[jira] Updated: (HADOOP-1219) Spurious progress messages should be discarded after a task is done

["Devaraj Das (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/HADOOP-1219?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Devaraj Das updated HADOOP-1219:
--------------------------------

    Status: Patch Available  (was: Open)

The patch does a check at the TaskTracker to make sure the progress report is ignored if the task had earlier reported as 'done'.

> Spurious progress messages should be discarded after a task is done
> -------------------------------------------------------------------
>
>                 Key: HADOOP-1219
>                 URL: https://issues.apache.org/jira/browse/HADOOP-1219
>             Project: Hadoop
>          Issue Type: Bug
>          Components: mapred
>            Reporter: Devaraj Das
>         Assigned To: Devaraj Das
>             Fix For: 0.13.0
>
>         Attachments: 1219.patch
>
>
> Noticed that the Map task's progress message sometimes goes through even after the fix for HADOOP-1191. In reality, this *could* happen since we don't check for thread.isInterrupted() at each step in the path of the client RPC call. So the situation described in HADOOP-1191 remains. The effect of this is that the TaskTracker finally kills the task for not sending progress report for the configured number of seconds (if the process happens to be alive then), or, assumes the process died due to some unknown reason when the process exits before the configured timeout.

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