In steady state oplog fetching, the oplog fetcher will retry (once by default) on the same node it is already syncing from. If we already know the node is down (due to missed heartbeats), we should instead fail and re-run sync source selection. This will result in a shorter time to majority write availability in the case where the retry takes a long time.
- is related to
-
SERVER-64635 Coverity analysis defect 121898: Parse warning
- Closed
- related to
-
SERVER-63792 Improve coverage of blackholing network requests
- Open
-
SERVER-63418 Oplog fetcher should abort if node goes down
- Backlog