SERVER-70571 Set reasonable value for shard split timeout

5 files changed, 16 insertions, 7 deletions

diff --git a/buildscripts/resmokeconfig/suites/shard_split_kill_primary_jscore_passthrough.yml b/buildscripts/resmokeconfig/suites/shard_split_kill_primary_jscore_passthrough.yml
index 9cf6b5b61fd..bb6c5699e2f 100644
--- a/buildscripts/resmokeconfig/suites/shard_split_kill_primary_jscore_passthrough.yml
+++ b/buildscripts/resmokeconfig/suites/shard_split_kill_primary_jscore_passthrough.yml
@@ -221,6 +221,8 @@ executor:
               blockTimeMS: 250
         shardSplitGarbageCollectionDelayMS: 1
         ttlMonitorSleepSecs: 1
+        # Unclean shutdown may cause blockTimestamp catch up to be very long
+        shardSplitTimeoutMS: 60000
       auth: ''
       keyFile: *keyFile
     num_nodes_per_replica_set: 3
diff --git a/jstests/serverless/shard_split_concurrent_reads_on_donor_blocking.js b/jstests/serverless/shard_split_concurrent_reads_on_donor_blocking.js
index bbd9138bb70..409b2ecdcb4 100644
--- a/jstests/serverless/shard_split_concurrent_reads_on_donor_blocking.js
+++ b/jstests/serverless/shard_split_concurrent_reads_on_donor_blocking.js
@@ -65,7 +65,9 @@ const tenantId = "tenantId";
 const test = new ShardSplitTest({
     recipientTagName: "recipientTag",
     recipientSetName: "recipientSet",
-    quickGarbageCollection: true
+    quickGarbageCollection: true,
+    // Increase timeout because blocking in the critical section contributes to operation latency.
+    nodeOptions: {setParameter: {shardSplitTimeoutMS: 100000}}
 });
 test.addRecipientNodes();
 
diff --git a/jstests/serverless/shard_split_concurrent_writes_on_donor_blocking.js b/jstests/serverless/shard_split_concurrent_writes_on_donor_blocking.js
index 01de40772a1..b6e53662901 100644
--- a/jstests/serverless/shard_split_concurrent_writes_on_donor_blocking.js
+++ b/jstests/serverless/shard_split_concurrent_writes_on_donor_blocking.js
@@ -29,7 +29,9 @@ const tenantMigrationTest = new ShardSplitTest({
     recipientSetName,
     quickGarbageCollection: true,
     allowStaleReadsOnDonor: true,
-    initiateWithShortElectionTimeout: true
+    initiateWithShortElectionTimeout: true,
+    // Increase timeout because blocking in the critical section contributes to operation latency.
+    nodeOptions: {setParameter: {shardSplitTimeoutMS: 100000}}
 });
 
 const donorPrimary = tenantMigrationTest.getDonorPrimary();
diff --git a/src/mongo/db/repl/repl_server_parameters.idl b/src/mongo/db/repl/repl_server_parameters.idl
index 65b87404f7e..14637377c6c 100644
--- a/src/mongo/db/repl/repl_server_parameters.idl
+++ b/src/mongo/db/repl/repl_server_parameters.idl
@@ -533,14 +533,13 @@ server_parameters:
         validator:
             gte: 1
 
-
     shardSplitTimeoutMS:
         description: >-
             Period of time, in milliseconds, after which a shard split should be interrupted.
         set_at: [ startup, runtime ]
         cpp_vartype: AtomicWord<int>
         cpp_varname: shardSplitTimeoutMS
-        default: 3600000
+        default: 10000
         validator:
             gte: 1
 
@@ -720,4 +719,4 @@ feature_flags:
                      allow kDowngradingFromLatestToLastLTS -> kUpgrading -> kUpgraded path.
         cpp_varname: feature_flags::gDowngradingToUpgrading
         default: false
-        
\ No newline at end of file
+
diff --git a/src/mongo/db/serverless/shard_split_donor_service.cpp b/src/mongo/db/serverless/shard_split_donor_service.cpp
index cabaeb0d722..80697cbe0f4 100644
--- a/src/mongo/db/serverless/shard_split_donor_service.cpp
+++ b/src/mongo/db/serverless/shard_split_donor_service.cpp
@@ -1039,8 +1039,7 @@ ShardSplitDonorService::DonorStateMachine::_handleErrorOrEnterAbortedState(
     ON_BLOCK_EXIT([&] {
         stdx::lock_guard<Latch> lg(_mutex);
         if (_abortSource) {
-            // Cancel source to ensure all child threads (RSM monitor, etc)
-            // terminate.
+            // Cancel source to ensure all child threads (RSM monitor, etc) terminate.
             _abortSource->cancel();
         }
     });
@@ -1049,6 +1048,11 @@ ShardSplitDonorService::DonorStateMachine::_handleErrorOrEnterAbortedState(
         stdx::lock_guard<Latch> lg(_mutex);
         if (isAbortedDocumentPersistent(lg, _stateDoc)) {
             // The document is already in aborted state. No need to write it.
+            LOGV2(8423376,
+                  "Shard split already aborted.",
+                  "id"_attr = _migrationId,
+                  "abortReason"_attr = _abortReason.value());
+
             return ExecutorFuture(**executor,
                                   DurableState{ShardSplitDonorStateEnum::kAborted, _abortReason});
         }