DanielLeens commented on issue #10571: URL: https://github.com/apache/seatunnel/issues/10571#issuecomment-4953706298
Thanks for reporting this, and thanks to the earlier analysis already captured in the thread. I rechecked the issue against the current discussion and the fix now being prepared in PR #11410. This still looks like a real bug in SQL Server CDC restore/resume offset handling after restart, not a Paimon sink-side problem. The key boundary here is that after Flink restore, SeaTunnel needs to resume from the full SQL Server CDC position, not a reduced one. If the restored state only keeps part of the position, events sharing the same commit/change boundary can be treated as already processed. That matches the symptom reported here: the first insert after restart may still be emitted, while later inserts can be skipped incorrectly. PR #11410 is targeting exactly that path by preserving `commit_lsn`, `change_lsn`, and `event_serial_no` in the incremental split state and by comparing the complete SQL Server position on resume. So this issue should stay open until that fix lands and is verified, but it now has a concrete implementation path. Once #11410 is merged into `dev`, the most useful next step will be to retest the same restart scenario against the latest `dev` build and confirm whether post-restart inserts continue arriving normally after the first row. -- 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]
