Currently chunk merges acquire the collection distributed lock. This means that it is not possible to run more than one merge at a time for the entire cluster or if there are any ongoing splits or balancing (again for the entire cluster).
This ticket is to figure out how to remove the dist lock acquisition from merges without causing the much more expensive moves to start failing, because the chunk being moved got merged.
- causes
-
SERVER-59929 unexpected slower update/insert operation bease of splitchunk and moveChunk
- Closed
- is depended on by
-
SERVER-57032 Acquire only local distLock for Sharding DDL coordinators
- Closed
-
SERVER-56654 Do not use the collection distributed lock for chunk splits
- Closed
- is duplicated by
-
SERVER-57115 Merge chunks should wait for any ongoing moves to complete
- Closed
-
SERVER-25359 Create collection-specific ResourceMutex map to allow split/merge/move chunk on different collections to proceed in parallel
- Closed
- is related to
-
SERVER-56786 There are three routing info refreshes and two chunk scans on the mergeChunks path
- Closed