[ 
https://issues.apache.org/jira/browse/FLINK-4807?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Till Rohrmann resolved FLINK-4807.
----------------------------------
       Resolution: Duplicate
    Fix Version/s: 1.5.0

> ResourceManager clean up JobManager's registration
> --------------------------------------------------
>
>                 Key: FLINK-4807
>                 URL: https://issues.apache.org/jira/browse/FLINK-4807
>             Project: Flink
>          Issue Type: Sub-task
>          Components: Cluster Management
>            Reporter: Kurt Young
>            Priority: Major
>             Fix For: 1.5.0
>
>
> When RM received a JM's registration, it will record it either with some 
> leaderid or leadership listener. We should make sure the finished / failed JM 
> can properly unregister itself with RM.
> We can make it happen by doing these two things:
> 1. If JM finds out job reaches a terminate state(either success or fail), it 
> should send an unregistration request to RM.
> 2. If (1) does not happen for various reasons, RM can rely on the heartbeat 
> manager to find out timeout JM and clear it up.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to