joeyutong commented on issue #1034: URL: https://github.com/apache/flink-agents/issues/1034#issuecomment-5422830414
Regarding boundary persistence: I suggest storing it in a separate compacted Kafka control topic, outside both Flink checkpoint state and the action-state data topic. This prevents the boundary from rolling back with an older checkpoint or being removed by `deleteRecords` on the data topic. The acknowledged `COMMITTED` boundary should be the logical point of no return: recovery must reject any older marker immediately, even if physical deletion has not finished. `deleteRecords` can then retry the same per-partition vector idempotently, with the operation moving to `APPLIED` after the resulting low watermarks have been verified. This gives crash and partial-partition failures an unambiguous recovery 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]
