[
https://issues.apache.org/jira/browse/MAPREDUCE-3675?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13187110#comment-13187110
]
Vinod Kumar Vavilapalli commented on MAPREDUCE-3675:
----------------------------------------------------
No maps jobs are valid use-cases. For jobs with no maps, and sometimes even no
reducers, I know of customers who depend on the side-effects caused by
setup/cleanup tasks and/or job-success state. This is generally useful in
workflow scenarios which run periodically and when they run into cases where
some job in the workflow doesn't have input data, but has to cause either some
side-effect or just finish successfully so that downstream jobs can proceed
without breakage.
Given that, I think this should be closed as invalid.
Because of the same, we are supporting 0maps/0reduces cases in mrv2 also.
> A job must not be submitted when there are no mappers to run
> ------------------------------------------------------------
>
> Key: MAPREDUCE-3675
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-3675
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: client
> Affects Versions: 1.0.0
> Reporter: Harsh J
> Priority: Minor
>
> Right now, one's able to run a job with an empty input directory and a job is
> indeed scheduled.
> The job runs no mappers, but any number of specified reducers are run anyway
> as dummy tasks.
> This should be/could be avoided. I do not see a use for such an allowance,
> and it looks mostly like a logic slip to me with empty arrays involved and
> accepted.
> The fix could be simply in the job submission code, where we can avoid
> submitting if the splits are nil.
--
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
