SERVER-57378 Fix data-race in the resharding commit monitor unit-tests

2 files changed, 14 insertions, 3 deletions

diff --git a/src/mongo/db/s/resharding/resharding_coordinator_commit_monitor.cpp b/src/mongo/db/s/resharding/resharding_coordinator_commit_monitor.cpp
index 775fba78a47..91c5868fccc 100644
--- a/src/mongo/db/s/resharding/resharding_coordinator_commit_monitor.cpp
+++ b/src/mongo/db/s/resharding/resharding_coordinator_commit_monitor.cpp
@@ -54,6 +54,7 @@ namespace resharding {
 namespace {
 
 MONGO_FAIL_POINT_DEFINE(failQueryingRecipients);
+MONGO_FAIL_POINT_DEFINE(hangBeforeQueryingRecipients);
 
 BSONObj makeCommandObj(const NamespaceString& ns) {
     auto command = _shardsvrReshardingOperationTime(ns);
@@ -139,6 +140,8 @@ Milliseconds CoordinatorCommitMonitor::_queryMaxRemainingOperationTimeForRecipie
                             ReadPreferenceSetting(ReadPreference::PrimaryOnly),
                             Shard::RetryPolicy::kIdempotent);
 
+    hangBeforeQueryingRecipients.pauseWhileSet();
+
     auto maxRemainingTime = Milliseconds(0);
     while (!ars.done()) {
         iassert(ErrorCodes::CallbackCanceled,
diff --git a/src/mongo/db/s/resharding/resharding_coordinator_commit_monitor_test.cpp b/src/mongo/db/s/resharding/resharding_coordinator_commit_monitor_test.cpp
index 2a123e257a4..63d56feb894 100644
--- a/src/mongo/db/s/resharding/resharding_coordinator_commit_monitor_test.cpp
+++ b/src/mongo/db/s/resharding/resharding_coordinator_commit_monitor_test.cpp
@@ -173,9 +173,17 @@ TEST_F(CoordinatorCommitMonitorTest, UnblocksWhenRecipientsWithinCommitThreshold
 }
 
 TEST_F(CoordinatorCommitMonitorTest, UnblocksWhenCancellationTokenIsCancelled) {
-    auto future = getCommitMonitor()->waitUntilRecipientsAreWithinCommitThreshold();
-    runOnMockingNextResponse([this] { cancelMonitor(); });
-    respondWithNotReadyToCommit();
+    auto future = [&] {
+        FailPointEnableBlock fp("hangBeforeQueryingRecipients");
+        auto future = getCommitMonitor()->waitUntilRecipientsAreWithinCommitThreshold();
+        fp->waitForTimesEntered(fp.initialTimesEntered() + 1);
+        // Cancels the monitor before waiting for the recipients to respond to the query. Once the
+        // fail-point is disabled, the monitor should cancel pending network operations and make the
+        // future ready. Thus, we do not block to run commands on behalf of the mocked network here.
+        cancelMonitor();
+        return future;
+    }();
+
     future.get();
 }