[jira] [Commented] (FLINK-16562) Handle JobManager termination future in place

["Chesnay Schepler (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FLINK-16562?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17068554#comment-17068554 ] 

Chesnay Schepler commented on FLINK-16562:
------------------------------------------

If you don't have the jobManagerTerminationFutures field, how would you wait for all jobmanagers to have shut down in Dispatcher#onStop?

> Handle JobManager termination future in place
> ---------------------------------------------
>
>                 Key: FLINK-16562
>                 URL: https://issues.apache.org/jira/browse/FLINK-16562
>             Project: Flink
>          Issue Type: Improvement
>          Components: Runtime / Coordination
>            Reporter: Zili Chen
>            Assignee: Zili Chen
>            Priority: Major
>             Fix For: 1.11.0
>
>
> After FLINK-11843 {{Dispatcher}} becomes a {{PermanentlyFencedRpcEndpoint}} and will be created as different instance in difference leader epoch. Thus, we don't have {{jobManagerTerminationFutures}} crosses multiple leader epoch that should be handled. Given the truth, we can remove {{jobManagerTerminationFutures}} field in {{Dispatcher}} and handle those futures in place, which will simplify the code and helps on further refactoring.
> CC [~trohrmann]
> I will create a branch later this week.



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