party/wiredtiger/src/log/log_slot.c
@@ -401,7 +401,7 @@ __wt_log_slot_switch(WT_SESSION_IMPL *session,
  *	Initialize the slot array.
  */
 int
-__wt_log_slot_init(WT_SESSION_IMPL *session)
+__wt_log_slot_init(WT_SESSION_IMPL *session, bool alloc)
 {
 	WT_CONNECTION_IMPL *conn;
 	WT_DECL_RET;
@@ -423,15 +423,17 @@ __wt_log_slot_init(WT_SESSION_IMPL *session)
 	 * switch log files very aggressively.  Scale back the buffer for
 	 * small log file sizes.
 	 */
-	log->slot_buf_size = (uint32_t)WT_MIN(
-	    (size_t)conn->log_file_max / 10, WT_LOG_SLOT_BUF_SIZE);
-	for (i = 0; i < WT_SLOT_POOL; i++) {
-		WT_ERR(__wt_buf_init(session,
-		    &log->slot_pool[i].slot_buf, log->slot_buf_size));
-		F_SET(&log->slot_pool[i], WT_SLOT_INIT_FLAGS);
+	if (alloc) {
+		log->slot_buf_size = (uint32_t)WT_MIN(
+		    (size_t)conn->log_file_max / 10, WT_LOG_SLOT_BUF_SIZE);
+		for (i = 0; i < WT_SLOT_POOL; i++) {
+			WT_ERR(__wt_buf_init(session,
+			    &log->slot_pool[i].slot_buf, log->slot_buf_size));
+			F_SET(&log->slot_pool[i], WT_SLOT_INIT_FLAGS);
+		}
+		WT_STAT_CONN_SET(session,
+		    log_buffer_size, log->slot_buf_size * WT_SLOT_POOL);
 	}
-	WT_STAT_CONN_SET(session,
-	    log_buffer_size, log->slot_buf_size * WT_SLOT_POOL);
 	/*
 	 * Set up the available slot from the pool the first time.
 	 */
diff --git a/src/third_party/wiredtiger/src/lsm/lsm_manager.c b/src/third_party/wiredtiger/src/lsm/lsm_manager.c
index b7d9086d10e..62da094b5f7 100644
--- a/src/third_party/wiredtiger/src/lsm/lsm_manager.c
+++ b/src/third_party/wiredtiger/src/lsm/lsm_manager.c
@@ -295,8 +295,10 @@ __wt_lsm_manager_destroy(WT_SESSION_IMPL *session)
 	    manager->lsm_workers == 0);
 	if (manager->lsm_workers > 0) {
 		/* Wait for the main LSM manager thread to finish. */
-		while (!F_ISSET(manager, WT_LSM_MANAGER_SHUTDOWN))
+		while (!F_ISSET(manager, WT_LSM_MANAGER_SHUTDOWN)) {
+			WT_STAT_CONN_INCR(session, conn_close_blocked_lsm);
 			__wt_yield();
+		}
 
 		/* Clean up open LSM handles. */
 		ret = __wt_lsm_tree_close_all(session);
diff --git a/src/third_party/wiredtiger/src/meta/meta_table.c b/src/third_party/wiredtiger/src/meta/meta_table.c
index aca69d0e6a2..a970694b9d7 100644
--- a/src/third_party/wiredtiger/src/meta/meta_table.c
+++ b/src/third_party/wiredtiger/src/meta/meta_table.c
@@ -230,12 +230,23 @@ __wt_metadata_remove(WT_SESSION_IMPL *session, const char *key)
 		WT_RET_MSG(session, EINVAL,
 		    "%s: remove not supported on the turtle file", key);
 
+	/*
+	 * Take, release, and reacquire the metadata cursor. It's complicated,
+	 * but that way the underlying meta-tracking function doesn't have to
+	 * open a second metadata cursor, it can use the session's cached one.
+	 */
 	WT_RET(__wt_metadata_cursor(session, &cursor));
 	cursor->set_key(cursor, key);
 	WT_ERR(cursor->search(cursor));
+	WT_ERR(__wt_metadata_cursor_release(session, &cursor));
+
 	if (WT_META_TRACKING(session))
 		WT_ERR(__wt_meta_track_update(session, key));
-	WT_ERR(cursor->remove(cursor));
+
+	WT_ERR(__wt_metadata_cursor(session, &cursor));
+	cursor->set_key(cursor, key);
+	ret = cursor->remove(cursor);
+
 err:	WT_TRET(__wt_metadata_cursor_release(session, &cursor));
 	return (ret);
 }
@@ -266,7 +277,9 @@ __wt_metadata_search(WT_SESSION_IMPL *session, const char *key, char **valuep)
 		 * that Coverity complains a lot, add an error check to get some
 		 * peace and quiet.
 		 */
-		if ((ret = __wt_turtle_read(session, key, valuep)) != 0)
+		WT_WITH_TURTLE_LOCK(session,
+		    ret = __wt_turtle_read(session, key, valuep));
+		if (ret != 0)
 			__wt_free(session, *valuep);
 		return (ret);
 	}
diff --git a/src/third_party/wiredtiger/src/meta/meta_turtle.c b/src/third_party/wiredtiger/src/meta/meta_turtle.c
index 5a089471059..f7ea6fe99c5 100644
--- a/src/third_party/wiredtiger/src/meta/meta_turtle.c
+++ b/src/third_party/wiredtiger/src/meta/meta_turtle.c
@@ -246,6 +246,9 @@ __wt_turtle_read(WT_SESSION_IMPL *session, const char *key, char **valuep)
 
 	*valuep = NULL;
 
+	/* Require single-threading. */
+	WT_ASSERT(session, F_ISSET(session, WT_SESSION_LOCKED_TURTLE));
+
 	/*
 	 * Open the turtle file; there's one case where we won't find the turtle
 	 * file, yet still succeed.  We create the metadata file before creating
@@ -302,6 +305,9 @@ __wt_turtle_update(WT_SESSION_IMPL *session, const char *key, const char *value)
 
 	fs = NULL;
 
+	/* Require single-threading. */
+	WT_ASSERT(session, F_ISSET(session, WT_SESSION_LOCKED_TURTLE));
+
 	/*
 	 * Create the turtle setup file: we currently re-write it from scratch
 	 * every time.
diff --git a/src/third_party/wiredtiger/src/os_posix/os_mtx_cond.c b/src/third_party/wiredtiger/src/os_posix/os_mtx_cond.c
index fe010b62305..e4a6683dee9 100644
--- a/src/third_party/wiredtiger/src/os_posix/os_mtx_cond.c
+++ b/src/third_party/wiredtiger/src/os_posix/os_mtx_cond.c
@@ -19,11 +19,19 @@ __wt_cond_alloc(WT_SESSION_IMPL *session, const char *name, WT_CONDVAR **condp)
 	WT_DECL_RET;
 
 	WT_RET(__wt_calloc_one(session, &cond));
-
 	WT_ERR(pthread_mutex_init(&cond->mtx, NULL));
 
-	/* Initialize the condition variable to permit self-blocking. */
+#ifdef HAVE_PTHREAD_COND_MONOTONIC
+	{
+	pthread_condattr_t condattr;
+
+	WT_ERR(pthread_condattr_init(&condattr));
+	WT_ERR(pthread_condattr_setclock(&condattr, CLOCK_MONOTONIC));
+	WT_ERR(pthread_cond_init(&cond->cond, &condattr));
+	}
+#else
 	WT_ERR(pthread_cond_init(&cond->cond, NULL));
+#endif
 
 	cond->name = name;
 	cond->waiters = 0;
@@ -79,7 +87,26 @@ __wt_cond_wait_signal(WT_SESSION_IMPL *session, WT_CONDVAR *cond,
 		goto skipping;
 
 	if (usecs > 0) {
-		__wt_epoch(session, &ts);
+		/*
+		 * Get the current time as the basis for calculating when the
+		 * wait should end.  Prefer a monotonic clock source to avoid
+		 * unexpectedly long sleeps when the system clock is adjusted.
+		 *
+		 * Failing that, query the time directly and don't attempt to
+		 * correct for the clock moving backwards, which would result
+		 * in a sleep that is too long by however much the clock is
+		 * updated.  This isn't as good as a monotonic clock source but
+		 * makes the window of vulnerability smaller (i.e., the
+		 * calculated time is only incorrect if the system clock
+		 * changes in between us querying it and waiting).
+		 */
+#ifdef HAVE_PTHREAD_COND_MONOTONIC
+		WT_SYSCALL_RETRY(clock_gettime(CLOCK_MONOTONIC, &ts), ret);
+		if (ret != 0)
+			WT_PANIC_MSG(session, ret, "clock_gettime");
+#else
+		__wt_epoch_raw(session, &ts);
+#endif
 		ts.tv_sec += (time_t)
 		    (((uint64_t)ts.tv_nsec + WT_THOUSAND * usecs) / WT_BILLION);
 		ts.tv_nsec = (long)
diff --git a/src/third_party/wiredtiger/src/os_posix/os_time.c b/src/third_party/wiredtiger/src/os_posix/os_time.c
index fe337fea7cf..25a08d62355 100644
--- a/src/third_party/wiredtiger/src/os_posix/os_time.c
+++ b/src/third_party/wiredtiger/src/os_posix/os_time.c
@@ -9,14 +9,12 @@
 #include "wt_internal.h"
 
 /*
- * __wt_epoch --
- *	Return the time since the Epoch.
+ * __wt_epoch_raw --
+ *	Return the time since the Epoch as reported by a system call.
  */
 void
-__wt_epoch(WT_SESSION_IMPL *session, struct timespec *tsp)
-    WT_GCC_FUNC_ATTRIBUTE((visibility("default")))
+__wt_epoch_raw(WT_SESSION_IMPL *session, struct timespec *tsp)
 {
-	struct timespec tmp;
 	WT_DECL_RET;
 
 	/*
@@ -28,19 +26,10 @@ __wt_epoch(WT_SESSION_IMPL *session, struct timespec *tsp)
 	tsp->tv_sec = 0;
 	tsp->tv_nsec = 0;
 
-	/*
-	 * Read into a local variable so that we're comparing the correct
-	 * value when we check for monotonic increasing time.  There are
-	 * many places we read into an unlocked global variable.
-	 */
 #if defined(HAVE_CLOCK_GETTIME)
-	WT_SYSCALL_RETRY(clock_gettime(CLOCK_REALTIME, &tmp), ret);
-	if (ret == 0) {
-		__wt_time_check_monotonic(session, &tmp);
-		tsp->tv_sec = tmp.tv_sec;
-		tsp->tv_nsec = tmp.tv_nsec;
+	WT_SYSCALL_RETRY(clock_gettime(CLOCK_REALTIME, tsp), ret);
+	if (ret == 0)
 		return;
-	}
 	WT_PANIC_MSG(session, ret, "clock_gettime");
 #elif defined(HAVE_GETTIMEOFDAY)
 	{
@@ -48,10 +37,8 @@ __wt_epoch(WT_SESSION_IMPL *session, struct timespec *tsp)
 
