SERVER-37051 Check for term change after fetching the queued metadata in ShardServerCatalogCacheLoader

(cherry picked from commit fe8f517a59d694b7577da564d19e4415e13831e8)

2 files changed, 18 insertions, 4 deletions

diff --git a/src/mongo/db/s/shard_server_catalog_cache_loader.cpp b/src/mongo/db/s/shard_server_catalog_cache_loader.cpp
index 3dc2366d3eb..52bba0249c9 100644
--- a/src/mongo/db/s/shard_server_catalog_cache_loader.cpp
+++ b/src/mongo/db/s/shard_server_catalog_cache_loader.cpp
@@ -690,7 +690,21 @@ void ShardServerCatalogCacheLoader::_schedulePrimaryGetChunksSince(
 
                 swCollectionAndChangedChunks =
                     _getLoaderMetadata(opCtx, nss, catalogCacheSinceVersion, termScheduled);
-                if (swCollectionAndChangedChunks.isOK()) {
+
+                const auto termAfterRefresh = [&] {
+                    stdx::lock_guard<stdx::mutex> lock(_mutex);
+                    return _term;
+                }();
+
+                if (termAfterRefresh != termScheduled) {
+                    // Raising a ConflictingOperationInProgress error here will cause the
+                    // CatalogCache to attempt the refresh as secondary instead of failing the
+                    // operation
+                    swCollectionAndChangedChunks = Status(
+                        ErrorCodes::ConflictingOperationInProgress,
+                        str::stream() << "Replication stepdown occurred during refresh for  '"
+                                      << nss.toString());
+                } else if (swCollectionAndChangedChunks.isOK()) {
                     // After finding metadata remotely, we must have found metadata locally.
                     invariant(!collAndChunks.changedChunks.empty());
                 }
@@ -760,12 +774,12 @@ StatusWith<CollectionAndChangedChunks> ShardServerCatalogCacheLoader::_getLoader
     OperationContext* opCtx,
     const NamespaceString& nss,
     const ChunkVersion& catalogCacheSinceVersion,
-    const long long term) {
+    long long expectedTerm) {
 
     // Get the enqueued metadata first. Otherwise we could miss data between reading persisted and
     // enqueued, if an enqueued task finished after the persisted read but before the enqueued read.
 
-    auto enqueuedRes = _getEnqueuedMetadata(nss, catalogCacheSinceVersion, term);
+    auto enqueuedRes = _getEnqueuedMetadata(nss, catalogCacheSinceVersion, expectedTerm);
     bool tasksAreEnqueued = std::move(enqueuedRes.first);
     CollectionAndChangedChunks enqueued = std::move(enqueuedRes.second);
 
diff --git a/src/mongo/db/s/shard_server_catalog_cache_loader.h b/src/mongo/db/s/shard_server_catalog_cache_loader.h
index 9b4e327caa8..ff246c0666e 100644
--- a/src/mongo/db/s/shard_server_catalog_cache_loader.h
+++ b/src/mongo/db/s/shard_server_catalog_cache_loader.h
@@ -412,7 +412,7 @@ private:
         OperationContext* opCtx,
         const NamespaceString& nss,
         const ChunkVersion& catalogCacheSinceVersion,
-        const long long term);
+        long long expectedTerm);
 
     /**
      * Loads chunk metadata from all in-memory tasks enqueued to update the shard persisted metadata