dawidwys commented on a change in pull request #18086:
URL: https://github.com/apache/flink/pull/18086#discussion_r768539890
##########
File path:
flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/CheckpointCoordinator.java
##########
@@ -1664,10 +1675,13 @@ public boolean restoreSavepoint(
allowNonRestored,
checkpointProperties);
- // register shared state - even before adding the checkpoint to the
store
- // because the latter might trigger subsumption so the ref counts must
be up-to-date
- savepoint.registerSharedStatesAfterRestored(
- completedCheckpointStore.getSharedStateRegistry());
+ // in claim mode we should not register any shared handles
+ if (restoreSettings.getRestoreMode() != RestoreMode.NO_CLAIM) {
+ // register shared state - even before adding the checkpoint to
the store
+ // because the latter might trigger subsumption so the ref counts
must be up-to-date
+ savepoint.registerSharedStatesAfterRestored(
Review comment:
I don't think so. If we restore from a savepoint (or externalized
checkpoints, it actually goes through the same codepath) we do not have any
prior checkpoints, the one added from `restoreSavepoint` will be the only one.
If we are doing a failover, the `RestoreMode` does not apply in such a case.
--
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]
