[ 
https://issues.apache.org/jira/browse/YARN-11566?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17765110#comment-17765110
 ] 

ASF GitHub Bot commented on YARN-11566:
---------------------------------------

zhengchenyu commented on PR #6068:
URL: https://github.com/apache/hadoop/pull/6068#issuecomment-1719245764

   There are three ways to solve this problem:
   
   1. Call forceKillApplication on all sub cluster.
   2. Store secondary sub cluster when submit unmanaged application, then 
forceKillApplication on these secondary sub cluster.
   3. forceKillApplication to secondary sub cluster when application is removed 
in nodemanager. 
   
   Solution 1 is the easiest way. But we may call 
   forceKillApplication to unused sub cluster. And If we kill application to 
home subcluster but not router, the app in secondary cluster will not be 
removed.
    
   Solution 2: The problem is that NM will store the meta info of app. Router 
does not know when NM store the info. So we must disable the cache.
    
   Here I will choose solution 3. UAM is managed by NM, so this solution is 
most reasonable.
   
   This PR is about solution 3.




> Yarn app kill command can not kill the application in secondary sub cluster.
> ----------------------------------------------------------------------------
>
>                 Key: YARN-11566
>                 URL: https://issues.apache.org/jira/browse/YARN-11566
>             Project: Hadoop YARN
>          Issue Type: Improvement
>            Reporter: zhengchenyu
>            Assignee: zhengchenyu
>            Priority: Major
>              Labels: pull-request-available
>
> When AMRMProxy is enable, the application may allocate container among multi 
> sub cluster. The application in secondary sub cluster will be labeled as 
> unmananged application. When we run 'yarn app -kill \{appid}', the unmananged 
> application will not be killed in secondary sub cluster.
> The unmanaged application will be removed util app attempt is expired after 
> 15 minute.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to