SEPURI-SAI-KRISHNA commented on issue #13324: URL: https://github.com/apache/seatunnel/issues/13324#issuecomment-7072525379
I would like to work on this. From the code on dev `7029c0969a`: `isNoErrorCompleted()` can only be true for a final checkpoint this coordinator completed itself, since a restored checkpoint is marked `restored`. The coordinator has then already put the same status into `checkpointCoordinatorFuture`: FINISHED, SUSPEND, or FAILED when `notifyCompleted` failed (the case #6193 added the IMap check for). So the IMap read only adds this failure mode. Plan: 1. `isNoErrorCompleted()` takes the terminal status from the coordinator's own future, not the IMap. 2. A failed terminal write is logged at ERROR with its cause. 3. Unit test from your probe (`set` throws, IMap stays RUNNING): true after FINISHED, false after FAILED. Out of scope: after a master failover the new coordinator still reads the IMap, so a lost FINISHED still re-runs the final checkpoint. Durable terminal writes change failover behavior; I would leave that to a separate issue. Draft #12043 also reworks `updateStatus`; I will rebase on whichever lands first. Does this direction work? -- 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]
