You are viewing a plain text version of this content. The canonical link for it is here.
Posted to common-dev@hadoop.apache.org by "Owen O'Malley (JIRA)" <ji...@apache.org> on 2007/10/03 09:08:54 UTC

[jira] Commented: (HADOOP-1983) jobs using pipes interface with tasks not using java output format have a good chance of not updating progress and timing out

    [ https://issues.apache.org/jira/browse/HADOOP-1983?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12532015 ] 

Owen O'Malley commented on HADOOP-1983:
---------------------------------------

Would adding a flush on the command socket when sending the progress address the problem? It is already limited to 1/second, so it shouldn't cause excessive load.

> jobs using pipes interface with tasks not using java output format have a good chance of not updating progress and timing out
> -----------------------------------------------------------------------------------------------------------------------------
>
>                 Key: HADOOP-1983
>                 URL: https://issues.apache.org/jira/browse/HADOOP-1983
>             Project: Hadoop
>          Issue Type: Bug
>          Components: pipes
>    Affects Versions: 0.15.0
>            Reporter: Christian Kunz
>
> When using C++-pipes interface, mappers/reducers not emitting any key-values pairs, but running longer than 'mapred.task.timeout' might timeout even when they send periodical status updates upstream.
> The cause of the problem is that all upstream messages are buffered. Progress and status updates should be flushed immediately.

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