SERVER-58752 Support retrying internal transactions on transient transaction errors

10 files changed, 93 insertions, 28 deletions

diff --git a/src/mongo/db/s/migration_destination_manager.cpp b/src/mongo/db/s/migration_destination_manager.cpp
index 1451fd2eced..5f81beb94fa 100644
--- a/src/mongo/db/s/migration_destination_manager.cpp
+++ b/src/mongo/db/s/migration_destination_manager.cpp
@@ -108,7 +108,8 @@ void checkOutSessionAndVerifyTxnState(OperationContext* opCtx) {
     TransactionParticipant::get(opCtx).beginOrContinue(opCtx,
                                                        *opCtx->getTxnNumber(),
                                                        boost::none /* autocommit */,
-                                                       boost::none /* startTransaction */);
+                                                       boost::none /* startTransaction */,
+                                                       boost::none /* txnRetryCounter */);
 }
 
 template <typename Callable>
@@ -921,7 +922,8 @@ void MigrationDestinationManager::_migrateThread() {
         txnParticipant.beginOrContinue(opCtx,
                                        *opCtx->getTxnNumber(),
                                        boost::none /* autocommit */,
-                                       boost::none /* startTransaction */);
+                                       boost::none /* startTransaction */,
+                                       boost::none /* txnRetryCounter */);
         _migrateDriver(opCtx);
     } catch (...) {
         _setStateFail(str::stream() << "migrate failed: " << redact(exceptionToStatus()));
diff --git a/src/mongo/db/s/resharding/resharding_data_copy_util.cpp b/src/mongo/db/s/resharding/resharding_data_copy_util.cpp
index 3461c15b815..745e83a40f9 100644
--- a/src/mongo/db/s/resharding/resharding_data_copy_util.cpp
+++ b/src/mongo/db/s/resharding/resharding_data_copy_util.cpp
@@ -249,7 +249,11 @@ boost::optional<SharedSemiFuture<void>> withSessionCheckedOut(OperationContext*
     auto txnParticipant = TransactionParticipant::get(opCtx);
 
     try {
-        txnParticipant.beginOrContinue(opCtx, txnNumber, boost::none, boost::none);
+        txnParticipant.beginOrContinue(opCtx,
+                                       txnNumber,
+                                       boost::none /* autocommit */,
+                                       boost::none /* startTransaction */,
+                                       boost::none /* txnRetryCounter */);
 
         if (stmtId && txnParticipant.checkStatementExecuted(opCtx, *stmtId)) {
             // Skip the incoming statement because it has already been logged locally.
diff --git a/src/mongo/db/s/resharding/resharding_oplog_application.cpp b/src/mongo/db/s/resharding/resharding_oplog_application.cpp
index 104e79cdd37..5a71b14a7da 100644
--- a/src/mongo/db/s/resharding/resharding_oplog_application.cpp
+++ b/src/mongo/db/s/resharding/resharding_oplog_application.cpp
@@ -86,7 +86,11 @@ void runWithTransaction(OperationContext* opCtx, unique_function<void(OperationC
         }
     });
 
-    txnParticipant.beginOrContinue(asr.opCtx(), txnNumber, false, true);
+    txnParticipant.beginOrContinue(asr.opCtx(),
+                                   txnNumber,
+                                   false /* autocommit */,
+                                   true /* startTransaction */,
+                                   boost::none /* txnRetryCounter */);
     txnParticipant.unstashTransactionResources(asr.opCtx(), "reshardingOplogApplication");
 
     func(asr.opCtx());
diff --git a/src/mongo/db/s/resharding/resharding_oplog_batch_applier_test.cpp b/src/mongo/db/s/resharding/resharding_oplog_batch_applier_test.cpp
index 9691a947954..f89da1f5b2d 100644
--- a/src/mongo/db/s/resharding/resharding_oplog_batch_applier_test.cpp
+++ b/src/mongo/db/s/resharding/resharding_oplog_batch_applier_test.cpp
@@ -177,8 +177,11 @@ public:
 
