Vivek1106-04 commented on PR #12165: URL: https://github.com/apache/seatunnel/pull/12165#issuecomment-6266752400
@SEZ9 @DanielLeens @nzw921rx The benchmark harness fix is now open as #12896 (benchmark code and tests only, same commit `1914eabf6` that the sweep in https://github.com/apache/seatunnel/pull/12418#issuecomment-6135494803 was run against). The open question from #12418 has had no answer yet, so I opened the PR with my current choices and listed the question in its description. Restating it here so it does not stall: 1. **JIT thresholds:** the fix forks with `-XX:CompileThresholdScaling=0.05` so the trigger path compiles within warmup. The alternative is JVM defaults with about 150 s of warmup per trial (not yet run as a full sweep, and it leaves the one-pipeline point at a lower compilation tier). 2. **Cost:** the trigger-count warmup adds about 19 minutes per JDK to the `CheckpointSchedulingBenchmark` sweep in `benchmarks.yml`. Dropping the one-pipeline point removes the noisiest point and about half of that time. @nzw921rx, either choice is a small change in #12896; please say which you prefer there. Plan for this PR, unchanged: - Head stays at `c773cc091` until the harness fix lands. - After #12896 merges I will rerun this PR against `dev` with the fixed harness and post the results here as well as on #12418. - If the rerun needs any commit on top of `c773cc091`, or a sync with `dev` (currently behind with no conflicts), I will say so explicitly in that comment so the CI attribution is redone on the new head. -- 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]
