SEPURI-SAI-KRISHNA commented on PR #13133: URL: https://github.com/apache/seatunnel/pull/13133#issuecomment-6647161375
Thanks for the fix, @nzw921rx. Since #13130 is marked help wanted, I checked this end to end on a real server, in case an independent result helps. I used a local engine build whose `CoordinatorService` and `RealtimeMetricsService` match `dev`, and applied only this PR's `getJobInfo` change as a patch. On JDK 17 I started a streaming FakeSource to Console job with observability disabled and `bucket_ms = 1000`, stopped it with a savepoint through `/stop-job`, then restored the same job ID with observability enabled and `bucket_ms = 2000`. I read the REST endpoints 25 and 45 seconds after the restore. | After restore | Without the change | With the change | |---|---|---| | `/metrics/realtime/jobs` | `enabled:false`, `bucketMs:1000` | `enabled:true`, `bucketMs:2000` | | `/edges` | `enabled:false`, `edges:[]` | `bucketMs:2000`, `edges` has data | | `/vertices` | `enabled:false`, `vertices:[]` | `bucketMs:2000`, `vertices` has data | So the bug reproduces without the change and is gone with it. This was one run of each case, using my own copy of your IT's flow rather than the IT itself, and I did not test the enabled to disabled direction. One optional thought, not needed for merge: `edges` had data here with the same observability settings as your IT, so a non-empty `edges` check could be added if you want the test to cover that part of the reported symptom. -- 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]