         MongoDOperationContextSession ocs(opCtx);
         auto txnParticipant = TransactionParticipant::get(opCtx);
-        txnParticipant.beginOrContinue(
-            opCtx, txnNumber, false /* autocommit */, true /* startTransaction */);
+        txnParticipant.beginOrContinue(opCtx,
+                                       txnNumber,
+                                       false /* autocommit */,
+                                       true /* startTransaction */,
+                                       boost::none /* txnRetryCounter */);
 
         txnParticipant.unstashTransactionResources(opCtx, "prepareTransaction");
 
@@ -198,8 +201,11 @@ public:
 
         MongoDOperationContextSession ocs(opCtx);
         auto txnParticipant = TransactionParticipant::get(opCtx);
-        txnParticipant.beginOrContinue(
-            opCtx, txnNumber, false /* autocommit */, boost::none /* startTransaction */);
+        txnParticipant.beginOrContinue(opCtx,
+                                       txnNumber,
+                                       false /* autocommit */,
+                                       boost::none /* startTransaction */,
+                                       boost::none /* txnRetryCounter */);
 
         txnParticipant.unstashTransactionResources(opCtx, "abortTransaction");
         txnParticipant.abortTransaction(opCtx);
diff --git a/src/mongo/db/s/resharding/resharding_oplog_session_application_test.cpp b/src/mongo/db/s/resharding/resharding_oplog_session_application_test.cpp
index 61a519c7493..29c6fd3c26b 100644
--- a/src/mongo/db/s/resharding/resharding_oplog_session_application_test.cpp
+++ b/src/mongo/db/s/resharding/resharding_oplog_session_application_test.cpp
@@ -79,7 +79,11 @@ public:
 
         MongoDOperationContextSession ocs(opCtx);
         auto txnParticipant = TransactionParticipant::get(opCtx);
-        txnParticipant.beginOrContinue(opCtx, txnNumber, boost::none, boost::none);
+        txnParticipant.beginOrContinue(opCtx,
+                                       txnNumber,
+                                       boost::none /* autocommit */,
+                                       boost::none /* startTransaction */,
+                                       boost::none /* txnRetryCounter */);
 
         WriteUnitOfWork wuow(opCtx);
         auto opTime = repl::getNextOpTime(opCtx);
@@ -102,8 +106,11 @@ public:
 
         MongoDOperationContextSession ocs(opCtx);
         auto txnParticipant = TransactionParticipant::get(opCtx);
-        txnParticipant.beginOrContinue(
-            opCtx, txnNumber, false /* autocommit */, true /* startTransaction */);
+        txnParticipant.beginOrContinue(opCtx,
+                                       txnNumber,
+                                       false /* autocommit */,
+                                       true /* startTransaction */,
+                                       boost::none /* txnRetryCounter */);
 
         txnParticipant.unstashTransactionResources(opCtx, "prepareTransaction");
 
@@ -123,8 +130,11 @@ public:
 
         MongoDOperationContextSession ocs(opCtx);
         auto txnParticipant = TransactionParticipant::get(opCtx);
-        txnParticipant.beginOrContinue(
-            opCtx, txnNumber, false /* autocommit */, boost::none /* startTransaction */);
+        txnParticipant.beginOrContinue(opCtx,
+                                       txnNumber,
+                                       false /* autocommit */,
+                                       boost::none /* startTransaction */,
+                                       boost::none /* txnRetryCounter */);
 
         txnParticipant.unstashTransactionResources(opCtx, "abortTransaction");
         txnParticipant.abortTransaction(opCtx);
@@ -255,7 +265,11 @@ public:
 
