You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Zili Chen (Jira)" <ji...@apache.org> on 2020/03/20 01:32:00 UTC

[jira] [Comment Edited] (FLINK-16637) Flink per job mode terminates before serving job cancellation result

    [ https://issues.apache.org/jira/browse/FLINK-16637?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17063025#comment-17063025 ] 

Zili Chen edited comment on FLINK-16637 at 3/20/20, 1:31 AM:
-------------------------------------------------------------

[~gjy] after "cancel" command received by the {{JobMaster}}, it causes

1. The job asynchronously cancelled, and later causes cluster shutdown.
2. The response asynchronously sent

I don't find an explicit synchronization between 1 and 2, so is the statement.

But generally 1 takes some time before later it causes the cluster shutdown, and thus 2 can nearly always happens before the cluster shutdown.

Possibly we flush outstanding response in rest server before it gets closed.


was (Author: tison):
[~gjy] after "cancel" command received by the {{JobMaster}}, it causes

1. The job asynchronously cancelled
2. The response asynchronously sent

I don't find an explicit synchronization between 1 and 2, so is the statement.

But generally 1 takes some time before later it causes the cluster shutdown, and thus 2 can nearly always happens before the cluster shutdown.

> Flink per job mode terminates before serving job cancellation result
> --------------------------------------------------------------------
>
>                 Key: FLINK-16637
>                 URL: https://issues.apache.org/jira/browse/FLINK-16637
>             Project: Flink
>          Issue Type: Bug
>          Components: Deployment / YARN
>    Affects Versions: 1.10.0
>            Reporter: Zili Chen
>            Priority: Major
>             Fix For: 1.10.1, 1.11.0
>
>         Attachments: yarn.log
>
>
> The {{MiniDispatcher}} no longer waits until the REST handler has served the cancellation result before shutting down. This behaviour seems to be introduced with FLINK-15116.
> See also [https://lists.apache.org/x/thread.html/rcadbd6ceede422bac8d4483fd0cdae58659fbff78533a399eb136743@%3Cdev.flink.apache.org%3E]



--
This message was sent by Atlassian Jira
(v8.3.4#803005)