[katello-devel] How are we gonna deploy

[Jeff Weiss]

On Thu, 2011-08-18 at 16:38 +0200, Lukas Zapletal wrote:
> On 08/18/2011 04:26 PM, Jeff Weiss wrote:
> >> An observer here, but i wanted to suggest torquebox.   The jboss
> >> >  container would work for candlepin and the jruby based rails container
> >> >  has some impressive performance stats.
> > Wouldn't that involve switching from the MRI interpreter to running
> > katello on JRuby?  How difficult is that migration?  I have heard it's
> > not too bad, but I can't imagine how switching interpreters (especially
> > ones as different as MRI and JVM/jruby) wouldn't cause a bunch of
> > regressions.  I'm not saying I'm opposed to the switch, if it results in
> > good performance, and stability and low resource usage.
> 
> Yeah that would be a massive change. I also wonder if jboss and/or 
> torquebox are in RHEL or Fedora.

They do not appear to be.

> It looks like the bet option would be to use apache2 and thin or nginx 
> and thin.
> 

I'd vote for apache2 for several reasons:
1) It's already installed and running because pulp is using it.
2) (speculating) nginx has no real performance advantage, the bottleneck
is the rails app, not the front end.
3) It's the standard http server for linux.

Jeff