You are viewing a plain text version of this content. The canonical link for it is here.

Posted to mapreduce-issues@hadoop.apache.org by "Jothi Padmanabhan (JIRA)" <ji...@apache.org> on 2009/07/15 06:51:14 UTC

[jira] Commented: (MAPREDUCE-55) Reduce task should stop shuffle-retrying in case of out-of-memory errors

    [ https://issues.apache.org/jira/browse/MAPREDUCE-55?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12731266#action_12731266 ] 

Jothi Padmanabhan commented on MAPREDUCE-55:
--------------------------------------------

Even the GetMapCompletionEvents thread is catching and ignoring OutOfMemory Errors. When it does this, some event completion updates are lost and the reducer just hangs.

> Reduce task should stop shuffle-retrying in case of out-of-memory errors
> ------------------------------------------------------------------------
>
>                 Key: MAPREDUCE-55
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-55
>             Project: Hadoop Map/Reduce
>          Issue Type: Bug
>            Reporter: Amareshwari Sriramadasu
>
> In ReduceTask, MapOutputCopier threads catch Throwble and retry happens for the shuffle. It should not retry incase of Errors suchas OutOfMemoryError etc.
> May be it should retry only in case of Connect/Read failures and die in all other cases. Thoughts?

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.