[
https://issues.apache.org/jira/browse/MAPREDUCE-727?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12776613#action_12776613
]
Todd Lipcon commented on MAPREDUCE-727:
---------------------------------------
bq. 'bin/hadoop job -submit' would be the same functionality as 'bin/hadoop
jar', hence there is no dissonance?
It seems strange to me from a user's perspective since the rest of the "job"
commands target single jobs, and that one could submit 0, 1, or N jobs.
Going a bit off-topic from this JIRA, I think a very useful mapred command line
would be: mapred job -submit -jar foo.jar -mapper com.mycorp.FooMapper -reducer
com.mycorp.FooReducer -input blah -output blah -inputformat blah -outputformat
blah -numReducers 3 ... this makes sense as a generic "always one job
submitted" interface, whereas we could keep "hadoop jar foo.jar" to mean "run
this jar with all of the hadoop environment stuff set up"
> Move the bin/hadoop jar command over to bin/mapred
> --------------------------------------------------
>
> Key: MAPREDUCE-727
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-727
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Reporter: Arun C Murthy
> Fix For: 0.21.0
>
>
> Currently 'bin/hadoop jar' is used to submit jobs, we should move it over to
> bin/mapred.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