 	WT_SYSCALL_RETRY(gettimeofday(&v, NULL), ret);
 	if (ret == 0) {
-		tmp.tv_sec = v.tv_sec;
-		tmp.tv_nsec = v.tv_usec * WT_THOUSAND;
-		__wt_time_check_monotonic(session, &tmp);
-		*tsp = tmp;
+		tsp->tv_sec = v.tv_sec;
+		tsp->tv_nsec = v.tv_usec * WT_THOUSAND;
 		return;
 	}
 	WT_PANIC_MSG(session, ret, "gettimeofday");
diff --git a/src/third_party/wiredtiger/src/os_win/os_time.c b/src/third_party/wiredtiger/src/os_win/os_time.c
index ba71341ab22..84c06bed6e5 100644
--- a/src/third_party/wiredtiger/src/os_win/os_time.c
+++ b/src/third_party/wiredtiger/src/os_win/os_time.c
@@ -9,24 +9,23 @@
 #include "wt_internal.h"
 
 /*
- * __wt_epoch --
- *	Return the time since the Epoch.
+ * __wt_epoch_raw --
+ *	Return the time since the Epoch as reported by the system.
  */
 void
-__wt_epoch(WT_SESSION_IMPL *session, struct timespec *tsp)
+__wt_epoch_raw(WT_SESSION_IMPL *session, struct timespec *tsp)
 {
-	struct timespec tmp;
 	FILETIME time;
 	uint64_t ns100;
 
+	WT_UNUSED(session);
+
 	GetSystemTimeAsFileTime(&time);
 
 	ns100 = (((int64_t)time.dwHighDateTime << 32) + time.dwLowDateTime)
 	    - 116444736000000000LL;
-	tmp.tv_sec = ns100 / 10000000;
-	tmp.tv_nsec = (long)((ns100 % 10000000) * 100);
-	__wt_time_check_monotonic(session, &tmp);
-	*tsp = tmp;
+	tsp->tv_sec = ns100 / 10000000;
+	tsp->tv_nsec = (long)((ns100 % 10000000) * 100);
 }
 
 /*
diff --git a/src/third_party/wiredtiger/src/reconcile/rec_write.c b/src/third_party/wiredtiger/src/reconcile/rec_write.c
index e59d9796352..688efa10398 100644
--- a/src/third_party/wiredtiger/src/reconcile/rec_write.c
+++ b/src/third_party/wiredtiger/src/reconcile/rec_write.c
@@ -45,12 +45,14 @@ typedef struct {
 	uint64_t orig_btree_checkpoint_gen;
 	uint64_t orig_txn_checkpoint_gen;
 
-	/* Track the page's maximum transaction ID. */
+	/* Track the oldest transaction running when reconciliation starts. */
+	uint64_t last_running;
+
 	uint64_t max_txn;
 
-	/* Track if all updates were skipped. */
-	uint64_t update_cnt;
-	uint64_t update_skip_cnt;
+	uint64_t update_mem_all;	/* Total update memory size */
+	uint64_t update_mem_saved;	/* Saved update memory size */
+	uint64_t update_mem_uncommitted;/* Uncommitted update memory size */
 
 	/*
 	 * When we can't mark the page clean (for example, checkpoint found some
@@ -338,7 +340,8 @@ static int  __rec_split_write(WT_SESSION_IMPL *,
 		WT_RECONCILE *, WT_BOUNDARY *, WT_ITEM *, bool);
 static int  __rec_update_las(
 		WT_SESSION_IMPL *, WT_RECONCILE *, uint32_t, WT_BOUNDARY *);
-static int  __rec_write_check_complete(WT_SESSION_IMPL *, WT_RECONCILE *);
+static int  __rec_write_check_complete(
+		WT_SESSION_IMPL *, WT_RECONCILE *, bool *);
 static int  __rec_write_init(WT_SESSION_IMPL *,
 		WT_REF *, uint32_t, WT_SALVAGE_COOKIE *, void *);
 static void __rec_write_page_status(WT_SESSION_IMPL *, WT_RECONCILE *);
@@ -437,7 +440,7 @@ __wt_reconcile(WT_SESSION_IMPL *session, WT_REF *ref,
 
 	/* Checks for a successful reconciliation. */
 	if (ret == 0)
-		ret = __rec_write_check_complete(session, r);
+		ret = __rec_write_check_complete(session, r, lookaside_retryp);
 
 	/* Wrap up the page reconciliation. */
 	if (ret == 0 && (ret = __rec_write_wrapup(session, r, page)) == 0)
@@ -448,14 +451,6 @@ __wt_reconcile(WT_SESSION_IMPL *session, WT_REF *ref,
 	/* Release the reconciliation lock. */
 	WT_PAGE_UNLOCK(session, page);
 
-	/*
-	 * If our caller can configure lookaside table reconciliation, flag if
-	 * that's worth trying. The lookaside table doesn't help if we skipped
-	 * updates, it can only help with older readers preventing eviction.
-	 */
-	if (lookaside_retryp != NULL && r->update_cnt == r->update_skip_cnt)
-		*lookaside_retryp = true;
-
 	/* Update statistics. */
 	WT_STAT_CONN_INCR(session, rec_pages);
 	WT_STAT_DATA_INCR(session, rec_pages);
@@ -558,13 +553,21 @@ __rec_las_checkpoint_test(WT_SESSION_IMPL *session, WT_RECONCILE *r)
 
 /*
  * __rec_write_check_complete --
- *	Check that reconciliation should complete
+ *	Check that reconciliation should complete.
  */
 static int
-__rec_write_check_complete(WT_SESSION_IMPL *session, WT_RECONCILE *r)
+__rec_write_check_complete(
+    WT_SESSION_IMPL *session, WT_RECONCILE *r, bool *lookaside_retryp)
 {
-	WT_BOUNDARY *bnd;
-	size_t i;
+	/*
+	 * Tests in this function are lookaside tests and tests to decide if
+	 * rewriting a page in memory is worth doing. In-memory configurations
+	 * can't use a lookaside table, and we ignore page rewrite desirability
+	 * checks for in-memory eviction because a small cache can force us to
+	 * rewrite every possible page.
+	 */
+	if (F_ISSET(r, WT_EVICT_IN_MEMORY))
+		return (0);
 
 	/*
 	 * If we have used the lookaside table, check for a lookaside table and
@@ -574,19 +577,62 @@ __rec_write_check_complete(WT_SESSION_IMPL *session, WT_RECONCILE *r)
 		return (EBUSY);
 
 	/*
-	 * If we are doing update/restore based eviction, confirm part of the
-	 * page is being discarded, or at least 10% of the updates won't have
-	 * to be re-instantiated. Otherwise, it isn't progress, don't bother.
+	 * Eviction can configure lookaside table reconciliation, consider if
+	 * it's worth giving up this reconciliation attempt and falling back to
+	 * using the lookaside table.  We continue with evict/restore if
+	 * switching to the lookaside doesn't make sense for any reason: we
+	 * won't retry an evict/restore reconciliation until/unless the
+	 * transactional system moves forward, so at worst it's a single wasted
+	 * effort.
+	 *
+	 * First, check if the lookaside table is a possible alternative.
 	 */
-	if (F_ISSET(r, WT_EVICT_UPDATE_RESTORE)) {
-		for (bnd = r->bnd, i = 0; i < r->bnd_entries; ++bnd, ++i)
-			if (bnd->supd == NULL)
-				break;
-		if (i == r->bnd_entries &&
-		    r->update_cnt / 10 >= r->update_skip_cnt)
-			return (EBUSY);
-	}
-	return (0);
+	if (lookaside_retryp == NULL)
+		return (0);
+
+	/*
+	 * We only suggest lookaside if currently in an evict/restore attempt
+	 * and some updates were saved.  Our caller sets the evict/restore flag
+	 * based on various conditions (like if this is a leaf page), which is
+	 * why we're testing that flag instead of a set of other conditions.
+	 * If no updates were saved, eviction will succeed without needing to
+	 * restore anything.
+	 */
+	if (!F_ISSET(r, WT_EVICT_UPDATE_RESTORE) || r->bnd->supd == NULL)
+		return (0);
+
+	/*
+	 * Check if this reconciliation attempt is making progress.  If there's
+	 * any sign of progress, don't fall back to the lookaside table.
+	 *
+	 * Check if the current reconciliation split, in which case we'll
+	 * likely get to write at least one of the blocks.  If that page is
+	 * empty, that's also progress.
+	 */
+	if (r->bnd_next != 1)
+		return (0);
+
+	/*
+	 * Check if the current reconciliation applied some updates, in which
+	 * case evict/restore should gain us some space.
+	 */
+	if (r->update_mem_saved != r->update_mem_all)
+		return (0);
+
+	/*
+	 * Check if lookaside eviction is possible.  If any of the updates we
+	 * saw were uncommitted, the lookaside table cannot be used: it only
+	 * helps with older readers preventing eviction.
+	 */
+	if (r->update_mem_uncommitted != 0)
+		return (0);
+
+	/*
+	 * The current evict/restore approach shows no signs of being useful,
+	 * lookaside is possible, suggest the lookaside table.
+	 */
+	*lookaside_retryp = true;
+	return (EBUSY);
 }
 
 /*
@@ -849,6 +895,16 @@ __rec_write_init(WT_SESSION_IMPL *session,
 	WT_ORDERED_READ(r->orig_write_gen, page->modify->write_gen);
 
 	/*
+	 * Cache the oldest running transaction ID.  This is used to check
+	 * whether updates seen by reconciliation have committed.  We keep a
+	 * cached copy to avoid races where a concurrent transaction could
+	 * abort while reconciliation is examining its updates.  This way, any
+	 * transaction running when reconciliation starts is considered
+	 * uncommitted.
+	 */
+	WT_ORDERED_READ(r->last_running, S2C(session)->txn_global.last_running);
+
+	/*
 	 * Lookaside table eviction is configured when eviction gets aggressive,
 	 * adjust the flags for cases we don't support.
 	 */
@@ -891,7 +947,7 @@ __rec_write_init(WT_SESSION_IMPL *session,
 	r->max_txn = WT_TXN_NONE;
 
 	/* Track if all updates were skipped. */
-	r->update_cnt = r->update_skip_cnt = 0;
+	r->update_mem_all = r->update_mem_saved = r->update_mem_uncommitted = 0;
 
 	/* Track if the page can be marked clean. */
 	r->leave_dirty = false;
@@ -1115,7 +1171,7 @@ __rec_txn_read(WT_SESSION_IMPL *session, WT_RECONCILE *r,
 	WT_DECL_ITEM(tmp);
 	WT_PAGE *page;
 	WT_UPDATE *append, *upd, *upd_list;
-	size_t notused;
+	size_t notused, update_mem;
 	uint64_t max_txn, min_txn, txnid;
 	bool append_origv, skipped;
 
@@ -1136,36 +1192,64 @@ __rec_txn_read(WT_SESSION_IMPL *session, WT_RECONCILE *r,
 	} else
 		upd_list = ins->upd;
 
-	++r->update_cnt;
-	for (skipped = false,
-	    max_txn = WT_TXN_NONE, min_txn = UINT64_MAX,
-	    upd = upd_list; upd != NULL; upd = upd->next) {
-		if ((txnid = upd->txnid) == WT_TXN_ABORTED)
-			continue;
+	skipped = false;
+	update_mem = 0;
+	max_txn = WT_TXN_NONE;
+	min_txn = UINT64_MAX;
 
