[ 
https://issues.apache.org/jira/browse/MAPREDUCE-3927?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13288615#comment-13288615
 ] 

Jason Lowe commented on MAPREDUCE-3927:
---------------------------------------

I'm not sure we should change the denominator of the progress calculation when 
map tasks fail.  As I see it, the map percentage is a representation of how 
many map tasks have been processed regardless of successful or failure outcome. 
 50% map progress to me means half of the map tasks have been processed and 
half have yet to finish processing.

For example, let's say we have 4 map tasks and are configured to allow at least 
one to fail.  When 1 out of 4 have completed, we're sitting at 25% map 
completion.  If another map task fails, I'd expect the completion to be 50% as 
we've processed two tasks and have yet to finish the other two.  However the 
proposed change will report 33% completion since it will remove the failed task 
from the total task count instead of adding it to the processed task count.  
Adding it to the processed task count would also be in line with how a job 
reports completed maps, see JobImpl.getCompletedMaps().

                
> Shuffle hang when set map.failures.percent
> ------------------------------------------
>
>                 Key: MAPREDUCE-3927
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-3927
>             Project: Hadoop Map/Reduce
>          Issue Type: Bug
>          Components: mrv2
>    Affects Versions: 0.23.3, 2.0.0-alpha, 2.0.1-alpha, trunk
>            Reporter: MengWang
>            Assignee: Bhallamudi Venkata Siva Kamesh
>            Priority: Critical
>              Labels: patch
>         Attachments: MAPREDUCE-3927-1.patch, MAPREDUCE-3927-2.patch, 
> MAPREDUCE-3927.patch, MAPREDUCE-3927.patch
>
>
> When set mapred.max.map.failures.percent and there does have some failed 
> maps, then shuffle will hang

--
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

        

Reply via email to