You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@oozie.apache.org by Rahul Bhattacharjee <ra...@gmail.com> on 2013/05/07 07:04:05 UTC

Oozie MR jobs.

Hi,

I have a question regarding Oozie, I am running a Oozie workflow , which
does a simple distcp and then runs a MR job.

It works just fine as expected. The question is that every action is
transformed into a MR job and the actual action specified in the workflow
is then fired from within the MR job (actually a Map only job) that Oozie
has created.

Example - my MR job is getting triggered from within the Map only job  that
oozie is creating and submitting to the cluster.

Why is this? Cann't the oozie server perform the individual actions itself
without wrapping it in a M only job, or is it just of off load the oozie
server?

Thanks,
Rahul

Re: Oozie MR jobs.

Posted by Harsh J <ha...@cloudera.com>.
While load is a secondary reason, the primary reason we spawn a
launcher job first is so that the Oozie server does not run *any* user
code (such as job drivers) and threaten its own stability.

On Tue, May 7, 2013 at 10:34 AM, Rahul Bhattacharjee
<ra...@gmail.com> wrote:
> Hi,
>
> I have a question regarding Oozie, I am running a Oozie workflow , which
> does a simple distcp and then runs a MR job.
>
> It works just fine as expected. The question is that every action is
> transformed into a MR job and the actual action specified in the workflow
> is then fired from within the MR job (actually a Map only job) that Oozie
> has created.
>
> Example - my MR job is getting triggered from within the Map only job  that
> oozie is creating and submitting to the cluster.
>
> Why is this? Cann't the oozie server perform the individual actions itself
> without wrapping it in a M only job, or is it just of off load the oozie
> server?
>
> Thanks,
> Rahul



-- 
Harsh J