-		/* Track the largest/smallest transaction IDs on the list. */
-		if (WT_TXNID_LT(max_txn, txnid))
-			max_txn = txnid;
-		if (WT_TXNID_LT(txnid, min_txn))
-			min_txn = txnid;
+	if (F_ISSET(r, WT_EVICTING)) {
+		/* Discard obsolete updates. */
+		if ((upd = __wt_update_obsolete_check(
+		    session, page, upd_list->next)) != NULL)
+			__wt_update_obsolete_free(session, page, upd);
+
+		for (upd = upd_list; upd != NULL; upd = upd->next) {
+			/* Track the total memory in the update chain. */
+			update_mem += WT_UPDATE_MEMSIZE(upd);
+
+			if ((txnid = upd->txnid) == WT_TXN_ABORTED)
+				continue;
 
-		/*
-		 * Find the first update we can use.
-		 */
-		if (F_ISSET(r, WT_EVICTING)) {
 			/*
+			 * Track the largest/smallest transaction IDs on the
+			 * list.
+			 */
+			if (WT_TXNID_LT(max_txn, txnid))
+				max_txn = txnid;
+			if (WT_TXNID_LT(txnid, min_txn))
+				min_txn = txnid;
+
+			/*
+			 * Find the first update we can use.
+			 *
 			 * Eviction can write any committed update.
 			 *
 			 * When reconciling for eviction, track whether any
 			 * uncommitted updates are found.
+			 *
+			 * When reconciling for eviction, track the memory held
+			 * by the update chain.
 			 */
-			if (__wt_txn_committed(session, txnid)) {
-				if (*updp == NULL)
-					*updp = upd;
-			} else
+			if (WT_TXNID_LE(r->last_running, txnid)) {
 				skipped = true;
-		} else {
+				continue;
+			}
+
+			if (*updp == NULL)
+				*updp = upd;
+		}
+	} else
+		for (upd = upd_list; upd != NULL; upd = upd->next) {
+			if ((txnid = upd->txnid) == WT_TXN_ABORTED)
+				continue;
+
+			/* Track the largest transaction ID on the list. */
+			if (WT_TXNID_LT(max_txn, txnid))
+				max_txn = txnid;
+
 			/*
+			 * Find the first update we can use.
+			 *
 			 * Checkpoint can only write updates visible as of its
 			 * snapshot.
 			 *
@@ -1180,7 +1264,8 @@ __rec_txn_read(WT_SESSION_IMPL *session, WT_RECONCILE *r,
 					skipped = true;
 			}
 		}
-	}
+
+	r->update_mem_all += update_mem;
 
 	/*
 	 * If all of the updates were aborted, quit. This test is not strictly
@@ -1227,12 +1312,6 @@ __rec_txn_read(WT_SESSION_IMPL *session, WT_RECONCILE *r,
 		    txnid != S2C(session)->txn_global.checkpoint_txnid ||
 		    WT_SESSION_IS_CHECKPOINT(session));
 #endif
-
-		/*
-		 * Track how many update chains we saw vs. how many update
-		 * chains had an entry we skipped.
-		 */
-		++r->update_skip_cnt;
 		return (0);
 	}
 
@@ -1276,6 +1355,23 @@ __rec_txn_read(WT_SESSION_IMPL *session, WT_RECONCILE *r,
 	if (skipped && !F_ISSET(r, WT_EVICT_UPDATE_RESTORE))
 		return (EBUSY);
 
+	/*
+	 * Track the memory required by the update chain.
+	 *
+	 * A page with no uncommitted (skipped) updates, that can't be evicted
+	 * because some updates aren't yet globally visible, can be evicted by
+	 * writing previous versions of the updates to the lookaside file. That
+	 * test is just checking if the skipped updates memory is zero.
+	 *
+	 * If that's not possible (there are skipped updates), we can rewrite
+	 * the pages in-memory, but we don't want to unless there's memory to
+	 * recover. That test is comparing the memory we'd recover to the memory
+	 * we'd have to re-instantiate as part of the rewrite.
+	 */
+	r->update_mem_saved += update_mem;
+	if (skipped)
+		r->update_mem_uncommitted += update_mem;
+
 	append_origv = false;
 	if (F_ISSET(r, WT_EVICT_UPDATE_RESTORE)) {
 		/*
@@ -1562,7 +1658,7 @@ __rec_child_modify(WT_SESSION_IMPL *session,
 	 * not reserved for our exclusive use, there are other page states that
 	 * must be considered.
 	 */
-	for (;; __wt_yield())
+	for (;; __wt_yield()) {
 		switch (r->tested_ref_state = ref->state) {
 		case WT_REF_DISK:
 			/* On disk, not modified by definition. */
@@ -1673,6 +1769,8 @@ __rec_child_modify(WT_SESSION_IMPL *session,
 
 		WT_ILLEGAL_VALUE(session);
 		}
+		WT_STAT_CONN_INCR(session, child_modify_blocked_page);
+	}
 
 in_memory:
 	/*
diff --git a/src/third_party/wiredtiger/src/schema/schema_alter.c b/src/third_party/wiredtiger/src/schema/schema_alter.c
index 26d800aa98e..edb3c6f77ae 100644
--- a/src/third_party/wiredtiger/src/schema/schema_alter.c
+++ b/src/third_party/wiredtiger/src/schema/schema_alter.c
@@ -61,13 +61,16 @@ __alter_colgroup(
 {
 	WT_COLGROUP *colgroup;
 	WT_DECL_RET;
+	WT_TABLE *table;
 
 	WT_ASSERT(session, F_ISSET(session, WT_SESSION_LOCKED_TABLE));
 
 	/* If we can get the colgroup, perform any potential alterations. */
 	if ((ret = __wt_schema_get_colgroup(
-	    session, uri, false, NULL, &colgroup)) == 0)
+	    session, uri, false, &table, &colgroup)) == 0) {
 		WT_TRET(__wt_schema_alter(session, colgroup->source, cfg));
+		WT_TRET(__wt_schema_release_table(session, table));
+	}
 
 	return (ret);
 }
@@ -82,11 +85,14 @@ __alter_index(
 {
 	WT_INDEX *idx;
 	WT_DECL_RET;
+	WT_TABLE *table;
 
 	/* If we can get the index, perform any potential alterations. */
 	if ((ret = __wt_schema_get_index(
-	    session, uri, false, NULL, &idx)) == 0)
+	    session, uri, false, &table, &idx)) == 0) {
 		WT_TRET(__wt_schema_alter(session, idx->source, cfg));
+		WT_TRET(__wt_schema_release_table(session, table));
+	}
 
 	return (ret);
 }
@@ -127,7 +133,8 @@ __alter_table(WT_SESSION_IMPL *session, const char *uri, const char *cfg[])
 			WT_ERR(__wt_schema_alter(
 			    session, colgroup->source, cfg));
 		}
-err:	__wt_schema_release_table(session, table);
+
+err:	WT_TRET(__wt_schema_release_table(session, table));
 	return (ret);
 }
 
diff --git a/src/third_party/wiredtiger/src/schema/schema_create.c b/src/third_party/wiredtiger/src/schema/schema_create.c
index 0677fa711a5..49c0274994c 100644
--- a/src/third_party/wiredtiger/src/schema/schema_create.c
+++ b/src/third_party/wiredtiger/src/schema/schema_create.c
@@ -273,7 +273,7 @@ err:	__wt_free(session, cgconf);
 	__wt_buf_free(session, &fmt);
 	__wt_buf_free(session, &namebuf);
 
-	__wt_schema_release_table(session, table);
+	WT_TRET(__wt_schema_release_table(session, table));
 	return (ret);
 }
 
@@ -540,7 +540,7 @@ err:	__wt_free(session, idxconf);
 	__wt_buf_free(session, &fmt);
 	__wt_buf_free(session, &namebuf);
 
-	__wt_schema_release_table(session, table);
+	WT_TRET(__wt_schema_release_table(session, table));
 	return (ret);
 }
 
@@ -615,7 +615,7 @@ err:		if (table != NULL) {
 		}
 	}
 	if (table != NULL)
-		__wt_schema_release_table(session, table);
+		WT_TRET(__wt_schema_release_table(session, table));
 	__wt_free(session, cgname);
 	__wt_free(session, tableconf);
 	return (ret);
diff --git a/src/third_party/wiredtiger/src/schema/schema_drop.c b/src/third_party/wiredtiger/src/schema/schema_drop.c
index 49801e4e5f9..18817c4fa2d 100644
--- a/src/third_party/wiredtiger/src/schema/schema_drop.c
+++ b/src/third_party/wiredtiger/src/schema/schema_drop.c
@@ -67,6 +67,7 @@ __drop_colgroup(
 	    session, uri, force, &table, &colgroup)) == 0) {
 		table->cg_complete = false;
 		WT_TRET(__wt_schema_drop(session, colgroup->source, cfg));
+		WT_TRET(__wt_schema_release_table(session, table));
 	}
 
 	WT_TRET(__wt_metadata_remove(session, uri));
@@ -90,6 +91,7 @@ __drop_index(
 	    session, uri, force, &table, &idx)) == 0) {
 		table->idx_complete = false;
 		WT_TRET(__wt_schema_drop(session, idx->source, cfg));
+		WT_TRET(__wt_schema_release_table(session, table));
 	}
 
 	WT_TRET(__wt_metadata_remove(session, uri));
@@ -151,7 +153,7 @@ __drop_table(WT_SESSION_IMPL *session, const char *uri, const char *cfg[])
 	WT_ERR(__wt_metadata_remove(session, uri));
 
 err:	if (table != NULL)
-		__wt_schema_release_table(session, table);
+		WT_TRET(__wt_schema_release_table(session, table));
 	return (ret);
 }
 
diff --git a/src/third_party/wiredtiger/src/schema/schema_list.c b/src/third_party/wiredtiger/src/schema/schema_list.c
index 74ef5135a4a..434038c6bc8 100644
--- a/src/third_party/wiredtiger/src/schema/schema_list.c
+++ b/src/third_party/wiredtiger/src/schema/schema_list.c
@@ -30,9 +30,12 @@ __schema_add_table(WT_SESSION_IMPL *session,
 	    session, name, namelen, ok_incomplete, &table));
 	WT_RET(ret);
 
-	bucket = table->name_hash % WT_HASH_ARRAY_SIZE;
-	TAILQ_INSERT_HEAD(&session->tables, table, q);
-	TAILQ_INSERT_HEAD(&session->tablehash[bucket], table, hashq);
+	if (!table->is_simple || F_ISSET(S2C(session), WT_CONN_TABLE_CACHE)) {
+		bucket = table->name_hash % WT_HASH_ARRAY_SIZE;
+		TAILQ_INSERT_HEAD(&session->tables, table, q);
+		TAILQ_INSERT_HEAD(&session->tablehash[bucket], table, hashq);
+	}
+
 	*tablep = table;
 
 	return (0);
@@ -112,11 +115,14 @@ __wt_schema_get_table(WT_SESSION_IMPL *session,
  * __wt_schema_release_table --
  *	Release a table handle.
  */
-void
+int
 __wt_schema_release_table(WT_SESSION_IMPL *session, WT_TABLE *table)
 {
 	WT_ASSERT(session, table->refcnt > 0);
-	--table->refcnt;
+	if (--table->refcnt == 0 &&
+	    table->is_simple && !F_ISSET(S2C(session), WT_CONN_TABLE_CACHE))
+		WT_RET(__wt_schema_destroy_table(session, &table));
+	return (0);
 }
 
 /*
@@ -229,9 +235,11 @@ __wt_schema_remove_table(WT_SESSION_IMPL *session, WT_TABLE *table)
 	uint64_t bucket;
 	WT_ASSERT(session, table->refcnt <= 1);
 
