cchung100m commented on PR #11818:
URL: https://github.com/apache/ozone/pull/11818#issuecomment-6496551839
Hi @sodonnel,
Thanks for the review. I checked against the code:
- Force close from ReplicationManager does require all unique origins and
closes only the highest-sid QC replicas, as you said.
- The report handler path is different:
`AbstractContainerReportHandler#updateContainerState` closes a CLOSING or
QUASI_CLOSED container on a single CLOSED replica whose BCSID matches the
current sid, with no origin count. A unit test with a mocked `ContainerManager`
confirmed this.
- QUASI_CLOSED containers can be replicated when
`QuasiClosedStuckReplicationCheck.shouldHandleAsQuasiClosedStuck` is true.
- HDDS-12409 attributes it to datanode `applyTransaction` / ratis group
removal bugs and left the question of whether a CLOSED sid may change to a
separate Jira, which this PR addresses.
- The change also affects DELETING/DELETED: without it, the sid bump makes
`replica BCSID <= container BCSID` true (HDDS-14416), so the replica is deleted
instead of the container being resurrected. The new test covers this.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
