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 "Jian He (JIRA)" <ji...@apache.org> on 2016/03/01 16:42:18 UTC

[jira] [Commented] (YARN-4750) App metrics may not be correct when an app is recovered

    [ https://issues.apache.org/jira/browse/YARN-4750?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15173917#comment-15173917 ] 

Jian He commented on YARN-4750:
-------------------------------

this was intentional, as the thought was persisting the metrics periodically while the app is running will cause too much load on state-store.

> App metrics may not be correct when an app is recovered
> -------------------------------------------------------
>
>                 Key: YARN-4750
>                 URL: https://issues.apache.org/jira/browse/YARN-4750
>             Project: Hadoop YARN
>          Issue Type: Bug
>            Reporter: Lavkesh Lahngir
>            Assignee: Lavkesh Lahngir
>
> App metrics(rather app attempt metrics) like Vcore-seconds and MB-seconds are saved in the state store when there is an attempt state transition. Values for running attempts will be in memory and will not be saved when there is an RM restart/failover. For recovered app metrics value will be reset. In that case, these values will be incomplete. 
> Was this intentional or have we not found a correct way to fix it?



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)