-	bucket = table->name_hash % WT_HASH_ARRAY_SIZE;
-	TAILQ_REMOVE(&session->tables, table, q);
-	TAILQ_REMOVE(&session->tablehash[bucket], table, hashq);
+	if (!table->is_simple || F_ISSET(S2C(session), WT_CONN_TABLE_CACHE)) {
+		bucket = table->name_hash % WT_HASH_ARRAY_SIZE;
+		TAILQ_REMOVE(&session->tables, table, q);
+		TAILQ_REMOVE(&session->tablehash[bucket], table, hashq);
+	}
 	return (__wt_schema_destroy_table(session, &table));
 }
 
@@ -249,3 +257,34 @@ __wt_schema_close_tables(WT_SESSION_IMPL *session)
 		WT_TRET(__wt_schema_remove_table(session, table));
 	return (ret);
 }
+
+/*
+ * __wt_schema_sweep_tables --
+ *	Close all idle, obsolete tables in a session.
+ */
+int
+__wt_schema_sweep_tables(WT_SESSION_IMPL *session)
+{
+	WT_TABLE *table, *next;
+	uint64_t schema_gen;
+	bool old_table_busy;
+
+	WT_ORDERED_READ(schema_gen, S2C(session)->schema_gen);
+	if (schema_gen == session->table_sweep_gen)
+		return (0);
+
+	old_table_busy = false;
+	TAILQ_FOREACH_SAFE(table, &session->tables, q, next)
+		if (table->schema_gen != schema_gen) {
+			if (table->refcnt == 0)
+				WT_RET(__wt_schema_remove_table(
+				    session, table));
+			else
+				old_table_busy = true;
+		}
+
+	if (!old_table_busy)
+		session->table_sweep_gen = schema_gen;
+
+	return (0);
+}
diff --git a/src/third_party/wiredtiger/src/schema/schema_open.c b/src/third_party/wiredtiger/src/schema/schema_open.c
index 44bd66e011a..2df6bae45f3 100644
--- a/src/third_party/wiredtiger/src/schema/schema_open.c
+++ b/src/third_party/wiredtiger/src/schema/schema_open.c
@@ -425,37 +425,40 @@ __schema_open_table(WT_SESSION_IMPL *session,
 	WT_DECL_RET;
 	WT_TABLE *table;
 	const char *tconfig;
-	char *tablename;
 
 	*tablep = NULL;
 
 	cursor = NULL;
 	table = NULL;
-	tablename = NULL;
 
 	WT_ASSERT(session, F_ISSET(session, WT_SESSION_LOCKED_TABLE));
 
+	WT_ERR(__wt_calloc_one(session, &table));
+	table->name_hash = __wt_hash_city64(name, namelen);
+
 	WT_ERR(__wt_scr_alloc(session, 0, &buf));
 	WT_ERR(__wt_buf_fmt(session, buf, "table:%.*s", (int)namelen, name));
-	WT_ERR(__wt_strndup(session, buf->data, buf->size, &tablename));
+	WT_ERR(__wt_strndup(session, buf->data, buf->size, &table->name));
 
+	/*
+	 * Don't hold the metadata cursor pinned, we call functions that use it
+	 * to retrieve column group information.
+	 */
 	WT_ERR(__wt_metadata_cursor(session, &cursor));
-	cursor->set_key(cursor, tablename);
-	WT_ERR(cursor->search(cursor));
-	WT_ERR(cursor->get_value(cursor, &tconfig));
-
-	WT_ERR(__wt_calloc_one(session, &table));
-	table->name = tablename;
-	tablename = NULL;
-	table->name_hash = __wt_hash_city64(name, namelen);
-
-	WT_ERR(__wt_config_getones(session, tconfig, "columns", &cval));
+	cursor->set_key(cursor, table->name);
+	if ((ret = cursor->search(cursor)) == 0 &&
+	    (ret = cursor->get_value(cursor, &tconfig)) == 0)
+		ret = __wt_strdup(session, tconfig, &table->config);
+	WT_TRET(__wt_metadata_cursor_release(session, &cursor));
+	WT_ERR(ret);
 
-	WT_ERR(__wt_config_getones(session, tconfig, "key_format", &cval));
+	WT_ERR(__wt_config_getones(session, table->config, "columns", &cval));
+	WT_ERR(__wt_config_getones(
+	    session, table->config, "key_format", &cval));
 	WT_ERR(__wt_strndup(session, cval.str, cval.len, &table->key_format));
-	WT_ERR(__wt_config_getones(session, tconfig, "value_format", &cval));
+	WT_ERR(__wt_config_getones(
+	    session, table->config, "value_format", &cval));
 	WT_ERR(__wt_strndup(session, cval.str, cval.len, &table->value_format));
-	WT_ERR(__wt_strdup(session, tconfig, &table->config));
 
 	/* Point to some items in the copy to save re-parsing. */
 	WT_ERR(__wt_config_getones(session, table->config,
@@ -491,7 +494,7 @@ __schema_open_table(WT_SESSION_IMPL *session,
 
 	if (table->ncolgroups > 0 && table->is_simple)
 		WT_ERR_MSG(session, EINVAL,
-		    "%s requires a table with named columns", tablename);
+		    "%s requires a table with named columns", table->name);
 
 	WT_ERR(__wt_calloc_def(session, WT_COLGROUPS(table), &table->cgroups));
 	WT_ERR(__wt_schema_open_colgroups(session, table));
@@ -509,9 +512,7 @@ __schema_open_table(WT_SESSION_IMPL *session,
 	if (0) {
 err:		WT_TRET(__wt_schema_destroy_table(session, &table));
 	}
-	WT_TRET(__wt_metadata_cursor_release(session, &cursor));
 
-	__wt_free(session, tablename);
 	__wt_scr_free(session, &buf);
 	return (ret);
 }
@@ -529,8 +530,8 @@ __wt_schema_get_colgroup(WT_SESSION_IMPL *session,
 	const char *tablename, *tend;
 	u_int i;
 
-	if (tablep != NULL)
-		*tablep = NULL;
+	WT_ASSERT(session, tablep != NULL);
+	*tablep = NULL;
 	*colgroupp = NULL;
 
 	tablename = uri;
@@ -547,15 +548,12 @@ __wt_schema_get_colgroup(WT_SESSION_IMPL *session,
 		colgroup = table->cgroups[i];
 		if (strcmp(colgroup->name, uri) == 0) {
 			*colgroupp = colgroup;
-			if (tablep != NULL)
-				*tablep = table;
-			else
-				__wt_schema_release_table(session, table);
+			*tablep = table;
 			return (0);
 		}
 	}
 
-	__wt_schema_release_table(session, table);
+	WT_RET(__wt_schema_release_table(session, table));
 	if (quiet)
 		WT_RET(ENOENT);
 	WT_RET_MSG(session, ENOENT, "%s not found in table", uri);
@@ -575,8 +573,8 @@ __wt_schema_get_index(WT_SESSION_IMPL *session,
 	const char *tablename, *tend;
 	u_int i;
 
-	if (tablep != NULL)
-		*tablep = NULL;
+	WT_ASSERT(session, tablep != NULL);
+	*tablep = NULL;
 	*indexp = NULL;
 
 	tablename = uri;
@@ -591,11 +589,8 @@ __wt_schema_get_index(WT_SESSION_IMPL *session,
 	for (i = 0; i < table->nindices; i++) {
 		idx = table->indices[i];
 		if (idx != NULL && strcmp(idx->name, uri) == 0) {
-			if (tablep != NULL)
-				*tablep = table;
-			else
-				__wt_schema_release_table(session, table);
 			*indexp = idx;
+			*tablep = table;
 			return (0);
 		}
 	}
@@ -603,15 +598,12 @@ __wt_schema_get_index(WT_SESSION_IMPL *session,
 	/* Otherwise, open it. */
 	WT_ERR(__wt_schema_open_index(
 	    session, table, tend + 1, strlen(tend + 1), indexp));
-	if (tablep != NULL)
-		*tablep = table;
+	*tablep = table;
+	return (0);
 
-err:	__wt_schema_release_table(session, table);
+err:	WT_TRET(__wt_schema_release_table(session, table));
 	WT_RET(ret);
 
-	if (*indexp != NULL)
-		return (0);
-
 	if (quiet)
 		WT_RET(ENOENT);
 	WT_RET_MSG(session, ENOENT, "%s not found in table", uri);
diff --git a/src/third_party/wiredtiger/src/schema/schema_rename.c b/src/third_party/wiredtiger/src/schema/schema_rename.c
index a374f4c2831..9effedd2cde 100644
--- a/src/third_party/wiredtiger/src/schema/schema_rename.c
+++ b/src/third_party/wiredtiger/src/schema/schema_rename.c
@@ -233,7 +233,7 @@ __rename_table(WT_SESSION_IMPL *session,
 	WT_ERR(__metadata_rename(session, uri, newuri));
 
 err:	if (table != NULL)
-		__wt_schema_release_table(session, table);
+		WT_TRET(__wt_schema_release_table(session, table));
 	return (ret);
 }
 
diff --git a/src/third_party/wiredtiger/src/schema/schema_stat.c b/src/third_party/wiredtiger/src/schema/schema_stat.c
index 345f9164e9b..9bcd2439619 100644
--- a/src/third_party/wiredtiger/src/schema/schema_stat.c
+++ b/src/third_party/wiredtiger/src/schema/schema_stat.c
@@ -19,14 +19,17 @@ __wt_curstat_colgroup_init(WT_SESSION_IMPL *session,
 	WT_COLGROUP *colgroup;
 	WT_DECL_ITEM(buf);
 	WT_DECL_RET;
+	WT_TABLE *table;
 
-	WT_RET(__wt_schema_get_colgroup(session, uri, false, NULL, &colgroup));
+	WT_RET(__wt_schema_get_colgroup(
+	    session, uri, false, &table, &colgroup));
 
 	WT_RET(__wt_scr_alloc(session, 0, &buf));
 	WT_ERR(__wt_buf_fmt(session, buf, "statistics:%s", colgroup->source));
 	ret = __wt_curstat_init(session, buf->data, NULL, cfg, cst);
 
 err:	__wt_scr_free(session, &buf);
+	WT_TRET(__wt_schema_release_table(session, table));
 	return (ret);
 }
 
@@ -41,14 +44,16 @@ __wt_curstat_index_init(WT_SESSION_IMPL *session,
 	WT_DECL_ITEM(buf);
 	WT_DECL_RET;
 	WT_INDEX *idx;
+	WT_TABLE *table;
 
-	WT_RET(__wt_schema_get_index(session, uri, false, NULL, &idx));
+	WT_RET(__wt_schema_get_index(session, uri, false, &table, &idx));
 
 	WT_RET(__wt_scr_alloc(session, 0, &buf));
 	WT_ERR(__wt_buf_fmt(session, buf, "statistics:%s", idx->source));
 	ret = __wt_curstat_init(session, buf->data, NULL, cfg, cst);
 
 err:	__wt_scr_free(session, &buf);
+	WT_TRET(__wt_schema_release_table(session, table));
 	return (ret);
 }
 
@@ -184,7 +189,7 @@ __wt_curstat_table_init(WT_SESSION_IMPL *session,
 
 	__wt_curstat_dsrc_final(cst);
 
-err:	__wt_schema_release_table(session, table);
+err:	WT_TRET(__wt_schema_release_table(session, table));
 
