sunm2n commented on PR #23390: URL: https://github.com/apache/kafka/pull/23390#issuecomment-6485343485
A user reported on [KAFKA-21034](https://issues.apache.org/jira/browse/KAFKA-21034) that they reproduced this crash loop on 4.3.2 with `exactly_once_v2` and a persistent global store, and requested backports to 4.3.x and 4.4.x. I've asked for their reproduction to validate the patch against it. This matches the path addressed here: since KAFKA-20396 (#21946), `RocksDBStore#openDB` wraps the invalid-state failure in `TaskCorruptedException`, so the wipe added in KAFKA-17411 for `ProcessorStateException` is no longer reached on this path. @UladzislauBlok, following up on your suggestion to get other opinions on where the cleanup should live: would you and @nicktelford be comfortable keeping this PR focused on the fix and considering consolidation of the two wipe paths in a follow-up on trunk? The current patch reuses the existing cleanup in `GlobalStreamThread`, preserves closing stores before deleting the directory, and prevents double cleanup during initialization failure, with regression coverage. Given the backport request, I'd prefer to keep the broader refactoring separate. I'm happy to file the follow-up ticket. If that works for you, I'd appreciate a re-review. @mjsax @bbejeck, could one of you review this approach and advise on backports to 4.3.x and 4.4.x? Could you also add the `ci-approved` label? The CI run on the current head is still pending approval. -- 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]
