Rangsh commented on issue #12063: URL: https://github.com/apache/seatunnel/issues/12063#issuecomment-5675138362
@SEZ9 thanks — aligned on your latest note. ### Status - Fork runs [`34354309911`](https://github.com/Rangsh/seatunnel/actions/runs/34354309911) / [`34354315647`](https://github.com/Rangsh/seatunnel/actions/runs/34354315647) remain **superseded** (baseline-only on `dev` @ `75fd4ed`; not a PR before/after). - Measurement head for #12173 stays **`7948e2640`** (spotless-only on top of `c553a8c04`); I am not pushing further unless a re-review asks for it. - I still cannot dispatch **Benchmarks** on `apache/seatunnel` (403: needs admin/dispatch permission). ### Ask: same-worker Benchmarks dispatch @nzw921rx (or another committer with workflow dispatch permission) — could you please trigger **Benchmarks** on `apache/seatunnel` with: | Input | Value | | --- | --- | | `seatunnel_ref` | `dev` | | `benchmarks` | `IMapJobStorageBenchmark` | | `custom_benchmarks` | `IMapJobStorageBenchmark.(runningJobGrowth\|completedJobHistoryGrowth)$` | | `pr_number` | `12173` | Expected same-worker order: **baseline → PR → PR → baseline** (head `7948e2640`). Once that run finishes, I will post the **Score / Error / CV** table here for both methods at `initialStoredJobCount=0` and `1000` from that single host, replacing the superseded fork table. ### Plan if `runningJobGrowth` @ `initialStoredJobCount=1000` CV stays high If the same-worker CV for `runningJobGrowth` @ `1000` remains in a high band (e.g. ~15%+): 1. **First**: request one follow-up same-worker Benchmarks dispatch with more JMH sampling (higher `-f` / `-i`) on the same selector, still at head `7948e2640`, to see whether CV drops with more iterations. 2. **If it stays high**: treat that cell as **inherently noisier** at pressure=1000 (growth + resident checks under larger IMap/WAL), and **not** over-read absolute Score deltas there; rely on (a) the fixture causal evidence already posted (Diagnostics wall/`FileMapStore.loadAll`), and (b) the other three cells (`runningJobGrowth@0`, `completedJobHistoryGrowth@0/1000`) for variance improvement. I will state that reading explicitly next to the table so remaining variance is not mistaken for an unresolved production regression. Happy to post the table as soon as the dispatch completes — please drop the run link here when it starts. -- 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]