 	__wt_scr_free(session, &buf);
 	return (ret);
diff --git a/src/third_party/wiredtiger/src/schema/schema_truncate.c b/src/third_party/wiredtiger/src/schema/schema_truncate.c
index 563bafa8ffc..3046305e819 100644
--- a/src/third_party/wiredtiger/src/schema/schema_truncate.c
+++ b/src/third_party/wiredtiger/src/schema/schema_truncate.c
@@ -33,7 +33,7 @@ __truncate_table(WT_SESSION_IMPL *session, const char *uri, const char *cfg[])
 		WT_ERR(__wt_schema_truncate(
 		    session, table->indices[i]->source, cfg));
 
-err:	__wt_schema_release_table(session, table);
+err:	WT_TRET(__wt_schema_release_table(session, table));
 	return (ret);
 }
 
diff --git a/src/third_party/wiredtiger/src/schema/schema_worker.c b/src/third_party/wiredtiger/src/schema/schema_worker.c
index 62cdd7d367b..9fb68720723 100644
--- a/src/third_party/wiredtiger/src/schema/schema_worker.c
+++ b/src/third_party/wiredtiger/src/schema/schema_worker.c
@@ -64,14 +64,17 @@ __wt_schema_worker(WT_SESSION_IMPL *session,
 		}
 	} else if (WT_PREFIX_MATCH(uri, "colgroup:")) {
 		WT_ERR(__wt_schema_get_colgroup(
-		    session, uri, false, NULL, &colgroup));
+		    session, uri, false, &table, &colgroup));
 		WT_ERR(__wt_schema_worker(session,
 		    colgroup->source, file_func, name_func, cfg, open_flags));
+		WT_ERR(__wt_schema_release_table(session, table));
 	} else if (WT_PREFIX_SKIP(tablename, "index:")) {
 		idx = NULL;
-		WT_ERR(__wt_schema_get_index(session, uri, false, NULL, &idx));
+		WT_ERR(__wt_schema_get_index(
+		    session, uri, false, &table, &idx));
 		WT_ERR(__wt_schema_worker(session, idx->source,
 		    file_func, name_func, cfg, open_flags));
+		WT_ERR(__wt_schema_release_table(session, table));
 	} else if (WT_PREFIX_MATCH(uri, "lsm:")) {
 		WT_ERR(__wt_lsm_tree_worker(session,
 		    uri, file_func, name_func, cfg, open_flags));
@@ -128,6 +131,6 @@ __wt_schema_worker(WT_SESSION_IMPL *session,
 		WT_ERR(__wt_bad_object_type(session, uri));
 
 err:	if (table != NULL)
-		__wt_schema_release_table(session, table);
+		WT_TRET(__wt_schema_release_table(session, table));
 	return (ret);
 }
diff --git a/src/third_party/wiredtiger/src/session/session_api.c b/src/third_party/wiredtiger/src/session/session_api.c
index 89a5a2c633d..386084b78d9 100644
--- a/src/third_party/wiredtiger/src/session/session_api.c
+++ b/src/third_party/wiredtiger/src/session/session_api.c
@@ -328,6 +328,7 @@ __session_open_cursor_int(WT_SESSION_IMPL *session, const char *uri,
 	WT_COLGROUP *colgroup;
 	WT_DATA_SOURCE *dsrc;
 	WT_DECL_RET;
+	WT_TABLE *table;
 
 	*cursorp = NULL;
 
@@ -355,9 +356,10 @@ __session_open_cursor_int(WT_SESSION_IMPL *session, const char *uri,
 			 * the underlying data source.
 			 */
 			WT_RET(__wt_schema_get_colgroup(
-			    session, uri, false, NULL, &colgroup));
+			    session, uri, false, &table, &colgroup));
 			WT_RET(__wt_open_cursor(
 			    session, colgroup->source, owner, cfg, cursorp));
+			WT_RET(__wt_schema_release_table(session, table));
 		} else if (WT_PREFIX_MATCH(uri, "config:"))
 			WT_RET(__wt_curconfig_open(
 			    session, uri, cfg, cursorp));
@@ -818,6 +820,8 @@ __session_reset(WT_SESSION *wt_session)
 
 	WT_TRET(__wt_session_reset_cursors(session, true));
 
+	WT_TRET(__wt_schema_sweep_tables(session));
+
 	/* Release common session resources. */
 	WT_TRET(__wt_session_release_resources(session));
 
@@ -1209,12 +1213,15 @@ __wt_session_range_truncate(WT_SESSION_IMPL *session,
 
 done:
 err:	/*
-	 * Close any locally-opened start cursor. Reset application cursors,
-	 * they've possibly moved and the application cannot use them.
+	 * Close any locally-opened start cursor.
+	 *
+	 * Reset application cursors, they've possibly moved and the
+	 * application cannot use them.  Note that we can make it here with a
+	 * NULL start cursor (e.g., if the truncate range is empty).
 	 */
 	if (local_start)
 		WT_TRET(start->close(start));
-	else
+	else if (start != NULL)
 		WT_TRET(start->reset(start));
 	if (stop != NULL)
 		WT_TRET(stop->reset(stop));
diff --git a/src/third_party/wiredtiger/src/session/session_dhandle.c b/src/third_party/wiredtiger/src/session/session_dhandle.c
index ffeb6137766..707e07ac11f 100644
--- a/src/third_party/wiredtiger/src/session/session_dhandle.c
+++ b/src/third_party/wiredtiger/src/session/session_dhandle.c
@@ -235,6 +235,7 @@ __wt_session_lock_dhandle(
 		lock_busy = true;
 
 		/* Give other threads a chance to make progress. */
+		WT_STAT_CONN_INCR(session, dhandle_lock_blocked);
 		__wt_yield();
 	}
 }
@@ -597,7 +598,9 @@ __wt_session_lock_checkpoint(WT_SESSION_IMPL *session, const char *checkpoint)
 	 * the underlying file are visible to the in-memory pages.
 	 */
 	WT_ERR(__wt_evict_file_exclusive_on(session));
-	WT_ERR(__wt_cache_op(session, WT_SYNC_DISCARD));
+	ret = __wt_cache_op(session, WT_SYNC_DISCARD);
+	__wt_evict_file_exclusive_off(session);
+	WT_ERR(ret);
 
 	/*
 	 * We lock checkpoint handles that we are overwriting, so the handle
diff --git a/src/third_party/wiredtiger/src/support/err.c b/src/third_party/wiredtiger/src/support/err.c
index 57efde72b23..f98b1943449 100644
--- a/src/third_party/wiredtiger/src/support/err.c
+++ b/src/third_party/wiredtiger/src/support/err.c
@@ -494,7 +494,18 @@ __wt_panic(WT_SESSION_IMPL *session)
     WT_GCC_FUNC_ATTRIBUTE((cold))
     WT_GCC_FUNC_ATTRIBUTE((visibility("default")))
 {
-	F_SET(S2C(session), WT_CONN_PANIC);
+	WT_CONNECTION_IMPL *conn;
+
+	conn = S2C(session);
+
+	/*
+	 * If the connection has already be marked for panic, just return the
+	 * error.
+	 */
+	if (F_ISSET(conn, WT_CONN_PANIC))
+		return (WT_PANIC);
+
+	F_SET(conn, WT_CONN_PANIC);
 	__wt_err(session, WT_PANIC, "the process must exit and restart");
 
 #if defined(HAVE_DIAGNOSTIC)
diff --git a/src/third_party/wiredtiger/src/support/rand.c b/src/third_party/wiredtiger/src/support/rand.c
index 4fae43edc8e..a5b229b9abc 100644
--- a/src/third_party/wiredtiger/src/support/rand.c
+++ b/src/third_party/wiredtiger/src/support/rand.c
@@ -120,15 +120,3 @@ __wt_random(WT_RAND_STATE volatile * rnd_state)
 
