SEZ9 commented on PR #11677: URL: https://github.com/apache/seatunnel/pull/11677#issuecomment-6183297556
@li3zhi4 Thanks for the pointer and the link to `c959fd1989`. The `substring(Math.min(logOffset, serverLogs.length()))` guard you describe is the right shape for PR11677-F7 / PR11677-F8 — a shorter log yielding an empty suffix and letting the Awaitility lambda retry is exactly what I was after. I'll confirm it against the diff on my next pass; I'm not asking for a forced-rotation test, and I appreciate you being upfront that the earlier run didn't exercise that path. Still open from the previous review: 1. **PR11677-F1 / F2 (SourceSplitEnumerator javadoc):** the rewritten contract applies to every enumerator implementation, but only the CDC base enumerator was hardened. Please either narrow the javadoc to what the engine actually guarantees today (noting the CDC-side hardening as implementation behavior), or outline how the other connectors' `addSplitsBack()` implementations are expected to conform. The first option is fine for this PR. 2. **PR11677-F3 / F4 (`if (running)` gate in `addSplitsBack()`):** splits returned before `run()` executes are dispatched only via the trailing `assignSplits()` call in `run()`. Please add a short comment documenting that invariant at the gate, and ideally a unit test covering the "addSplitsBack before run" path so it doesn't regress silently. 3. **PR11677-F5 (30-second wait for the injected sink failure):** this is still inconsistent with the 2-minute budgets elsewhere in the same test. Either align it to the same budget or explain why 30s is sufficient here. 4. **PR11677-F6 (replay before `DROP TRIGGER` takes effect):** a brief note on why the restart budget can't be exhausted in that window would be enough — or, if it can, a small adjustment so the failing insert can't be replayed more than once. Once those are addressed I'm happy to take another look. <!-- streview-comment:1761 --> -- 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]
