Update upstream source from tag 'upstream/6.0.0'master

Update to upstream version '6.0.0'
with Debian dir 5604a80ec1c96ca76f25f40d78e6ef855abec322

1 files changed, 18 insertions, 41 deletions

diff --git a/src/mongo/db/s/migration_util.cpp b/src/mongo/db/s/migration_util.cpp
index f0058b60318..ec63e7477af 100644
--- a/src/mongo/db/s/migration_util.cpp
+++ b/src/mongo/db/s/migration_util.cpp
@@ -42,7 +42,7 @@
 #include "mongo/db/catalog/collection_catalog_helper.h"
 #include "mongo/db/catalog_raii.h"
 #include "mongo/db/commands.h"
-#include "mongo/db/concurrency/exception_util.h"
+#include "mongo/db/concurrency/write_conflict_exception.h"
 #include "mongo/db/dbdirectclient.h"
 #include "mongo/db/logical_session_cache.h"
 #include "mongo/db/namespace_string.h"
@@ -521,27 +521,14 @@ void resubmitRangeDeletionsOnStepUp(ServiceContext* serviceContext) {
             FindCommandRequest findCommand(NamespaceString::kRangeDeletionNamespace);
             findCommand.setFilter(BSON(RangeDeletionTask::kProcessingFieldName << true));
             auto cursor = client.find(std::move(findCommand));
-
-            auto retFuture = ExecutorFuture<void>(getMigrationUtilExecutor(serviceContext));
-
-            int rangeDeletionsMarkedAsProcessing = 0;
-            while (cursor->more()) {
-                retFuture = migrationutil::submitRangeDeletionTask(
+            if (cursor->more()) {
+                return migrationutil::submitRangeDeletionTask(
                     opCtx.get(),
                     RangeDeletionTask::parse(IDLParserErrorContext("rangeDeletionRecovery"),
                                              cursor->next()));
-                rangeDeletionsMarkedAsProcessing++;
-            }
-
-            if (rangeDeletionsMarkedAsProcessing > 1) {
-                LOGV2_WARNING(
-                    6695800,
-                    "Rescheduling several range deletions marked as processing. Orphans count "
-                    "may be off while they are not drained",
-                    "numRangeDeletionsMarkedAsProcessing"_attr = rangeDeletionsMarkedAsProcessing);
+            } else {
+                return ExecutorFuture<void>(getMigrationUtilExecutor(serviceContext));
             }
-
-            return retFuture;
         })
         .then([serviceContext] {
             ThreadClient tc("ResubmitRangeDeletions", serviceContext);
@@ -661,8 +648,9 @@ void submitOrphanRangesForCleanup(OperationContext* opCtx) {
         if (tenantDbName.dbName() == NamespaceString::kLocalDb)
             continue;
 
-        for (auto&& coll : catalog->range(tenantDbName)) {
-            auto uuid = coll->uuid();
+        for (auto collIt = catalog->begin(opCtx, tenantDbName); collIt != catalog->end(opCtx);
+             ++collIt) {
+            auto uuid = collIt.uuid().get();
             auto nss = catalog->lookupNSSByUUID(opCtx, uuid).get();
             LOGV2_DEBUG(22034,
                         2,
@@ -718,8 +706,8 @@ void persistUpdatedNumOrphans(OperationContext* opCtx,
                                << BSON("$exists" << true));
     try {
         PersistentTaskStore<RangeDeletionTask> store(NamespaceString::kRangeDeletionNamespace);
-        ScopedRangeDeleterLock rangeDeleterLock(opCtx, MODE_IX);
-        // TODO (SERVER-65996) Remove writeConflictRetry loop
+        ScopedRangeDeleterLock rangeDeleterLock(opCtx, collectionUuid);
+        // TODO (SERVER-54284) Remove writeConflictRetry loop
         writeConflictRetry(
             opCtx, "updateOrphanCount", NamespaceString::kRangeDeletionNamespace.ns(), [&] {
                 store.update(opCtx,
@@ -843,7 +831,6 @@ void persistCommitDecision(OperationContext* opCtx,
     store.upsert(opCtx,
                  BSON(MigrationCoordinatorDocument::kIdFieldName << migrationDoc.getId()),
                  migrationDoc.toBSON());
-    ShardingStatistics::get(opCtx).countDonorMoveChunkCommitted.addAndFetch(1);
 