 	return ((z << 16) + (w & 65535));
 }
-
-/*
- * __wt_random64 --
- *	Return a 64-bit pseudo-random number.
- */
-uint64_t
-__wt_random64(WT_RAND_STATE volatile * rnd_state)
-    WT_GCC_FUNC_ATTRIBUTE((visibility("default")))
-{
-	return (((uint64_t)__wt_random(rnd_state) << 32) +
-	    __wt_random(rnd_state));
-}
diff --git a/src/third_party/wiredtiger/src/support/stat.c b/src/third_party/wiredtiger/src/support/stat.c
index 8b72e653658..c9e577ac3b6 100644
--- a/src/third_party/wiredtiger/src/support/stat.c
+++ b/src/third_party/wiredtiger/src/support/stat.c
@@ -842,11 +842,19 @@ static const char * const __stats_connection_desc[] = {
 	"thread-state: active filesystem write calls",
 	"thread-yield: application thread time evicting (usecs)",
 	"thread-yield: application thread time waiting for cache (usecs)",
+	"thread-yield: connection close blocked waiting for transaction state stabilization",
+	"thread-yield: connection close yielded for lsm manager shutdown",
+	"thread-yield: data handle lock yielded",
+	"thread-yield: log server sync yielded for log write",
 	"thread-yield: page acquire busy blocked",
 	"thread-yield: page acquire eviction blocked",
 	"thread-yield: page acquire locked blocked",
 	"thread-yield: page acquire read blocked",
 	"thread-yield: page acquire time sleeping (usecs)",
+	"thread-yield: page delete rollback yielded for instantiation",
+	"thread-yield: page reconciliation yielded due to child modification",
+	"thread-yield: reference for page index and slot yielded",
+	"thread-yield: tree descend one level yielded for split page index update",
 	"transaction: number of named snapshots created",
 	"transaction: number of named snapshots dropped",
 	"transaction: transaction begins",
@@ -1129,11 +1137,19 @@ __wt_stat_connection_clear_single(WT_CONNECTION_STATS *stats)
 		/* not clearing thread_write_active */
 	stats->application_evict_time = 0;
 	stats->application_cache_time = 0;
+	stats->txn_release_blocked = 0;
+	stats->conn_close_blocked_lsm = 0;
+	stats->dhandle_lock_blocked = 0;
+	stats->log_server_sync_blocked = 0;
 	stats->page_busy_blocked = 0;
 	stats->page_forcible_evict_blocked = 0;
 	stats->page_locked_blocked = 0;
 	stats->page_read_blocked = 0;
 	stats->page_sleep = 0;
+	stats->page_del_rollback_blocked = 0;
+	stats->child_modify_blocked_page = 0;
+	stats->page_index_slot_blocked = 0;
+	stats->tree_descend_blocked = 0;
 	stats->txn_snapshots_created = 0;
 	stats->txn_snapshots_dropped = 0;
 	stats->txn_begin = 0;
@@ -1475,12 +1491,25 @@ __wt_stat_connection_aggregate(
 	    WT_STAT_READ(from, application_evict_time);
 	to->application_cache_time +=
 	    WT_STAT_READ(from, application_cache_time);
+	to->txn_release_blocked += WT_STAT_READ(from, txn_release_blocked);
+	to->conn_close_blocked_lsm +=
+	    WT_STAT_READ(from, conn_close_blocked_lsm);
+	to->dhandle_lock_blocked += WT_STAT_READ(from, dhandle_lock_blocked);
+	to->log_server_sync_blocked +=
+	    WT_STAT_READ(from, log_server_sync_blocked);
 	to->page_busy_blocked += WT_STAT_READ(from, page_busy_blocked);
 	to->page_forcible_evict_blocked +=
 	    WT_STAT_READ(from, page_forcible_evict_blocked);
 	to->page_locked_blocked += WT_STAT_READ(from, page_locked_blocked);
 	to->page_read_blocked += WT_STAT_READ(from, page_read_blocked);
 	to->page_sleep += WT_STAT_READ(from, page_sleep);
+	to->page_del_rollback_blocked +=
+	    WT_STAT_READ(from, page_del_rollback_blocked);
+	to->child_modify_blocked_page +=
+	    WT_STAT_READ(from, child_modify_blocked_page);
+	to->page_index_slot_blocked +=
+	    WT_STAT_READ(from, page_index_slot_blocked);
+	to->tree_descend_blocked += WT_STAT_READ(from, tree_descend_blocked);
 	to->txn_snapshots_created +=
 	    WT_STAT_READ(from, txn_snapshots_created);
 	to->txn_snapshots_dropped +=
diff --git a/src/third_party/wiredtiger/src/support/time.c b/src/third_party/wiredtiger/src/support/time.c
new file mode 100644
index 00000000000..0e4562c0234
--- /dev/null
+++ b/src/third_party/wiredtiger/src/support/time.c
@@ -0,0 +1,89 @@
+/*-
+ * Public Domain 2014-2017 MongoDB, Inc.
+ * Public Domain 2008-2014 WiredTiger, Inc.
+ *
+ * This is free and unencumbered software released into the public domain.
+ *
+ * Anyone is free to copy, modify, publish, use, compile, sell, or
+ * distribute this software, either in source code form or as a compiled
+ * binary, for any purpose, commercial or non-commercial, and by any
+ * means.
+ *
+ * In jurisdictions that recognize copyright laws, the author or authors
+ * of this software dedicate any and all copyright interest in the
+ * software to the public domain. We make this dedication for the benefit
+ * of the public at large and to the detriment of our heirs and
+ * successors. We intend this dedication to be an overt act of
+ * relinquishment in perpetuity of all present and future rights to this
+ * software under copyright law.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "wt_internal.h"
+
+/*
+ * __time_check_monotonic --
+ *	Check and prevent time running backward.  If we detect that it has, we
+ *	set the time structure to the previous values, making time stand still
+ *	until we see a time in the future of the highest value seen so far.
+ */
+static void
+__time_check_monotonic(WT_SESSION_IMPL *session, struct timespec *tsp)
+{
+	/*
+	 * Detect time going backward.  If so, use the last
+	 * saved timestamp.
+	 */
+	if (session == NULL)
+		return;
+
+	if (tsp->tv_sec < session->last_epoch.tv_sec ||
+	     (tsp->tv_sec == session->last_epoch.tv_sec &&
+	     tsp->tv_nsec < session->last_epoch.tv_nsec)) {
+		WT_STAT_CONN_INCR(session, time_travel);
+		*tsp = session->last_epoch;
+	} else
+		session->last_epoch = *tsp;
+}
+
+/*
+ * __wt_epoch --
+ *	Return the time since the Epoch, adjusted so it never appears to go
+ *	backwards.
+ */
+void
+__wt_epoch(WT_SESSION_IMPL *session, struct timespec *tsp)
+    WT_GCC_FUNC_ATTRIBUTE((visibility("default")))
+{
+	struct timespec tmp;
+
+	/*
+	 * Read into a local variable so that we're comparing the correct
+	 * value when we check for monotonic increasing time.  There are
+	 * many places we read into an unlocked global variable.
+	 */
+	__wt_epoch_raw(session, &tmp);
+	__time_check_monotonic(session, &tmp);
+	*tsp = tmp;
+}
+
+/*
+ * __wt_seconds --
+ *	Return the seconds since the Epoch.
+ */
+void
+__wt_seconds(WT_SESSION_IMPL *session, time_t *timep)
+{
+	struct timespec t;
+
+	__wt_epoch(session, &t);
+
+	*timep = t.tv_sec;
+}
diff --git a/src/third_party/wiredtiger/src/txn/txn.c b/src/third_party/wiredtiger/src/txn/txn.c
index ea7faa2e966..7665ba56adc 100644
--- a/src/third_party/wiredtiger/src/txn/txn.c
+++ b/src/third_party/wiredtiger/src/txn/txn.c
@@ -503,13 +503,17 @@ __wt_txn_commit(WT_SESSION_IMPL *session, const char *cfg[])
 	WT_CONNECTION_IMPL *conn;
 	WT_DECL_RET;
 	WT_TXN *txn;
+	WT_TXN_GLOBAL *txn_global;
 	WT_TXN_OP *op;
 	u_int i;
-	bool did_update;
+	bool did_update, locked;
 
 	txn = &session->txn;
 	conn = S2C(session);
+	txn_global = &conn->txn_global;
 	did_update = txn->mod_count != 0;
+	locked = false;
+
 	WT_ASSERT(session, !F_ISSET(txn, WT_TXN_ERROR) || !did_update);
 
 	if (!F_ISSET(txn, WT_TXN_RUNNING))
@@ -580,6 +584,14 @@ __wt_txn_commit(WT_SESSION_IMPL *session, const char *cfg[])
 		 * This is particularly important for checkpoints.
 		 */
 		__wt_txn_release_snapshot(session);
+		/*
+		 * We hold the visibility lock for reading from the time
+		 * we write our log record until the time we release our
+		 * transaction so that the LSN any checkpoint gets will
+		 * always reflect visible data.
+		 */
+		__wt_readlock(session, &txn_global->visibility_rwlock);
+		locked = true;
 		ret = __wt_txn_log_commit(session, cfg);
 	}
 
@@ -590,6 +602,9 @@ __wt_txn_commit(WT_SESSION_IMPL *session, const char *cfg[])
 	 * Nothing can fail after this point.
 	 */
 	if (ret != 0) {
+		if (locked)
+			__wt_readunlock(session,
+			    &txn_global->visibility_rwlock);
 		WT_TRET(__wt_txn_rollback(session, cfg));
 		return (ret);
 	}
@@ -600,6 +615,8 @@ __wt_txn_commit(WT_SESSION_IMPL *session, const char *cfg[])
 	txn->mod_count = 0;
 
 	__wt_txn_release(session);
+	if (locked)
+		__wt_readunlock(session, &txn_global->visibility_rwlock);
 	return (0);
 }
 
@@ -770,6 +787,7 @@ __wt_txn_global_init(WT_SESSION_IMPL *session, const char *cfg[])
 	    &txn_global->id_lock, "transaction id lock"));
 	WT_RET(__wt_rwlock_init(session, &txn_global->scan_rwlock));
 	WT_RET(__wt_rwlock_init(session, &txn_global->nsnap_rwlock));
+	WT_RET(__wt_rwlock_init(session, &txn_global->visibility_rwlock));
 	txn_global->nsnap_oldest_id = WT_TXN_NONE;
 	TAILQ_INIT(&txn_global->nsnaph);
 
@@ -801,6 +819,7 @@ __wt_txn_global_destroy(WT_SESSION_IMPL *session)
 	__wt_spin_destroy(session, &txn_global->id_lock);
 	__wt_rwlock_destroy(session, &txn_global->scan_rwlock);
 	__wt_rwlock_destroy(session, &txn_global->nsnap_rwlock);
+	__wt_rwlock_destroy(session, &txn_global->visibility_rwlock);
 	__wt_free(session, txn_global->states);
 }
 
diff --git a/src/third_party/wiredtiger/src/txn/txn_log.c b/src/third_party/wiredtiger/src/txn/txn_log.c
index 2931dc1ce82..09a8c4d9663 100644
--- a/src/third_party/wiredtiger/src/txn/txn_log.c
+++ b/src/third_party/wiredtiger/src/txn/txn_log.c
@@ -289,16 +289,20 @@ int
 __wt_txn_checkpoint_log(
     WT_SESSION_IMPL *session, bool full, uint32_t flags, WT_LSN *lsnp)
 {
+	WT_CONNECTION_IMPL *conn;
 	WT_DECL_ITEM(logrec);
 	WT_DECL_RET;
 	WT_ITEM *ckpt_snapshot, empty;
 	WT_LSN *ckpt_lsn;
 	WT_TXN *txn;
+	WT_TXN_GLOBAL *txn_global;
 	uint8_t *end, *p;
 	size_t recsize;
 	uint32_t i, rectype = WT_LOGREC_CHECKPOINT;
 	const char *fmt = WT_UNCHECKED_STRING(IIIIu);
 
+	conn = S2C(session);
+	txn_global = &conn->txn_global;
 	txn = &session->txn;
 	ckpt_lsn = &txn->ckpt_lsn;
 
@@ -320,6 +324,15 @@ __wt_txn_checkpoint_log(
 		txn->full_ckpt = true;
 		WT_ERR(__wt_log_flush_lsn(session, ckpt_lsn, true));
 		/*
+		 * We take and immediately release the visibility lock.
+		 * Acquiring the write lock guarantees that any transaction
+		 * that has written to the log has also made its transaction
+		 * visible at this time.
+		 */
+		__wt_writelock(session, &txn_global->visibility_rwlock);
+		__wt_writeunlock(session, &txn_global->visibility_rwlock);
+
+		/*
 		 * We need to make sure that the log records in the checkpoint
 		 * LSN are on disk.  In particular to make sure that the
 		 * current log file exists.
@@ -363,20 +376,20 @@ __wt_txn_checkpoint_log(
 		    txn->ckpt_nsnapshot, ckpt_snapshot));
 		logrec->size += (uint32_t)recsize;
 		WT_ERR(__wt_log_write(session, logrec, lsnp,
-		    F_ISSET(S2C(session), WT_CONN_CKPT_SYNC) ?
+		    F_ISSET(conn, WT_CONN_CKPT_SYNC) ?
 		    WT_LOG_FSYNC : 0));
 
 		/*
 		 * If this full checkpoint completed successfully and there is
-		 * no hot backup in progress and this is not recovery, tell
-		 * the logging subsystem the checkpoint LSN so that it can
-		 * archive.  Do not update the logging checkpoint LSN if this
-		 * is during a clean connection close, only during a full
-		 * checkpoint.  A clean close may not update any metadata LSN
-		 * and we do not want to archive in that case.
+		 * no hot backup in progress and this is not an unclean
+		 * recovery, tell the logging subsystem the checkpoint LSN so
+		 * that it can archive.  Do not update the logging checkpoint
+		 * LSN if this is during a clean connection close, only during
+		 * a full checkpoint.  A clean close may not update any
+		 * metadata LSN and we do not want to archive in that case.
 		 */
-		if (!S2C(session)->hot_backup &&
-		    !F_ISSET(S2C(session), WT_CONN_RECOVERING) &&
+		if (!conn->hot_backup &&
+		    !FLD_ISSET(conn->log_flags, WT_CONN_LOG_RECOVER_DIRTY) &&
 		    txn->full_ckpt)
 			__wt_log_ckpt(session, ckpt_lsn);
 
diff --git a/src/third_party/wiredtiger/src/txn/txn_recover.c b/src/third_party/wiredtiger/src/txn/txn_recover.c
index 30932195b1e..29f4dee5199 100644
--- a/src/third_party/wiredtiger/src/txn/txn_recover.c
+++ b/src/third_party/wiredtiger/src/txn/txn_recover.c
@@ -20,6 +20,7 @@ typedef struct {
 	} *files;
 	size_t file_alloc;		/* Allocated size of files array. */
 	u_int max_fileid;		/* Maximum file ID seen. */
+	WT_LSN max_lsn;			/* Maximum checkpoint LSN seen. */
 	u_int nfiles;			/* Number of files in the metadata. */
 
