DDL coordinators share a common retry logic that enumerates the exceptions that must be retried. All other errors result in the operation being interrupted and the coordinator getting released.
Since the rename coordinator must never be released in case of unexpected errors, add a specialized logic to indefinitely retry the execution.
- is related to
-
SERVER-65340 Operations hang when re-using dropped unsharded collections
-
- Closed
-
- related to
-
SERVER-117624 Long running index builds may create temporary unavailability upon rename in sharded clusters
-
- Needs Scheduling
-