Open Stack

On Thu, Dec 19, 2013 at 12:48 AM, Jay Pipes <jaypipes at gmail.com> wrote:
> On 12/19/2013 12:36 AM, Joe Topjian wrote:

> Sure. So this patch added in the multiplexed Neutron server functionality:
>
> https://review.openstack.org/#/c/37131/
>
> It's not in Havana, but we backport that sucker into our Grizzly and Havana
> deployments:
>
> https://gist.github.com/alanmeadows/7770570
> https://review.openstack.org/#/c/63020/
>
> I would highly suggest you do the same, set workers=10 or so and retry your
> concurrent launch scenario...

+1

I too back ported that to my Grizzly install, really hoped it would
make it to Havana.

Without that patch I could not live (and possibly some keystone multi
worker stuff, basicly everything at
http://blog.gridcentric.com/bid/318277/Boosting-OpenStack-s-Parallel-Performance).
 I'm running a server per core (well thread really)  on my controller
node for me that's 24 and seems to handle upto the 1k instance range
which is as far as I've really gone.

I was seeing similar lockups when launching about 10 instances at once
& my users like to do 100's.

I think the cascading failure is because many tools try and get info
from neutron , that's what I was seeing anyway.  After I got the
neutron-server bits into a working multi server the next thing that
bound up was the neutron-dhcp-agent, I also found neutron makes much
greater demands on keystone than nova-networking does so needed to
scale out there too, though that wasn't noticeable until somewhere
around 300 instances or so, which is about 1/3 of what I was
targeting, so likely fine for your class room case but something you
want to think about for migrating/creating larger environments.

-Jon