DanielLeens commented on issue #12382: URL: https://github.com/apache/seatunnel/issues/12382#issuecomment-5979750533
Classification: F / Connector-V2 PostgreSQL CDC restore LSN handling, covered by active #12454. Thank you for the real-run comparison. The three missing-row runs now show the causal boundary directly: the stored commit-end LSN, the first LSN after restore, the locator LSN filtered as already processed, and the `id=15` transaction start are the same value. The post-#12454 runs hitting that boundary resume without filtering `id=15` and deliver it. The rate comparison is small, but this boundary evidence is decisive for the reported symptom. I rechecked the current #12454 implementation. Its offset carries the last processed message type while retaining reads of legacy savepoints, and its locator tests cover COMMIT, BEGIN, legacy-without-type, and an already processed INSERT at the shared-LSN boundary. The PostgreSQL CDC IT also now saves a quiet boundary, writes `id=15` while stopped, then proves both that recovered `id=15` and a post-reconnect `id=16` reach the sink. Please do not open a second production-fix PR. If any further deterministic assertion is needed, add it to #12454 so the recovery-format compatibility, unit coverage, and real CDC regression remain one reviewable change. #12454 currently has a successful Build but is still under review, so this is not fixed in `dev` yet. -- 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]
