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 "Junping Du (JIRA)" <ji...@apache.org> on 2014/11/17 16:44:34 UTC

[jira] [Commented] (MAPREDUCE-6164) "mapreduce.reduce.shuffle.fetch.retry.timeout-ms" should be set to 3 minutes instead of 30 seconds by default to be consistent with other retry timeout

    [ https://issues.apache.org/jira/browse/MAPREDUCE-6164?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14214751#comment-14214751 ] 

Junping Du commented on MAPREDUCE-6164:
---------------------------------------

Deliver a quick patch to fix it.

> "mapreduce.reduce.shuffle.fetch.retry.timeout-ms" should be set to 3 minutes instead of 30 seconds by default to be consistent with other retry timeout 
> --------------------------------------------------------------------------------------------------------------------------------------------------------
>
>                 Key: MAPREDUCE-6164
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-6164
>             Project: Hadoop Map/Reduce
>          Issue Type: Bug
>            Reporter: Junping Du
>            Assignee: Junping Du
>         Attachments: MAPREDUCE-6164.patch
>
>
> In MAPREDUCE-5891, we are adding retry logic to MAPREDUCE shuffle stage for fetcher can be survival during NM downtime (with shuffle service down as well). In many places, we are setting the default timeout to be 3 minutes (connection timeout, etc.) to tolerant possible more time for NM down, but we are making "mapreduce.reduce.shuffle.fetch.retry.timeout-ms" to be 30 seconds which is not consistent here. We should change this to 180 seconds. 



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