SEZ9 commented on issue #12063: URL: https://github.com/apache/seatunnel/issues/12063#issuecomment-5565331307
Suggested first step: run the `Benchmarks Diagnostics` workflow with GC and JFR profiling for both methods (initialStoredJobCount=0 and 1000) and attach the results here. Since both share the same job-lifecycle growth fixture, a common cause (per-invocation allocation, IMap growth, or FileMapStore/WAL fsync behavior) is likely; GC pause distribution should quickly confirm or rule out allocation pressure. Fixture-side checks worth doing before touching production code: confirm the IMap is reset between iterations (so map size does not grow monotonically across forks/iterations), check whether the WAL/FileMapStore writes to a disk that other CI jobs share, and try increasing warmup/measurement iterations or forks to see if the CV drops. If variance disappears with a fixture change, that should be the fix rather than a production change. Please reference this issue in any PR and include a before/after table (Score, Error, CV) for both methods from the same runner, plus a note confirming that running-job and completed-job state durability semantics (WAL/FileMapStore persistence) are unchanged. <!-- streview-comment:873 --> -- 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]
