JayModha100 commented on PR #13272: URL: https://github.com/apache/seatunnel/pull/13272#issuecomment-6895431445
@DanielLeens thanks for the review. Pushed `bbac55a6f3` with updates: 1. **Engine IT:** added `SavepointRestoreIT#testQueueMetricsStableAcrossRepeatedSavepointRestores` (parallelism 5, same job ID, 3 savepoint/restore cycles). It asserts `subtaskCount` and `queueCapacity` stay at the baseline on every fresh bucket after each restore, plus source progress and no duplicate offsets. With the cleanup call disabled it fails in cycle 1 (`subtaskCount drifted ... expected: <5> but was: <7>`); with the fix it passes. It covers the default blocking queue; Disruptor remains covered by the unit tests, and the MySQL CDC idle-reader case is not covered. 2. **Comments and docs:** added comments at the cleanup call site and the name matcher, and an EN/ZH note in `realtime-observability.md`. 3. **`RealtimeMetricsServiceTest` clock:** not changed in this PR. The tests build buckets from `System.currentTimeMillis()` and the service samples the clock again when filtering its 60 s window, so they are not strictly deterministic. The two samples are taken close together, so a failure would need a pause of about a minute or a clock jump. Making it deterministic needs a clock hook in `RealtimeMetricsService` (for example an overload taking an explicit "now"), which I'd rather not add to this narrow fix. Happy to do it in this PR or a follow-up, whichever you prefer. CI is running on the new head; I'll post the results for the failed lanes once it finishes. -- 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]
