SEZ9 commented on issue #11676: URL: https://github.com/apache/seatunnel/issues/11676#issuecomment-5215602362
Thanks for the detailed report and for confirming this is distinct from #10574 (Kafka restore-offset) and the 2023 restore-path issues. Since you verified it on current `dev` with a regression test, could you share that test (or open a draft PR with it)? That would greatly speed up review and validation of a fix. Could you attach the full enumerator/reader logs around the restart (including reader registration and any `IncrementalSourceEnumerator` assignment logs)? Confirming whether the reader's split request arrives before the enumerator finishes restoring state would help pinpoint the race in the restored-split re-assignment path. As a possible workaround while this is investigated: does a full job restart (rather than an in-place failure recovery) resume data flow correctly from the checkpoint/savepoint? Knowing this would help scope the bug to the enumerator restore/assignment race versus checkpoint state itself. <!-- streview-comment:70 --> -- 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]
