Open Stack

Excerpts from Ian McLeod's message of 2013-08-08 07:29:14 -0700:
> On Wed, 2013-08-07 at 11:00 -0700, Clint Byrum wrote:
> > Flexibility can also be looked at as complexity. Complexity in Nova
> > is expensive in many ways. But just poking a few more holes into
> > Nova via the API would be a simple and effective way to manage that
> > complexity. Exposing some lower level things would get you the needed
> > pieces for your iPXE setup, and Monitoring disk IO is what ceilometer
> > does. Is there something else non-obvious that you need inside Nova?
> 
> If Ceilometer can provide the disk IO data, and the Nova API can be
> expanded to allow direct specification of a kernel, ramdisk _and_ kernel
> command line for Xen PV (and ideally, any other backing hypervisor that
> supports direct boot), then we'd likely have what we need to do this
> outside of Nova.
> 
> There may be some performance penalty that comes with adding an extra
> "hop" when moving around kernels and ramdisks.  (I'm guessing the path
> would become origin -> glance -> compute node versus, potentially,
> origin directly to compute node.)
> 

This hop penalty is already paid every time an image is booted for the
first time on a compute host. OpenStack has a vested interest in making
this as fast as possible. Another option is to look at what Ironic is
doing to address the same problem.