[jira] Commented: (HADOOP-3976) JobTracker.close() gets stuck occasionally

["Doug Cutting (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HADOOP-3976?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12625681#action_12625681 ] 

Doug Cutting commented on HADOOP-3976:
--------------------------------------

I think Thread#interrupt() would work if no code in the thread caught and ignored InterruptedException, but, unfortunately, there's a lot of code in Hadoop that does catch and ignore this exception.  Given that, the best practice is to both use a flag and an interrupt, belt and suspenders.  +1 for the patch.

> JobTracker.close() gets stuck occasionally
> ------------------------------------------
>
>                 Key: HADOOP-3976
>                 URL: https://issues.apache.org/jira/browse/HADOOP-3976
>             Project: Hadoop Core
>          Issue Type: Bug
>          Components: mapred
>            Reporter: Amar Kamat
>         Attachments: HADOOP-3976-v1.patch
>
>
> JobTracker.close() shuts down all the worker threads by interrupting them and then doing a join. This will work if the thread is in {{WAITING}} state. Most of the time the worker threads are in the {{RUNNING}} state and the JT waits forever on the join(). 

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