SEZ9 commented on PR #13060: URL: https://github.com/apache/seatunnel/pull/13060#issuecomment-6817545831
@zephyr45 Thanks for the update on `7c272c5a`, and for inspecting the attempt 1 logs before rerunning only the two failed setup/download jobs in run `41960786230`. On F4 (Spark task retries getting a fresh transactional namespace): separating Kafka's retry-stable partition index from the task-attempt index used by other sinks, via the default `SinkWriter.Context#getStableSubtaskIndex()` in `9e4aad72`, sounds like the right direction. I'll confirm it against the diff. I'll also go through your corrected compatibility assessment and your per-finding status against the diff for the remaining items from my earlier review: - F1: Flink contiguous-chunk ListState redistribution vs. the modulo assumption in `validateRestoredStateOwner` - F2: scale-down restore when `transaction_prefix` is not configured and each writer has a random prefix - F3: a scale-up writer starting its transaction sequence at 1 regardless of checkpoint id - F7: aligning the `transaction_prefix` option row and the engine scope of the rescaling claim in the Kafka sink docs For F5/F6 (repeated legacy shared-namespace fencing), I'm fine treating that as a deferred optimization in a separate follow-up rather than in this PR. No new code asks on this head; I'll update my review once I've finished verifying the above. <!-- streview-comment:2287 --> -- 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]