         MongoDOperationContextSession ocs(opCtx);
         auto txnParticipant = TransactionParticipant::get(opCtx);
-        txnParticipant.beginOrContinue(opCtx, txnNumber, boost::none, boost::none);
+        txnParticipant.beginOrContinue(opCtx,
+                                       txnNumber,
+                                       boost::none /* autocommit */,
+                                       boost::none /* startTransaction */,
+                                       boost::none /* txnRetryCounter */);
         ASSERT_TRUE(bool(txnParticipant.checkStatementExecuted(opCtx, stmtId)));
     }
 };
diff --git a/src/mongo/db/s/resharding/resharding_txn_cloner_test.cpp b/src/mongo/db/s/resharding/resharding_txn_cloner_test.cpp
index e302f943693..ba1efd5a017 100644
--- a/src/mongo/db/s/resharding/resharding_txn_cloner_test.cpp
+++ b/src/mongo/db/s/resharding/resharding_txn_cloner_test.cpp
@@ -243,9 +243,17 @@ protected:
         auto txnParticipant = TransactionParticipant::get(opCtx);
         ASSERT(txnParticipant);
         if (multiDocTxn) {
-            txnParticipant.beginOrContinue(opCtx, txnNum, false, true);
+            txnParticipant.beginOrContinue(opCtx,
+                                           txnNum,
+                                           false /* autocommit */,
+                                           true /* startTransaction */,
+                                           boost::none /* txnRetryCounter */);
         } else {
-            txnParticipant.beginOrContinue(opCtx, txnNum, boost::none, boost::none);
+            txnParticipant.beginOrContinue(opCtx,
+                                           txnNum,
+                                           boost::none /* autocommit */,
+                                           boost::none /* startTransaction */,
+                                           boost::none /* txnRetryCounter */);
         }
     }
 
@@ -389,8 +397,11 @@ protected:
 
         MongoDOperationContextSession ocs(opCtx);
         auto txnParticipant = TransactionParticipant::get(opCtx);
-        txnParticipant.beginOrContinue(
-            opCtx, txnNumber, false /* autocommit */, true /* startTransaction */);
+        txnParticipant.beginOrContinue(opCtx,
+                                       txnNumber,
+                                       false /* autocommit */,
+                                       true /* startTransaction */,
+                                       boost::none /* txnRetryCounter */);
 
         txnParticipant.unstashTransactionResources(opCtx, "prepareTransaction");
 
@@ -410,8 +421,11 @@ protected:
 
         MongoDOperationContextSession ocs(opCtx);
         auto txnParticipant = TransactionParticipant::get(opCtx);
-        txnParticipant.beginOrContinue(
-            opCtx, txnNumber, false /* autocommit */, boost::none /* startTransaction */);
+        txnParticipant.beginOrContinue(opCtx,
+                                       txnNumber,
+                                       false /* autocommit */,
+                                       boost::none /* startTransaction */,
+                                       boost::none /* txnRetryCounter */);
 
         txnParticipant.unstashTransactionResources(opCtx, "abortTransaction");
         txnParticipant.abortTransaction(opCtx);
diff --git a/src/mongo/db/s/resharding_destined_recipient_test.cpp b/src/mongo/db/s/resharding_destined_recipient_test.cpp
index 98d534ca694..6592bb4f35d 100644
--- a/src/mongo/db/s/resharding_destined_recipient_test.cpp
+++ b/src/mongo/db/s/resharding_destined_recipient_test.cpp
@@ -70,7 +70,11 @@ void runInTransaction(OperationContext* opCtx, Callable&& func) {
 
     auto txnParticipant = TransactionParticipant::get(opCtx);
     ASSERT(txnParticipant);
-    txnParticipant.beginOrContinue(opCtx, txnNum, false, true);
+    txnParticipant.beginOrContinue(opCtx,
+                                   txnNum,
+                                   false /* autocommit */,
+                                   true /* startTransaction */,
+                                   boost::none /* txnRetryCounter */);
     txnParticipant.unstashTransactionResources(opCtx, "SetDestinedRecipient");
 
     func();
diff --git a/src/mongo/db/s/session_catalog_migration_destination.cpp b/src/mongo/db/s/session_catalog_migration_destination.cpp
index 80226a9c463..8350f974581 100644
--- a/src/mongo/db/s/session_catalog_migration_destination.cpp
+++ b/src/mongo/db/s/session_catalog_migration_destination.cpp
@@ -247,7 +247,11 @@ ProcessOplogResult processSessionOplog(const BSONObj& oplogBSON,
     auto txnParticipant = TransactionParticipant::get(opCtx);
 
