davidzollo commented on PR #11244: URL: https://github.com/apache/seatunnel/pull/11244#issuecomment-4894450528
Thanks for the improvement. Excluding `engine_runningJobMetrics` from IMAP persistence looks like a reasonable direction to avoid WAL growth for high-frequency observability-only snapshots. One behavior worth making explicit: after an engine restart, recovered tasks will call `TaskExecutionContext#getOrCreateMetricsContext(...)`. If the previous `engine_runningJobMetrics` snapshot is no longer restored, this path falls back to a new `SeaTunnelMetricsContext`, so the running-job metrics shown after restart will be rebuilt from subsequent reports rather than continuing from the pre-restart snapshot. That seems acceptable if this map is intentionally treated as auxiliary runtime state, but it may be worth documenting this reset/rebuild behavior more directly. One implementation detail I would suggest tightening: `FileMapStore` now initializes `mapStorage` to `NoOpMapStorage.INSTANCE` by default. To keep the no-op behavior strictly scoped to `RUNNING_JOB_METRICS`, would it be safer to assign the no-op storage only inside the `RUNNING_JOB_METRICS` branch and keep other maps fail-fast if storage initialization did not complete? That avoids accidentally hiding initialization failures for persistent engine maps. -- 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]
