hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Nathan Marz (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HADOOP-5330) Zombie tasks remain after jobs finish/fail/get killed
Date Fri, 27 Mar 2009 05:59:51 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-5330?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12689819#action_12689819
] 

Nathan Marz commented on HADOOP-5330:
-------------------------------------

The tasks aren't actually shown in the jobtracker UI. They're processes with "attempt***"
in the name running throughout the cluster. After a few weeks there are about a hundred of
them in our 40 node cluster. The only way to get rid of them is to kill -9 them - a normal
kill does not work.

> Zombie tasks remain after jobs finish/fail/get killed
> -----------------------------------------------------
>
>                 Key: HADOOP-5330
>                 URL: https://issues.apache.org/jira/browse/HADOOP-5330
>             Project: Hadoop Core
>          Issue Type: Bug
>    Affects Versions: 0.19.1
>            Reporter: Nathan Marz
>
> I'm seeing a lot of "task attempts" around our hadoop cluster for jobs that are no longer
around. The attempts seem to be "hung", as they sit there forever. Additionally, they seem
to take up map and reduce slots in the cluster unless MapReduce is restarted. This causes
real jobs to be unable to utilize the whole cluster.

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


Mime
View raw message