DanielLeens commented on PR #13060: URL: https://github.com/apache/seatunnel/pull/13060#issuecomment-6914638500
Thanks, @zephyr45 and @SEZ9. I checked the same head, `7c272c5a`. The F3 warning is already present in both Kafka sink guides next to the stateful recovery description: when there is no retained state, the new writer only fences sequence one, and a later abandoned transaction can hold up `read_committed` consumers until `transaction.timeout.ms`. A checkpoint counter cannot identify an ID that was never captured in checkpoint state. For F5/F6, `KafkaSinkWriter.abortRestoredTransactions` deduplicates per writer, then probes each prefix assigned to that writer. It does not elect one owner for the legacy shared prefix across writers, so the one-time migration cost is still there. That is the distinction I would keep in the review. One update to my earlier CI note: [Apache Build now reports success](https://github.com/apache/seatunnel/runs/127712599445), and the [fork Build](https://github.com/zephyr45/seatunnel/actions/runs/41960786230) succeeded at the same head. The PR is conflict-free and GitHub now reports review required. I would leave the remaining F3/F5/F6 review disposition to @SEZ9's pending recheck and the final approval to a maintainer. -- 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]
