yuqi1129 opened a new issue, #13180: URL: https://github.com/apache/gravitino/issues/13180
### Describe the subtask `JobManager.runJob` submits to the job executor **before** `entityStore.put`. If the registration fails or the server dies after submission, the executor runs a job Gravitino does not know about; a retry submits it twice. Job/template OCC and version-checked status transitions exist (#12669, #12992), but they cannot cover this window because the executor is not part of the store transaction. Proposed change: - Insert the job row in a `SUBMITTING` state first, submit, then CAS to `QUEUED`/`RUNNING` with the executor's id. - The status poller reconciles `SUBMITTING` rows against the executor and treats the executor as the source of truth when they disagree. - Extend `TestJobManagerMultiNode`: executor accepts, registration fails / server dies. Design: implementation design Part 4 (S3). Gate G4. ### Parent issue https://github.com/apache/gravitino/issues/10238 -- 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]
