Jackeyzhe commented on PR #29239: URL: https://github.com/apache/flink/pull/29239#issuecomment-5817186235
I found one mixed-state case that the new `hasKnownActivity()` guard does not cover. If a subtask is already idle, register a silent split (UNKNOWN), then register another split and call `markIdle()` on it. The latter makes `hasKnownActivity()` true, but the UNKNOWN split keeps the combined `idle` false, so `updateCombinedWatermark()` forwards `markActive()` even though no split has produced a record or watermark. A temporary `ProgressiveTimestampsAndWatermarksSubtaskIdlenessTest` on 627b6ce expected the downstream idle updates to remain `[true]`, but got `[true, false]`. Could we distinguish known ACTIVE from the mixed UNKNOWN+IDLE case for the downstream activity signal, while retaining UNKNOWN's watermark-holding behavior? A regression for the mixed state would help. -- 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]