     try {
-        txnParticipant.beginOrContinue(opCtx, result.txnNum, boost::none, boost::none);
+        txnParticipant.beginOrContinue(opCtx,
+                                       result.txnNum,
+                                       boost::none /* autocommit */,
+                                       boost::none /* startTransaction */,
+                                       boost::none /* txnRetryCounter */);
         if (txnParticipant.checkStatementExecuted(opCtx, stmtIds.front())) {
             // Skip the incoming statement because it has already been logged locally
             return lastResult;
diff --git a/src/mongo/db/s/session_catalog_migration_destination_test.cpp b/src/mongo/db/s/session_catalog_migration_destination_test.cpp
index 7d770e47cff..446fedd435a 100644
--- a/src/mongo/db/s/session_catalog_migration_destination_test.cpp
+++ b/src/mongo/db/s/session_catalog_migration_destination_test.cpp
@@ -183,7 +183,11 @@ public:
         opCtx->setTxnNumber(txnNum);
         MongoDOperationContextSession ocs(opCtx);
         auto txnParticipant = TransactionParticipant::get(opCtx);
-        txnParticipant.beginOrContinue(opCtx, txnNum, boost::none, boost::none);
+        txnParticipant.beginOrContinue(opCtx,
+                                       txnNum,
+                                       boost::none /* autocommit */,
+                                       boost::none /* startTransaction */,
+                                       boost::none /* txnRetryCounter */);
     }
 
     void checkOplog(const repl::OplogEntry& originalOplog, const repl::OplogEntry& oplogToCheck) {
@@ -250,8 +254,11 @@ public:
             initializeOperationSessionInfo(innerOpCtx.get(), insertBuilder.obj(), true, true, true);
             MongoDOperationContextSession sessionTxnState(innerOpCtx.get());
             auto txnParticipant = TransactionParticipant::get(innerOpCtx.get());
-            txnParticipant.beginOrContinue(
-                innerOpCtx.get(), *sessionInfo.getTxnNumber(), boost::none, boost::none);
+            txnParticipant.beginOrContinue(innerOpCtx.get(),
+                                           *sessionInfo.getTxnNumber(),
+                                           boost::none /* autocommit */,
+                                           boost::none /* startTransaction */,
+                                           boost::none /* txnRetryCounter */);
 
             const auto reply = write_ops_exec::performInserts(innerOpCtx.get(), insertRequest);
             ASSERT(reply.results.size() == 1);
@@ -1879,7 +1886,11 @@ TEST_F(SessionCatalogMigrationDestinationTest,
         auto txnParticipant = TransactionParticipant::get(opCtx);
 
         txnParticipant.refreshFromStorageIfNeeded(opCtx);
-        txnParticipant.beginOrContinue(opCtx, 3, boost::none, boost::none);
+        txnParticipant.beginOrContinue(opCtx,
+                                       3,
+                                       boost::none /* autocommit */,
+                                       boost::none /* startTransaction */,
+                                       boost::none /* txnRetryCounter */);
     }
 
     OperationSessionInfo sessionInfo2;
diff --git a/src/mongo/db/s/txn_two_phase_commit_cmds.cpp b/src/mongo/db/s/txn_two_phase_commit_cmds.cpp
index b228c301d95..3fce3dda429 100644
--- a/src/mongo/db/s/txn_two_phase_commit_cmds.cpp
+++ b/src/mongo/db/s/txn_two_phase_commit_cmds.cpp
@@ -314,7 +314,8 @@ public:
                 txnParticipant.beginOrContinue(opCtx,
                                                *opCtx->getTxnNumber(),
                                                false /* autocommit */,
-                                               boost::none /* startTransaction */);
+                                               boost::none /* startTransaction */,
+                                               *opCtx->getTxnRetryCounter());
 
                 if (txnParticipant.transactionIsCommitted())
                     return;
@@ -336,7 +337,8 @@ public:
                 txnParticipant.beginOrContinue(opCtx,
                                                *opCtx->getTxnNumber(),
                                                false /* autocommit */,
-                                               boost::none /* startTransaction */);
+                                               boost::none /* startTransaction */,
+                                               *opCtx->getTxnRetryCounter());
 
                 invariant(!txnParticipant.transactionIsOpen(),
                           "The participant should not be in progress after we waited for the "