 	WT_LSN ckpt_lsn;		/* Start LSN for main recovery loop. */
@@ -342,6 +343,10 @@ __recovery_setup_file(WT_RECOVERY *r, const char *uri, const char *config)
 	    "Recovering %s with id %" PRIu32 " @ (%" PRIu32 ", %" PRIu32 ")",
 	    uri, fileid, lsn.l.file, lsn.l.offset);
 
+	if ((!WT_IS_MAX_LSN(&lsn) && !WT_IS_INIT_LSN(&lsn)) &&
+	    (WT_IS_MAX_LSN(&r->max_lsn) || __wt_log_cmp(&lsn, &r->max_lsn) > 0))
+		r->max_lsn = lsn;
+
 	return (0);
 
 }
@@ -428,6 +433,7 @@ __wt_txn_recover(WT_SESSION_IMPL *session)
 	WT_RET(__wt_open_internal_session(conn, "txn-recover",
 	    false, WT_SESSION_NO_LOGGING, &session));
 	r.session = session;
+	WT_MAX_LSN(&r.max_lsn);
 
 	F_SET(conn, WT_CONN_RECOVERING);
 	WT_ERR(__wt_metadata_search(session, WT_METAFILE_URI, &config));
@@ -443,9 +449,29 @@ __wt_txn_recover(WT_SESSION_IMPL *session)
 	 */
 	if (!FLD_ISSET(S2C(session)->log_flags, WT_CONN_LOG_EXISTED) ||
 	    WT_IS_MAX_LSN(&metafile->ckpt_lsn)) {
+		/*
+		 * Detect if we're going from logging disabled to enabled.
+		 * We need to know this to verify LSNs and start at the correct
+		 * log file later.  If someone ran with logging, then disabled
+		 * it and removed all the log files and then turned logging back
+		 * on, we have to start logging in the log file number that is
+		 * larger than any checkpoint LSN we have from the earlier time.
+		 */
 		WT_ERR(__recovery_file_scan(&r));
+		/*
+		 * The array can be re-allocated in recovery_file_scan.  Reset
+		 * our pointer after scanning all the files.
+		 */
+		metafile = &r.files[WT_METAFILE_ID];
 		conn->next_file_id = r.max_fileid;
-		goto done;
+
+		if (FLD_ISSET(conn->log_flags, WT_CONN_LOG_ENABLED) &&
+		    WT_IS_MAX_LSN(&metafile->ckpt_lsn) &&
+		    !WT_IS_MAX_LSN(&r.max_lsn)) {
+			WT_ERR(__wt_log_reset(session, r.max_lsn.l.file));
+			goto ckpt;
+		} else
+			goto done;
 	}
 
 	/*
@@ -488,6 +514,11 @@ __wt_txn_recover(WT_SESSION_IMPL *session)
 
 	/* Scan the metadata to find the live files and their IDs. */
 	WT_ERR(__recovery_file_scan(&r));
+	/*
+	 * Clear this out.  We no longer need it and it could have been
+	 * re-allocated when scanning the files.
+	 */
+	metafile = NULL;
 
 	/*
 	 * We no longer need the metadata cursor: close it to avoid pinning any
@@ -535,6 +566,8 @@ __wt_txn_recover(WT_SESSION_IMPL *session)
 	 * this is not a read-only connection.
 	 * We can consider skipping it in the future.
 	 */
+	if (needs_rec)
+		FLD_SET(conn->log_flags, WT_CONN_LOG_RECOVER_DIRTY);
 	if (WT_IS_INIT_LSN(&r.ckpt_lsn))
 		WT_ERR(__wt_log_scan(session, NULL,
 		    WT_LOGSCAN_FIRST | WT_LOGSCAN_RECOVER,
@@ -554,11 +587,12 @@ __wt_txn_recover(WT_SESSION_IMPL *session)
 	 * open is fast and keep the metadata up to date with the checkpoint
 	 * LSN and archiving.
 	 */
-	WT_ERR(session->iface.checkpoint(&session->iface, "force=1"));
+ckpt:	WT_ERR(session->iface.checkpoint(&session->iface, "force=1"));
 
 done:	FLD_SET(conn->log_flags, WT_CONN_LOG_RECOVER_DONE);
 err:	WT_TRET(__recovery_free(&r));
 	__wt_free(session, config);
+	FLD_CLR(conn->log_flags, WT_CONN_LOG_RECOVER_DIRTY);
 
 	if (ret != 0)
 		__wt_err(session, ret, "Recovery failed");
diff --git a/src/third_party/wiredtiger/test/mciproject.yml b/src/third_party/wiredtiger/test/mciproject.yml
index 6456475aa00..50a910d9e58 100644
--- a/src/third_party/wiredtiger/test/mciproject.yml
+++ b/src/third_party/wiredtiger/test/mciproject.yml
@@ -157,20 +157,6 @@ buildvariants:
     - name: unit-test
     - name: fops
 
-- name: solaris
-  display_name: Solaris
-  run_on:
-  - solaris
-  expansions:
-    make_command: PATH=/opt/mongodbtoolchain/bin:$PATH gmake
-    test_env_vars: LD_LIBRARY_PATH=`pwd`/.libs
-    smp_command: -j $(kstat cpu | sort -u | grep -c "^module")
-    configure_env_vars: PATH=/opt/mongodbtoolchain/bin:$PATH CFLAGS="-m64"
-  tasks:
-    - name: compile
-    - name: unit-test
-    - name: fops
-
 - name: windows-64
   display_name: Windows 64-bit
   run_on:
diff --git a/src/third_party/wiredtiger/test/recovery/random-abort.c b/src/third_party/wiredtiger/test/recovery/random-abort.c
index febe6530534..b53383e5730 100644
--- a/src/third_party/wiredtiger/test/recovery/random-abort.c
+++ b/src/third_party/wiredtiger/test/recovery/random-abort.c
@@ -47,9 +47,9 @@ static bool inmem;
 #define	RECORDS_FILE	"records-%" PRIu32
 
 #define	ENV_CONFIG_DEF						\
-    "create,log=(file_max=10M,archive=false,enabled)"
+    "create,log=(file_max=10M,enabled)"
 #define	ENV_CONFIG_TXNSYNC					\
-    "create,log=(file_max=10M,archive=false,enabled),"		\
+    "create,log=(file_max=10M,enabled),"		\
     "transaction_sync=(enabled,method=none)"
 #define	ENV_CONFIG_REC "log=(recover=on)"
 #define	MAX_VAL	4096
diff --git a/src/third_party/wiredtiger/test/suite/test_bug018.py b/src/third_party/wiredtiger/test/suite/test_bug018.py
new file mode 100644
index 00000000000..7d20ebcaacb
--- /dev/null
+++ b/src/third_party/wiredtiger/test/suite/test_bug018.py
@@ -0,0 +1,98 @@
+#!/usr/bin/env python
+#
+# Public Domain 2014-2017 MongoDB, Inc.
+# Public Domain 2008-2014 WiredTiger, Inc.
+#
+# This is free and unencumbered software released into the public domain.
+#
+# Anyone is free to copy, modify, publish, use, compile, sell, or
+# distribute this software, either in source code form or as a compiled
+# binary, for any purpose, commercial or non-commercial, and by any
+# means.
+#
+# In jurisdictions that recognize copyright laws, the author or authors
+# of this software dedicate any and all copyright interest in the
+# software to the public domain. We make this dedication for the benefit
+# of the public at large and to the detriment of our heirs and
+# successors. We intend this dedication to be an overt act of
+# relinquishment in perpetuity of all present and future rights to this
+# software under copyright law.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+# EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+# IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+# OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+# ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+# OTHER DEALINGS IN THE SOFTWARE.
+
+from helper import copy_wiredtiger_home
+import os
+import wiredtiger, wttest
+
+# test_bug018.py
+#   JIRA WT-3590: if writing table data fails during close then tables
+# that were updated within the same transaction could get out of sync with
+# each other.
+class test_bug018(wttest.WiredTigerTestCase):
+    '''Test closing/reopening/recovering tables when writes fail'''
+
+    conn_config = 'log=(enabled)'
+
+    def setUp(self):
+        # This test uses Linux-specific code so skip on any other system.
+        if os.name != 'posix' or os.uname()[0] != 'Linux':
+            self.skipTest('Linux-specific test skipped on ' + os.name)
+        super(test_bug018, self).setUp()
+
+    def create_table(self, uri):
+        self.session.create(uri, 'key_format=S,value_format=S')
+        return self.session.open_cursor(uri)
+
+    def test_bug018(self):
+        '''Test closing multiple tables'''
+        basename = 'bug018.'
+        baseuri = 'file:' + basename
+        c1 = self.create_table(baseuri + '01.wt')
+        c2 = self.create_table(baseuri + '02.wt')
+
+        self.session.begin_transaction()
+        c1['key'] = 'value'
+        c2['key'] = 'value'
+        self.session.commit_transaction()
+
+        # Simulate a write failure by closing the file descriptor for the second
+        # table out from underneath WiredTiger.  We do this right before
+        # closing the connection so that the write error happens during close
+        # when writing out the final data.  Allow table 1 to succeed and force
+        # an erorr writing out table 2.
+        #
+        # This is Linux-specific code to figure out the file descriptor.
+        for f in os.listdir('/proc/self/fd'):
+            try:
+                if os.readlink('/proc/self/fd/' + f).endswith(basename + '02.wt'):
+                    os.close(int(f))
+            except OSError:
+                pass
+
+        # Expect an error and messages, so turn off stderr checking.
+        with self.expectedStderrPattern(''):
+            try:
+                self.close_conn()
+            except wiredtiger.WiredTigerError:
+                self.conn = None
+
+        # Make a backup for forensics in case something goes wrong.
+        backup_dir = 'BACKUP'
+        copy_wiredtiger_home('.', backup_dir, True)
+
+        # After reopening and running recovery both tables should be in
+        # sync even though table 1 was successfully written and table 2
+        # had an error on close.
+        self.open_conn()
+        c1 = self.session.open_cursor(baseuri + '01.wt')
+        c2 = self.session.open_cursor(baseuri + '02.wt')
+        self.assertEqual(list(c1), list(c2))
+
+if __name__ == '__main__':
+    wttest.run()
diff --git a/src/third_party/wiredtiger/test/suite/test_inmem01.py b/src/third_party/wiredtiger/test/suite/test_inmem01.py
index 388485db29b..d280642942a 100644
--- a/src/third_party/wiredtiger/test/suite/test_inmem01.py
+++ b/src/third_party/wiredtiger/test/suite/test_inmem01.py
@@ -108,12 +108,15 @@ class test_inmem01(wttest.WiredTigerTestCase):
 
