Thanks for this. I have been banging up against errors caused by the
ReduceTasks failing because of waits for about a week now. This fixes it.
Dennis
Doug Cutting (JIRA) wrote:
> [ http://issues.apache.org/jira/browse/HADOOP-262?page=all ]
>
> Doug Cutting resolved HADOOP-262:
> ---------------------------------
>
> Fix Version: 0.3.1
> (was: 0.3.0)
> Resolution: Fixed
>
> I just committed this. Thanks, Mahadev.
>
>
>> the reduce tasks do not report progress if they the map output locations is empty.
>> ----------------------------------------------------------------------------------
>>
>> Key: HADOOP-262
>> URL: http://issues.apache.org/jira/browse/HADOOP-262
>> Project: Hadoop
>> Type: Bug
>>
>
>
>> Components: mapred
>> Versions: 0.3.0
>> Reporter: Mahadev konar
>> Assignee: Mahadev konar
>> Fix For: 0.3.1
>> Attachments: reduce_sleep.patch
>>
>> The ReduceTaskRunner should report progress even if the number of ouput locations
to copy is empty. In case, the last few maps are running on a tasktracker that goes down,
all the reduce tasks waiting for these mapoutputs would fail.
>>
>
>
|