-
Type: Bug
-
Resolution: Fixed
-
Priority: Major - P3
-
Affects Version/s: None
-
Component/s: Replication
-
None
-
Fully Compatible
-
ALL
-
Repl 2020-02-24
-
21
Before SERVER-45574, OplogFetcher used RemoteCommandRetryScheduler on the ReplicationExecutor to run oplog fetching queries. And all connections in the executor thread pool run LogicalTimeMetadataHook on incoming command replies to advance the logical clock.
After SERVER-45574, OplogFetcher uses DBClientConnection and DBClientCursor for oplog fetching. So we should add cluster time metadata handling to the DBClientConnection's metadata reader. Otherwise, the syncing node could advance the commit point (in processMetadata) beyond the logical clock.
- related to
-
SERVER-35663 Replication recovery does not update the logical clock
- Closed
-
SERVER-27749 Integrate LogicalTimeMetadata with metadata processing
- Closed
-
SERVER-45574 Replace OplogFetcher with NewOplogFetcher
- Closed