[
https://issues.apache.org/jira/browse/KYLIN-5292?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17632661#comment-17632661
]
Liu Zhao commented on KYLIN-5292:
---------------------------------
What I understand is that JobProcessContex is located in different processes,
registered in the Job node, and the kill operation is in the Query node, and
the broadcast event does not make the jobServer get Process to kill in its
JobProcessContext.
Therefore, we can judge whether the job status is from running to discarded in
the received event. If so, we can perform the destruction process in the
listener.
I will test and verify it on Monday.
!screenshot-1.png!
> the task that has been submitted to yarn after discard is still executing
> -------------------------------------------------------------------------
>
> Key: KYLIN-5292
> URL: https://issues.apache.org/jira/browse/KYLIN-5292
> Project: Kylin
> Issue Type: Bug
> Components: Job Engine
> Affects Versions: v4.0.1
> Reporter: Liu Zhao
> Priority: Major
> Attachments: image-2022-11-11-20-50-31-861.png,
> image-2022-11-11-20-50-54-135.png, image-2022-11-11-20-53-02-878.png,
> image-2022-11-11-20-53-21-382.png, image-2022-11-11-20-53-38-208.png,
> image-2022-11-11-20-53-57-941.png, screenshot-1.png
>
>
> When I discard a task that has been submitted to Yarn, I find that the task
> has not been killed, and the local Sparksubmit subprocess is still alive. The
> following are my environment, phenomena and troubleshooting:
> env: Job and query are in different processes（server.model=job and
> server.model=query）
> phenomena:
> !image-2022-11-11-20-50-31-861.png!
> !image-2022-11-11-20-50-54-135.png!
> !image-2022-11-11-20-53-02-878.png!
> !image-2022-11-11-20-53-21-382.png!
> !image-2022-11-11-20-53-38-208.png!
> !image-2022-11-11-20-53-57-941.png!
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
