Hello All,

Is there a "community standard" / "best" way to deploy HBase to a cluster?  
We're in the process of setting up a ~15 node cluster and I'm curious how  
you all go about your deployments. Do you package the code into an RPM,  
place it into a central YUM repository, and then drive the install via  
Puppet? Or do you rsync the code and use shell scripts? Or all the above?

Thanks so much for your input!

~ Peter

I've used puppet for the job.  Adobe posted a set of puppet scripts a while
back that would deploy a tar.gz.  They were my starting point (here:
http://hstack.org/hstack-automated-deployment-using-puppet/).  Those
configs and values are pretty old so make sure you look at the docs for
whatever version you are deploying. Even if you deploy through rpm/deb I
would highly recommend some sort of config control; be it puppet, chef, or
cf engine.  The time taken to learn and set that up will pay itself back in
the long term. There are a lot of different moving pieces in a hadoop MR /
HDFS / HBase cluster and they will take some tuning.

On Wed, May 30, 2012 at 2:08 PM, Peter Naudus <[EMAIL PROTECTED]> wrote:

> Hello All,
>
> Is there a "community standard" / "best" way to deploy HBase to a cluster?
> We're in the process of setting up a ~15 node cluster and I'm curious how
> you all go about your deployments. Do you package the code into an RPM,
> place it into a central YUM repository, and then drive the install via
> Puppet? Or do you rsync the code and use shell scripts? Or all the above?
>
> Thanks so much for your input!
>
> ~ Peter
>

We package into RPMs and manage configuration with Puppet.

Have a look at Apache Bigtop (incubating) for RPM and DEB package build harness and integration/smoke tests. It's a promising project.

You can also Google around for presentations on this topic by Arvind at StumbleUpon. He talks about how they use tarballs for deploy and Puppet for configuration management and Fabric for coordination.

I'd say RPM package management and Puppet configuration management is common, maybe most common, but that last bit is just a guess.

Hope that helps,

     - Any

On May 30, 2012, at 2:08 PM, "Peter Naudus" <[EMAIL PROTECTED]> wrote:

> Hello All,
>
> Is there a "community standard" / "best" way to deploy HBase to a cluster? We're in the process of setting up a ~15 node cluster and I'm curious how you all go about your deployments. Do you package the code into an RPM, place it into a central YUM repository, and then drive the install via Puppet? Or do you rsync the code and use shell scripts? Or all the above?
>
> Thanks so much for your input!
>
> ~ Peter