     if (hangInPersistMigrateCommitDecisionThenSimulateErrorUninterruptible.shouldFail()) {
         hangInPersistMigrateCommitDecisionThenSimulateErrorUninterruptible.pauseWhileSet(opCtx);
@@ -863,7 +850,6 @@ void persistAbortDecision(OperationContext* opCtx,
     store.upsert(opCtx,
                  BSON(MigrationCoordinatorDocument::kIdFieldName << migrationDoc.getId()),
                  migrationDoc.toBSON());
-    ShardingStatistics::get(opCtx).countDonorMoveChunkAborted.addAndFetch(1);
 
     if (hangInPersistMigrateAbortDecisionThenSimulateErrorUninterruptible.shouldFail()) {
         hangInPersistMigrateAbortDecisionThenSimulateErrorUninterruptible.pauseWhileSet(opCtx);
@@ -1000,6 +986,11 @@ void markAsReadyRangeDeletionTaskLocally(OperationContext* opCtx, const UUID& mi
 }
 
 void deleteMigrationCoordinatorDocumentLocally(OperationContext* opCtx, const UUID& migrationId) {
+    // Before deleting the migration coordinator document, ensure that in the case of a crash, the
+    // node will start-up from at least the configTime, which it obtained as part of recovery of the
+    // shardVersion, which will ensure that it will see at least the same shardVersion.
+    VectorClockMutable::get(opCtx)->waitForDurableConfigTime().get(opCtx);
+
     PersistentTaskStore<MigrationCoordinatorDocument> store(
         NamespaceString::kMigrationCoordinatorsNamespace);
     store.remove(opCtx,
@@ -1135,7 +1126,7 @@ void recoverMigrationCoordinations(OperationContext* opCtx,
                 hangInRefreshFilteringMetadataUntilSuccessThenSimulateErrorUninterruptible
                     .pauseWhileSet();
                 uasserted(ErrorCodes::InternalError,
-                          "simulate an error response for forceGetCurrentMetadata");
+                          "simulate an error response for forceShardFilteringMetadataRefresh");
             }
 
             auto setFilteringMetadata = [&opCtx, &currentMetadata, &doc, &cancellationToken]() {
@@ -1175,15 +1166,6 @@ void recoverMigrationCoordinations(OperationContext* opCtx,
                           "coordinatorDocumentUUID"_attr = doc.getCollectionUuid());
                 }
 
-                // TODO SERVER-71918 once the drop collection coordinator starts persisting the
-                // config time we can remove this. Since the collection has been dropped,
-                // persist config time inclusive of the drop collection event before deleting
-                // leftover migration metadata.
-                // This will ensure that in case of stepdown the new
-                // primary won't read stale data from config server and think that the sharded
-                // collection still exists.
-                VectorClockMutable::get(opCtx)->waitForDurableConfigTime().get(opCtx);
-
                 deleteRangeDeletionTaskOnRecipient(opCtx, doc.getRecipientShardId(), doc.getId());
                 deleteRangeDeletionTaskLocally(opCtx, doc.getId());
                 coordinator.forgetMigration(opCtx);
@@ -1316,15 +1298,10 @@ void resumeMigrationRecipientsOnStepUp(OperationContext* opCtx) {
             const auto& nss = doc.getNss();
 
             // Register this receiveChunk on the ActiveMigrationsRegistry before completing step-up
-            // to prevent a new migration from starting while a receiveChunk was ongoing. Wait for
-            // any migrations that began in a previous term to complete if there are any.
+            // to prevent a new migration from starting while a receiveChunk was ongoing.
             auto scopedReceiveChunk(
                 uassertStatusOK(ActiveMigrationsRegistry::get(opCtx).registerReceiveChunk(
-                    opCtx,
-                    nss,
-                    doc.getRange(),
-                    doc.getDonorShardIdForLoggingPurposesOnly(),
-                    true /* waitForOngoingMigrations */)));
+                    opCtx, nss, doc.getRange(), doc.getDonorShardIdForLoggingPurposesOnly())));
 
             const auto mdm = MigrationDestinationManager::get(opCtx);
             uassertStatusOK(