         cursor.reset()
         # Spin inserting to give eviction a chance to reclaim space
+        sleeps = 0
         inserted = False
         for i in range(1, 1000):
             try:
                 cursor[ds.key(1)] = ds.value(1)
             except wiredtiger.WiredTigerError:
                 cursor.reset()
+                sleeps = sleeps + 1
+                self.assertLess(sleeps, 60 * 5)
                 sleep(1)
                 continue
             inserted = True
diff --git a/src/third_party/wiredtiger/test/suite/test_las.py b/src/third_party/wiredtiger/test/suite/test_las.py
new file mode 100644
index 00000000000..d0bd1d108fa
--- /dev/null
+++ b/src/third_party/wiredtiger/test/suite/test_las.py
@@ -0,0 +1,60 @@
+#!/usr/bin/env python
+#
+# Public Domain 2014-2017 MongoDB, Inc.
+# Public Domain 2008-2014 WiredTiger, Inc.
+#
+# This is free and unencumbered software released into the public domain.
+#
+# Anyone is free to copy, modify, publish, use, compile, sell, or
+# distribute this software, either in source code form or as a compiled
+# binary, for any purpose, commercial or non-commercial, and by any
+# means.
+#
+# In jurisdictions that recognize copyright laws, the author or authors
+# of this software dedicate any and all copyright interest in the
+# software to the public domain. We make this dedication for the benefit
+# of the public at large and to the detriment of our heirs and
+# successors. We intend this dedication to be an overt act of
+# relinquishment in perpetuity of all present and future rights to this
+# software under copyright law.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+# EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+# IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+# OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+# ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+# OTHER DEALINGS IN THE SOFTWARE.
+
+import wiredtiger, wttest
+from wtdataset import SimpleDataSet
+
+# test_las.py
+#       Smoke tests to ensure lookaside tables are working.
+class test_las(wttest.WiredTigerTestCase):
+    # Force a small cache.
+    def conn_config(self):
+        return 'cache_size=1GB'
+
+    @wttest.longtest('lookaside table smoke test')
+    def test_las(self):
+        # Create a small table.
+        uri = "table:test_las"
+        nrows = 100
+        ds = SimpleDataSet(self, uri, nrows, key_format="S")
+        ds.populate()
+
+        # Take a snapshot.
+        self.session.snapshot("name=xxx")
+
+        # Insert a large number of records, we'll hang if the lookaside table
+        # isn't doing its thing.
+        c = self.session.open_cursor(uri)
+        bigvalue = "abcde" * 100
+        for i in range(1, 1000000):
+            c.set_key(ds.key(nrows + i))
+            c.set_value(bigvalue)
+            self.assertEquals(c.insert(), 0)
+
+if __name__ == '__main__':
+    wttest.run()
diff --git a/src/third_party/wiredtiger/test/suite/test_txn02.py b/src/third_party/wiredtiger/test/suite/test_txn02.py
index 01626057b9e..76a325743e9 100644
--- a/src/third_party/wiredtiger/test/suite/test_txn02.py
+++ b/src/third_party/wiredtiger/test/suite/test_txn02.py
@@ -169,7 +169,6 @@ class test_txn02(wttest.WiredTigerTestCase, suite_subprocess):
             try:
                 session = backup_conn.open_session()
             finally:
-                session.checkpoint("force")
                 self.check(backup_conn.open_session(), None, committed)
                 # Sleep long enough so that the archive thread is guaranteed
                 # to run before we close the connection.
diff --git a/src/third_party/wiredtiger/test/suite/test_txn05.py b/src/third_party/wiredtiger/test/suite/test_txn05.py
index 7aaff221ba4..7099bc972aa 100644
--- a/src/third_party/wiredtiger/test/suite/test_txn05.py
+++ b/src/third_party/wiredtiger/test/suite/test_txn05.py
@@ -134,12 +134,12 @@ class test_txn05(wttest.WiredTigerTestCase, suite_subprocess):
                  session = backup_conn.open_session()
             finally:
                 self.check(session, None, committed)
-                # Force a checkpoint because we don't record the recovery
-                # checkpoint as available for archiving.
-                session.checkpoint("force")
                 # Sleep long enough so that the archive thread is guaranteed
                 # to run before we close the connection.
                 time.sleep(1.0)
+                if count == 0:
+                    first_logs = \
+                        fnmatch.filter(os.listdir(self.backup_dir), "*Log*")
                 backup_conn.close()
             count += 1
         #
@@ -149,6 +149,11 @@ class test_txn05(wttest.WiredTigerTestCase, suite_subprocess):
         #
         cur_logs = fnmatch.filter(os.listdir(self.backup_dir), "*Log*")
         for o in orig_logs:
+            # Creating the backup was effectively an unclean shutdown so
+            # even after sleeping, we should never archive log files
+            # because a checkpoint has not run.  Later opens and runs of
+            # recovery will detect a clean shutdown and allow archiving.
+            self.assertEqual(True, o in first_logs)
             if self.archive == 'true':
                 self.assertEqual(False, o in cur_logs)
             else:
diff --git a/src/third_party/wiredtiger/test/suite/test_txn09.py b/src/third_party/wiredtiger/test/suite/test_txn09.py
index 768d714e248..b8a3d7f38ae 100644
--- a/src/third_party/wiredtiger/test/suite/test_txn09.py
+++ b/src/third_party/wiredtiger/test/suite/test_txn09.py
@@ -26,8 +26,8 @@
 # ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
 # OTHER DEALINGS IN THE SOFTWARE.
 #
-# test_txn02.py
-#   Transactions: commits and rollbacks
+# test_txn09.py
+#   Transactions: recovery toggling logging
 #
 
 import fnmatch, os, shutil, time
diff --git a/src/third_party/wiredtiger/test/suite/test_txn16.py b/src/third_party/wiredtiger/test/suite/test_txn16.py
new file mode 100644
index 00000000000..929da2291c7
--- /dev/null
+++ b/src/third_party/wiredtiger/test/suite/test_txn16.py
@@ -0,0 +1,140 @@
+#!/usr/bin/env python
+#
+# Public Domain 2014-2017 MongoDB, Inc.
+# Public Domain 2008-2014 WiredTiger, Inc.
+#
+# This is free and unencumbered software released into the public domain.
+#
+# Anyone is free to copy, modify, publish, use, compile, sell, or
+# distribute this software, either in source code form or as a compiled
+# binary, for any purpose, commercial or non-commercial, and by any
+# means.
+#
+# In jurisdictions that recognize copyright laws, the author or authors
+# of this software dedicate any and all copyright interest in the
+# software to the public domain. We make this dedication for the benefit
+# of the public at large and to the detriment of our heirs and
+# successors. We intend this dedication to be an overt act of
+# relinquishment in perpetuity of all present and future rights to this
+# software under copyright law.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+# EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+# IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+# OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+# ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+# OTHER DEALINGS IN THE SOFTWARE.
+#
+# test_txn16.py
+#   Recovery: Test that toggling between logging and not logging does not
+#   continue to generate more log files.
+#
+
+import fnmatch, os, shutil, time
+from suite_subprocess import suite_subprocess
+import wttest
+
+class test_txn16(wttest.WiredTigerTestCase, suite_subprocess):
+    t1 = 'table:test_txn16_1'
+    t2 = 'table:test_txn16_2'
+    t3 = 'table:test_txn16_3'
+    nentries = 1000
+    create_params = 'key_format=i,value_format=i'
+    # Set the log file size small so we generate checkpoints
+    # with LSNs in different files.
+    conn_config = 'config_base=false,' + \
+        'log=(archive=false,enabled,file_max=100K),' + \
+        'transaction_sync=(method=dsync,enabled)'
+    conn_on = 'config_base=false,' + \
+        'log=(archive=false,enabled,file_max=100K),' + \
+        'transaction_sync=(method=dsync,enabled)'
+    conn_off = 'config_base=false,log=(enabled=false)'
+
+    def populate_table(self, uri):
+        self.session.create(uri, self.create_params)
+        c = self.session.open_cursor(uri, None, None)
+        # Populate with an occasional checkpoint to generate
+        # some varying LSNs.
+        for i in range(self.nentries):
+            c[i] = i + 1
+            if i % 900 == 0:
+                self.session.checkpoint()
+        c.close()
+
+    def copy_dir(self, olddir, newdir):
+        ''' Simulate a crash from olddir and restart in newdir. '''
+        # with the connection still open, copy files to new directory
+        shutil.rmtree(newdir, ignore_errors=True)
+        os.mkdir(newdir)
+        for fname in os.listdir(olddir):
+            fullname = os.path.join(olddir, fname)
+            # Skip lock file on Windows since it is locked
+            if os.path.isfile(fullname) and \
+                "WiredTiger.lock" not in fullname and \
+                "Tmplog" not in fullname and \
+                "Preplog" not in fullname:
+                shutil.copy(fullname, newdir)
+        # close the original connection.
+        self.close_conn()
+
+    def run_toggle(self, homedir):
+        loop = 0
+        # Record original log files.  There should never be overlap
+        # with these even after they're removed.
+        orig_logs = fnmatch.filter(os.listdir(homedir), "*Log*")
+        while loop < 3:
+            # Reopen with logging on to run recovery first time
+            on_conn = self.wiredtiger_open(homedir, self.conn_on)
+            on_conn.close()
+            if loop > 0:
+                # Get current log files.
+                cur_logs = fnmatch.filter(os.listdir(homedir), "*Log*")
+                scur = set(cur_logs)
+                sorig = set(orig_logs)
+                # There should never be overlap with the log files that
+                # were there originally.  Mostly this checks that after
+                # opening with logging disabled and then re-enabled, we
+                # don't see log file 1.
+                self.assertEqual(scur.isdisjoint(sorig), True)
+                if loop > 1:
+                    # We should be creating the same log files each time.
+                    for l in cur_logs:
+                        self.assertEqual(l in last_logs, True)
+                    for l in last_logs:
+                        self.assertEqual(l in cur_logs, True)
+                last_logs = cur_logs
+            loop += 1
+            # Remove all log files before opening without logging.
+            cur_logs = fnmatch.filter(os.listdir(homedir), "*Log*")
+            for l in cur_logs:
+                path=homedir + "/" + l
+                os.remove(path)
+            off_conn = self.wiredtiger_open(homedir, self.conn_off)
+            off_conn.close()
+
+    def test_recovery(self):
+        ''' Check log file creation when toggling. '''
+
+        # Here's the strategy:
+        #    - With logging populate 4 tables.  Checkpoint
+        #      them at different times.
+        #    - Copy to a new directory to simulate a crash.
+        #    - Close the original connection.
+        #    On both a "copy" to simulate a crash and the original (3x):
+        #    - Record log files existing.
+        #    - Reopen with logging to run recovery.  Close connection.
+        #    - Record log files existing.
+        #    - Remove all log files.
+        #    - Open connection with logging disabled.
+        #    - Record log files existing.  Verify we don't keep adding.
+        #
+        self.populate_table(self.t1)
+        self.populate_table(self.t2)
+        self.populate_table(self.t3)
+        self.copy_dir(".", "RESTART")
+        self.run_toggle(".")
+        self.run_toggle("RESTART")
+
+if __name__ == '__main__':
+    wttest.run()
diff --git a/version.json b/version.json
index aeb6f194474..16ecc44565a 100644
--- a/version.json
+++ b/version.json
@@ -1,4 +1,4 @@
 {
-    "githash": "cf38c1b8a0a8dca4a11737581beafef4fe120bcd",
-    "version": "3.4.7"
+    "githash": "fd954412dfc10e4d1e3e2dd4fac040f8b476b268",
+    "version": "3.4.14"
 }
\ No newline at end of file