-
Type: Bug
-
Resolution: Fixed
-
Priority: Major - P3
-
Affects Version/s: None
-
Component/s: Replication
-
Fully Compatible
-
ALL
-
v5.0
-
Repl 2021-05-31, Repl 2021-06-14
-
121
During recovery, we assume that the oplogApplicationStartPoint exists in the oplog.
The oplogApplicationStartPoint derives from the stable timestamp, which is not guaranteed to be in the oplog. If the stable timestamp was greater than the oplogTruncateAfterPoint, then we move this point back to the truncate after point, which is guaranteed to be in the oplog. However, if the stable timestamp is less than the oplogTruncateAfterPoint, we don't do anything to fix the oplogApplicationStartPoint.
Because of SERVER-51158, we guarantee that an oplog entry less than or equal to the stable timestamp exists in the oplog, so we should move the oplogApplicationStartPoint back to that oplog entry.
- is related to
-
SERVER-51049 Cannot assume recovery timestamp can be found in oplog
- Closed
-
SERVER-51158 Must not truncate entire oplog before truncate point
- Closed