DanielLeens commented on issue #12565: URL: https://github.com/apache/seatunnel/issues/12565#issuecomment-6415643253
Thanks @WXPNG for following up. I need to correct the implementation status in the earlier replies: #12567 was closed by its author on 2026-10-15 **without merging**. Current `dev` (`4f5e3293740`) does not contain that per-table restore-offset fix. I am sorry the older “fix in progress in #12567” wording could now suggest an active PR. I also checked open #11206: it addresses wildcard discovery of tables created while a job is running and the corresponding sink registration. It does not claim to fix this savepoint → add-table → restore case, where already-streaming tables can lose changes (and where an incorrect restart position can also replay them under `exactly_once`). This issue therefore stays open; I added `cdc`, `connectors-v2`, and `help wanted` after checking the current source, recent commits, and PRs. A contributor can revive/rebase #12567 or open a focused replacement, but please carry forward the unresolved restart-position review from that PR: prove the old table's own checkpoint offset is retained through split reconstruction, assignment, and the fetcher, and cover both missing and duplicate records with `incremental.parallelism > 1`. Run GitHub CI on the final head and link the new PR here. If you can also recheck your original old-table/new-table row counts against it, that will give us an independent end-to-end acceptance signal. -- 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]
