It worked. There's runtime logical bugs left to squash. But the scheduler
that I wrote has started getting used. Like you said, its more of a Java
thing.

To use my custom scheduler, I made the JAR using Eclipse. Just did a mvn
eclipse:clean and mvn eclipse:eclipse to get the project workable in
Eclipse. Then just a right-click and "Export -> Jar" did the trick. Thank
you for the quick response, Harsh :)

Sincerely,
Sreejith R


On Tue, Jul 9, 2013 at 8:31 AM, Harsh J <[email protected]> wrote:

> To compile it to a jar, rely on your build framework. Or pack it up
> with "jar cf jarfile dirs…"? This is more of a Java question, can be
> googled and is specific to the tool you use.
>
> To set it into effect, for 1.x, drop the jar into the
> $HADOOP_HOME/lib/ directory and set the
> mapred.jobtracker.taskScheduler property in mapred-site.xml to your
> fully qualified classname, and restart JT.
>
> P.s. I'd recommend writing a 2.x scheduler as thats where the
> project's moving and a load of things have changed. The 1.x scheduler
> won't be compatible with the new 2.x releases.
>
> On Mon, Jul 8, 2013 at 4:05 AM, Sreejith Ramakrishnan
> <[email protected]> wrote:
> > Hey,
> >
> > I just wrote a task scheduler for hadoop (1.0.3). It's based on MR1
> > syntax.
> >
> > I want to know how to compile it (to a jar?) and where to put the
> compiled
> > target so that I can make hadoop use it.  Hope some experts can shed some
> > light.
> >
> > Sincerely,
> > Sreejith R
> >
> > P. S: I had asked another question recently. I got very good answers.  I
> > was so busy I couldn't even thank them. Special thanks to Kun Ling
>
>
>
> --
> Harsh J
>

Reply via email to