You are viewing a plain text version of this content. The canonical link for it is here.
Posted to yarn-issues@hadoop.apache.org by "Varun Saxena (JIRA)" <ji...@apache.org> on 2014/08/20 16:00:45 UTC

[jira] [Updated] (YARN-2432) RMStateStore should process the pending events before clo

     [ https://issues.apache.org/jira/browse/YARN-2432?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Varun Saxena updated YARN-2432:
-------------------------------

    Summary: RMStateStore should process the pending events before clo  (was: RMStateStore should process the pending events before close)

> RMStateStore should process the pending events before clo
> ---------------------------------------------------------
>
>                 Key: YARN-2432
>                 URL: https://issues.apache.org/jira/browse/YARN-2432
>             Project: Hadoop YARN
>          Issue Type: Bug
>          Components: resourcemanager
>            Reporter: Varun Saxena
>            Assignee: Varun Saxena
>
> Refer to discussion on YARN-2136 (https://issues.apache.org/jira/browse/YARN-2136?focusedCommentId=14097266&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-14097266). 
> As pointed out by [~jianhe], we should process the dispatcher event queue before closing the state store by flipping over the following statements in code.
> {code:title=RMStateStore.java|borderStyle=solid}
>  protected void serviceStop() throws Exception {
>     closeInternal();
>     dispatcher.stop();
>   }
> {code}
> Currently, if the state store is being stopped on events such as switching to standby, it will first close the state store(in case of ZKRMStateStore, close connection with ZK) and then process the pending events. Instead, we should first process the pending events and then call close.



--
This message was sent by Atlassian JIRA
(v6.2#6252)