SERVER-47018 Attach topology version to shutdown errors while in quiesce mode

3 files changed, 37 insertions, 0 deletions

diff --git a/src/mongo/s/commands/strategy.cpp b/src/mongo/s/commands/strategy.cpp
index 2334232a604..23b21731fa6 100644
--- a/src/mongo/s/commands/strategy.cpp
+++ b/src/mongo/s/commands/strategy.cpp
@@ -77,6 +77,7 @@
 #include "mongo/s/cluster_commands_helpers.h"
 #include "mongo/s/commands/cluster_explain.h"
 #include "mongo/s/grid.h"
+#include "mongo/s/mongos_topology_coordinator.h"
 #include "mongo/s/query/cluster_cursor_manager.h"
 #include "mongo/s/query/cluster_find.h"
 #include "mongo/s/session_catalog_router.h"
@@ -891,6 +892,26 @@ void runCommand(OperationContext* opCtx,
     }
 }
 
+/**
+ * Attaches the topology version to the response.
+ */
+void attachTopologyVersionDuringShutdown(OperationContext* opCtx,
+                                         const DBException& ex,
+                                         BSONObjBuilder* errorBuilder) {
+    // Only attach the topology version if the mongos is in quiesce mode. If the mongos is in
+    // quiesce mode, this shutdown error is due to mongos rather than a shard.
+    auto code = ex.code();
+    if (code && ErrorCodes::isA<ErrorCategory::ShutdownError>(code)) {
+        if (auto mongosTopCoord = MongosTopologyCoordinator::get(opCtx);
+            mongosTopCoord && mongosTopCoord->inQuiesceMode()) {
+            // Append the topology version to the response.
+            const auto topologyVersion = mongosTopCoord->getTopologyVersion();
+            BSONObjBuilder topologyVersionBuilder(errorBuilder->subobjStart("topologyVersion"));
+            topologyVersion.serialize(&topologyVersionBuilder);
+        }
+    }
+}
+
 }  // namespace
 
 DbResponse Strategy::queryOp(OperationContext* opCtx, const NamespaceString& nss, DbMessage* dbm) {
@@ -1078,10 +1099,13 @@ DbResponse Strategy::clientCommand(OperationContext* opCtx, const Message& m) {
         if (propagateException) {
             throw;
         }
+
         reply->reset();
         auto bob = reply->getBodyBuilder();
         CommandHelpers::appendCommandStatusNoThrow(bob, ex.toStatus());
         appendRequiredFieldsToResponse(opCtx, &bob);
+
+        attachTopologyVersionDuringShutdown(opCtx, ex, &errorBuilder);
         bob.appendElements(errorBuilder.obj());
     }
 
diff --git a/src/mongo/s/mongos_topology_coordinator.h b/src/mongo/s/mongos_topology_coordinator.h
index eccf83c8a31..6d9ddc0f286 100644
--- a/src/mongo/s/mongos_topology_coordinator.h
+++ b/src/mongo/s/mongos_topology_coordinator.h
@@ -81,6 +81,12 @@ public:
         return _topologyVersion;
     }
 
+    bool inQuiesceMode() const {
+        stdx::lock_guard lk(_mutex);
+        return _inQuiesceMode;
+    }
+
+
 private:
     using SharedPromiseOfMongosIsMasterResponse =
         SharedPromise<std::shared_ptr<const MongosIsMasterResponse>>;
diff --git a/src/mongo/s/server.cpp b/src/mongo/s/server.cpp
index 05d228e5ba3..cf86f70a63d 100644
--- a/src/mongo/s/server.cpp
+++ b/src/mongo/s/server.cpp
@@ -138,6 +138,8 @@ MONGO_FAIL_POINT_DEFINE(failReplicaSetChangeConfigServerUpdateHook);
 
 namespace {
 
+MONGO_FAIL_POINT_DEFINE(pauseWhileKillingOperationsAtShutdown);
+
 #if defined(_WIN32)
 const ntservice::NtServiceDefaultStrings defaultServiceStrings = {
     L"MongoS", L"MongoDB Router", L"MongoDB Sharding Router"};
@@ -315,6 +317,11 @@ void cleanupTask(const ShutdownTaskArgs& shutdownArgs) {
 
         if (serviceContext) {
             serviceContext->setKillAllOperations();
+
+            if (MONGO_unlikely(pauseWhileKillingOperationsAtShutdown.shouldFail())) {
+                LOGV2(4701800, "pauseWhileKillingOperationsAtShutdown failpoint enabled");
+                sleepsecs(1);
+            }
         }
 
         // Perform all shutdown operations after setKillAllOperations is called in order to ensure