SERVER-60682 Exempt transaction coordinators and journal flusher from acquiring storage tickets

Co-authored-by: Max Hirschhorn max.hirschhorn@mongodb.com
(cherry picked from commit 1bdff76322b144ef27060fe79324fe3cce4bb17a)

2 files changed, 12 insertions, 0 deletions

diff --git a/src/mongo/db/repl/replication_consistency_markers_impl.cpp b/src/mongo/db/repl/replication_consistency_markers_impl.cpp
index df8c3ebdd6d..e7151875651 100644
--- a/src/mongo/db/repl/replication_consistency_markers_impl.cpp
+++ b/src/mongo/db/repl/replication_consistency_markers_impl.cpp
@@ -36,6 +36,7 @@
 #include "mongo/db/catalog/collection_options.h"
 #include "mongo/db/catalog_raii.h"
 #include "mongo/db/concurrency/d_concurrency.h"
+#include "mongo/db/concurrency/lock_state.h"
 #include "mongo/db/concurrency/write_conflict_exception.h"
 #include "mongo/db/repl/optime.h"
 #include "mongo/db/repl/replication_coordinator.h"
@@ -498,6 +499,12 @@ ReplicationConsistencyMarkersImpl::refreshOplogTruncateAfterPointIfPrimary(
     }
     ON_BLOCK_EXIT([&] { opCtx->recoveryUnit()->setPrepareConflictBehavior(originalBehavior); });
 
+    // Exempt storage ticket acquisition in order to avoid starving upstream requests waiting
+    // for durability. SERVER-60682 is an example with more pending prepared transactions than
+    // storage tickets; the transaction coordinator could not persist the decision and
+    // had to unnecessarily wait for prepared transactions to expire to make forward progress.
+    SkipTicketAcquisitionForLock skipTicketAcquisition(opCtx);
+
     // The locks necessary to write to the oplog truncate after point's collection and read from the
     // oplog collection must be taken up front so that the mutex can also be taken around both
     // operations without causing deadlocks.
diff --git a/src/mongo/db/s/transaction_coordinator_util.cpp b/src/mongo/db/s/transaction_coordinator_util.cpp
index 0bc0d019c6c..8fa377734ef 100644
--- a/src/mongo/db/s/transaction_coordinator_util.cpp
+++ b/src/mongo/db/s/transaction_coordinator_util.cpp
@@ -36,6 +36,7 @@
 #include "mongo/client/remote_command_retry_scheduler.h"
 #include "mongo/db/commands/txn_cmds_gen.h"
 #include "mongo/db/commands/txn_two_phase_commit_cmds_gen.h"
+#include "mongo/db/concurrency/lock_state.h"
 #include "mongo/db/curop.h"
 #include "mongo/db/dbdirectclient.h"
 #include "mongo/db/ops/write_ops.h"
@@ -411,6 +412,10 @@ Future<repl::OpTime> persistDecision(txn::AsyncWorkScheduler& scheduler,
             return scheduler.scheduleWork(
                 [lsid, txnNumber, participants, decision](OperationContext* opCtx) {
                     FlowControl::Bypass flowControlBypass(opCtx);
+                    // Do not acquire a storage ticket in order to avoid unnecessary serialization
+                    // with other prepared transactions that are holding a storage ticket
+                    // themselves; see SERVER-60682.
+                    SkipTicketAcquisitionForLock skipTicketAcquisition(opCtx);
                     getTransactionCoordinatorWorkerCurOpRepository()->set(
                         opCtx, lsid, txnNumber, CoordinatorAction::kWritingDecision);
                     return persistDecisionBlocking(opCtx, lsid, txnNumber, participants, decision);