[
https://issues.apache.org/jira/browse/FLINK-26063?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17504937#comment-17504937
]
Roman Khachatryan commented on FLINK-26063:
-------------------------------------------
So far I've only been able to reproduce the issue on a cluster.
The same
[program|https://github.com/rkhachatryan/flink/blob/c24c94261c23cbebd664ecf02c61e937bb7d1e36/flink-end-to-end-tests/flink-changelog-test/src/main/java/org/apache/flink/tests/ChangelogTestProgram.java#L53]
on minicluster doesn't fail.
> [Changelog] Incorrect key group logged for PQ.poll and remove
> -------------------------------------------------------------
>
> Key: FLINK-26063
> URL: https://issues.apache.org/jira/browse/FLINK-26063
> Project: Flink
> Issue Type: Bug
> Components: Runtime / State Backends
> Affects Versions: 1.15.0
> Reporter: Roman Khachatryan
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.15.0
>
>
> Key group is logged so that state changes can be re-distributed or shuffled.
> It is currently obtained from keyContext during poll() and remove()
> operations.
> However, keyContext is not updated when dequeing processing time timers.
> The impact is relatively small for remove(): in the worst case, the operation
> will be ignored.
> poll() should probably be replaced with remove() anyways - see FLINK-26062.
> One way to solve this problem is to extract key group from the polled element
> - if it is a timer.
> cc: [~masteryhx], [~ym], [~yunta]
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
