hadoop-yarn-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Daryn Sharp (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (YARN-338) RM renews tokens even when maxDate will soon be exceeded
Date Tue, 07 May 2013 17:25:16 GMT

    [ https://issues.apache.org/jira/browse/YARN-338?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13651085#comment-13651085

Daryn Sharp commented on YARN-338:

I think I may have filed a jira about this a long time back, but I couldn't find it in a quick
search.  I had a patch that simply stopped renewing if the new expiration equaled the old
expiration.  Internal discussion also was how the RM should kill the job when the token actually
expires – to avoid tying up cluster resources with needless task and app retries that will
inevitably die a slow death.
> RM renews tokens even when maxDate will soon be exceeded
> --------------------------------------------------------
>                 Key: YARN-338
>                 URL: https://issues.apache.org/jira/browse/YARN-338
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: resourcemanager
>    Affects Versions: 0.23.3, 3.0.0, 2.0.0-alpha
>            Reporter: Daryn Sharp
> The RM will renew tokens 90% of the way to the next expiration.  When the max lifetime
is approaching, the next expiration is always the max lifetime.  The RM starts to unnecessarily
renew more and more frequent as that hard limit approaches.  The RM should stop renewing when
the last expiration matches the new expiration.

This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

View raw message