hadoop-mapreduce-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ravi Gummadi (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (MAPREDUCE-2761) New TaskController code doesn't run on Windows
Date Tue, 02 Aug 2011 06:16:27 GMT

    [ https://issues.apache.org/jira/browse/MAPREDUCE-2761?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13076060#comment-13076060

Ravi Gummadi commented on MAPREDUCE-2761:

Linux task controller seems to be giving NPE on linux while trying to kill/signal task process.
And this is causing task processes not getting killed and there are about 100 tasks running
on each TT node. And then tasks/jobs start failing. May need a separate JIRA for this.

Vinay, please provide the exception seen here.

> New TaskController code doesn't run on Windows
> ----------------------------------------------
>                 Key: MAPREDUCE-2761
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-2761
>             Project: Hadoop Map/Reduce
>          Issue Type: Bug
>          Components: task-controller, tasktracker
>    Affects Versions:, 0.23.0
>            Reporter: Todd Lipcon
> After MAPREDUCE-2178, TaskController assumes that pids are always available. The shell
executor object that's used to launch a JVM isn't retained, but rather the pid is set when
the task heartbeats. On Windows, there are no pids, and since the ShellCommandExecutor object
is no longer around, we can't call process.destroy(). So, the TaskController doesn't work
on Cygwin anymore.

This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira


View raw message