DanielLeens commented on issue #13137: URL: https://github.com/apache/seatunnel/issues/13137#issuecomment-6657729803
Thanks for the careful scope statement and the sanitized endpoint evidence. I checked current `dev` (`3cb9efc2a15223372bc1f87a92d027845a8ca800`). `TaskExecutionService.provideDynamicMetrics()` still builds its collection from both `finishedExecutionContexts` and `executionContexts`; the active map overwrites an entry with the same `TaskGroupLocation`. `notifyCleanTaskGroupContext()` removes a finished context only when that cleanup path is invoked. This makes aggregation scope a valid investigation boundary, but it does not by itself prove that a previous execution is being counted twice: a same-location collision is overwritten. Before changing the engine, please provide a minimized reproduction that includes: 1. The complete sanitized job configuration and the exact submit, savepoint-stop, and restore commands, including the reused job ID and the `dev` commit. 2. The task-mapping and realtime-edges responses immediately before the stop and after the restored attempt becomes RUNNING, with the task-group IDs, queue IDs, and observation timestamps retained. 3. Whether any worker or master failover occurred, plus the relevant worker logs around task-group finish, context cleanup, and restore. With that evidence, the regression test can state the required invariant precisely: whether the endpoint must report only active queue-bearing task groups, or must separately expose retained finished contexts. Please keep the scope to realtime-observability accounting; this report does not yet establish a task-thread leak or data-correctness impact. -- 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]
