-
Type:
Bug
-
Resolution: Fixed
-
Priority:
Major - P3
-
Affects Version/s: 6.0.20, 8.0.5, 7.0.17
-
Component/s: None
-
None
-
Catalog and Routing
-
Fully Compatible
-
ALL
-
CAR Team 2025-03-03, CAR Team 2025-03-17
-
None
-
None
-
None
-
None
-
None
-
None
-
None
As part of its commit phase, the Sharding DDL coordinator for dropDatabase(dbName) requests (through a remote command) to each data-bearing shard the execution of a routine that removes from its local catalog and storage of any meta/data related to dbName; such a routine includes the generation of a c op entry, which the change streams machinery interprets as "a dropDatabase(dbName) user request has been completed" event.
We should ensure that the commit of a dropDatabase(dbName) request on a sharded cluster is matched by a single user-visible op entry generated by the primary shard of dbName (the ones generated by other data-bearing shards may be hidden through the {fromMigrate: true} op entry field).
- is depended on by
-
TOOLS-3776 Investigate changes in SERVER-101522: A sharded cluster should generate a single user-visible change event for "dbName" when dropDatabase("dbName") is committed.
-
- Accepted
-
- is related to
-
SERVER-34554 Database drops do not have a total ordering in a change stream
-
- Backlog
-
-
SERVER-101540 Add a test case to verify the drop of a database is exactly matched by a single user-visible dropDatabase change event when run against a sharded cluster
-
- Closed
-