SEZ9 commented on issue #12061: URL: https://github.com/apache/seatunnel/issues/12061#issuecomment-5564334212
Thanks @nzw921rx for the expanded methodology and for clarifying expectations; that lines up with what I asked for above. To keep this thread coordinated: @modelpath-dev, the sequence is evidence first, PR later. Concretely, the first update we'd like to see here is: 1. One affected parameter combination from the existing IMap job-storage benchmark (taskGroupStateTransition or runningMetricsReport at 1000 tasks), reproduced with the same runner class, JDK, benchmark arguments, and storage configuration as the original runs. 2. The JMH JSON output for that reproduction, plus whichever of CPU, wall, lock, GC, or JFR profiles you were able to capture around the anomalous iterations. 3. A short note separating fixture/environment variance from behavior in the production storage and metrics-report paths, including hypotheses you have already ruled out and any open questions. A consistent local environment is fine for this stage, and existing experiments you have already run are useful inputs, so please post what you have rather than restarting. Once the evidence points at a specific change point, we can decide together whether the right outcome is a benchmark correction, a production change, or additional diagnostics. If it ends up being a code change, it must preserve task-group transition and metrics-report semantics (including persistence and failure behavior), ship with targeted tests, and report comparable before/after latency and variance. Please link any focused PR back here when it is ready. <!-- streview-comment:857 --> -- 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]
