[ http://issues.apache.org/jira/browse/HADOOP-597?page=all ]
Owen O'Malley updated HADOOP-597:
---------------------------------
Attachment: map-output-protect.patch
Here is a patch that fixes the problem. I use a flag to mark that the fault was
on the input and therefore the map output should be declared lost. I'm not
really that happy with all of the catches that rethrow, but I didn't see a more
straightforward way of getting the right effect.
> transmission errors to the reduce will cause map output to be considered lost
> -----------------------------------------------------------------------------
>
> Key: HADOOP-597
> URL: http://issues.apache.org/jira/browse/HADOOP-597
> Project: Hadoop
> Issue Type: Bug
> Affects Versions: 0.7.0
> Reporter: Owen O'Malley
> Assigned To: Owen O'Malley
> Fix For: 0.7.1
>
> Attachments: map-output-protect.patch
>
>
> When the mapOutput servlet gets an IOException from reading the map output
> file, it correctly declares the map output lost. However, an IOException
> writing to the socket should NOT cause the map output to be declared lost.
> Unfortunately, the current implementation always declares the map output
> lost. When the map output is lost, the map is re-run on a different computer.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://issues.apache.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
