-
Type:
Bug
-
Resolution: Fixed
-
Priority:
Major - P3
-
Affects Version/s: None
-
Component/s: None
-
None
-
Storage Execution
-
Fully Compatible
-
ALL
-
Storage Execution 2026-11-23
-
200
-
None
-
None
-
None
-
None
-
None
-
None
-
None
In the normal flow a replicated table drop should only be applied after the oldest timestamp has advanced past the drop timestamp, and so there should not be any remaining readers pinning the table. We therefore chose to make this a fatal error, as it's most likely not a transient condition and it's better to just restart a secondary than stall replication forever.
Magic Restore turns out to be an exception to this. The collection drop and table drop can be applied with no delay, and while most of the possible readers are stopped during magic restore, FTDC is not. FTDC can briefly pin the collection catalog and thus a dropped table, making the invariant that there are no readers fail.
- is duplicated by
-
SERVER-133311 Should a replicated ident drop be fatal when a reader still holds the ident?
-
- Closed
-