DanielLeens commented on issue #11433: URL: https://github.com/apache/seatunnel/issues/11433#issuecomment-5199226222
Thanks for the follow-up details. The Doris version (`2.1.11`) and the comparison with `sink.enable-2pc = false` help narrow this down quite a bit. That makes the issue more specific: this is not just general Doris eventual-consistency behavior on unique-key tables. The real gap is the SeaTunnel 2PC completion contract. When 2PC is enabled, SeaTunnel can currently report the sink side as finished once Doris reaches `COMMITTED`, even though the downstream-visible state is still `VISIBLE`-pending. Your note about enabling 2PC specifically to reduce source/sink inconsistency after worker crashes is exactly why this distinction matters. So the right next step is still the fix in PR #11562, but the acceptance bar should be explicit: 1. verify the behavior against Doris `2.1.11`; 2. confirm `sink.enable-2pc = true` waits for the intended visible/usable state before treating the sink commit as complete; 3. keep the `sink.enable-2pc = false` path unchanged, since that path already matches the current non-2PC expectation. This issue should stay open until that behavior is covered and verified on the PR path. -- 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]
