SEZ9 commented on issue #13052: URL: https://github.com/apache/seatunnel/issues/13052#issuecomment-6538609991
Hi @KaustAbhinand, thanks for offering — there is no active assignment on this issue as far as I can see, so please go ahead and pick it up. To recap the scope: deferred checkpoint error handlers in `CheckpointCoordinator` are not tied to a coordinator generation, so a stale error raised before `restoreCoordinator()` runs can still fail the restored coordinator and overwrite its `errorByPhysicalVertex`. A few concrete asks for the PR: - Tie each deferred error handler to the coordinator generation it was created in (or an equivalent token), and drop/ignore errors whose generation no longer matches the current one after `restoreCoordinator()`. - Make sure `errorByPhysicalVertex` is only written by errors belonging to the current generation. - Add a unit test that reproduces the race (error scheduled before restore, delivered after) and verifies the restored coordinator is not failed and its `errorByPhysicalVertex` is untouched. - Link this issue from the PR description. Let me know if you have questions once you start digging in. <!-- streview-comment:2162 --> -- 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]
