[jira] [Commented] (MAPREDUCE-4121) dump the threads stack trace to stdout before killing a Task in timeout

["Julien Le Dem (Commented) (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/MAPREDUCE-4121?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13249628#comment-13249628 ] 

Julien Le Dem commented on MAPREDUCE-4121:
------------------------------------------

I see several ways of doing this.
one is to add a jstack() method to org.apache.hadoop.util.Shell to run jstack (or kill -3) on the command line.
The other one is to use sun.tools.jstack.JStack.main() directly (if the class is present).
These all require the PID of the process to send the signal to.

org.apache.hadoop.mapred.TaskTracker would call jstack() before killing the Task

                
> dump the threads stack trace to stdout before killing a Task in timeout
> -----------------------------------------------------------------------
>
>                 Key: MAPREDUCE-4121
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-4121
>             Project: Hadoop Map/Reduce
>          Issue Type: Improvement
>            Reporter: Julien Le Dem
>
> Typically when a job fails because of tasks timing out we investigate the issue by running the job again and triggering a dump of the thread stack traces of one of the tasks with jstack/"kill -3" before it times out.
> It would be convenient if the Task tracker could do the same right before killing tasks in time out. This usually points at the offending code.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira