DanielLeens commented on issue #12382: URL: https://github.com/apache/seatunnel/issues/12382#issuecomment-5917737506
Classification: F / PostgreSQL CDC restore LSN handling, covered by an active PR. The new controls materially narrow the investigation. The uninstrumented failures establish that the loss is not caused by the tracing agent, and the locator logs plus the `+312`-byte movement provide a concrete ambiguous-boundary hypothesis. They do not by themselves prove the causal link to `id=15`, so it is important to keep that distinction explicit. I checked the current source and #12454. That active PR owns the same recovery boundary: it adds message-type-aware PostgreSQL offset handling and deterministic shared-LSN coverage while retaining reads of existing savepoints. It is not merged, so this issue is not fixed in `dev` yet and no duplicate production fix should be opened. Please add the proposed real-run comparison to the evidence for #12454: record the inserted row's WAL LSN range, the stored offset, and the locator's `already processed` LSN in the same failing run. The result should confirm or reject the hypothesized equality before we treat the test fixture as proof of this production symptom. Keep the two-revision/load matrix and reached/not-reached accounting intact. -- 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]
