Oplog sampling can take a long time during startup depending on the size of the oplog. Since the restore procedure includes multiple rounds of restarting the server, we should opt out of oplog sampling until the restore is completed. We currently do this when recoverFromOplogAsStandalone is set to true (since it sets disallowUserWrites) but not during other parts of the restore procedure.
Originally we thought SERVER-60761 changed this behavior but on further inspection it seems like oplog sampling was never disabled during these sections of restore, which can lead to HELP tickets around a long restore procedure.
- is related to
-
SERVER-60761 Global readOnly flag should not be switched on and off after startup
-
- Closed
-
- related to
-
SERVER-94725 Remove redundant conditional check for oplog truncate marker creation
-
- Open
-
-
-
- Backlog
-