[
https://issues.apache.org/jira/browse/FLINK-38454?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Timo Walther updated FLINK-38454:
---------------------------------
Affects Version/s: 2.1.0
> Inconsistent idle behavior between StatusWatermarkValve and
> CombinedWatermarkStatus
> -----------------------------------------------------------------------------------
>
> Key: FLINK-38454
> URL: https://issues.apache.org/jira/browse/FLINK-38454
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Task
> Affects Versions: 2.1.0
> Reporter: Timo Walther
> Assignee: Timo Walther
> Priority: Major
>
> Both StatusWatermarkValve and CombinedWatermarkStatus derive watermarks and
> deal with idle inputs (from input channel or input split respectively).
> However, cases where all inputs are marked as idle are handled differently.
> In both cases, the maximum watermark should be derived from all idle inputs.
> From comments in StatusWatermarkValve (which implements this already):
> {code}
> // now that all subpartitions are idle and no subpartitions will continue to
> advance
> // its watermark, we should "flush" all watermarks across all subpartitions;
> // effectively, this
> // means emitting the max watermark across all subpartitions as the new
> watermark.
> {code}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
