Open Stack

> We currently have two OpenStack management nodes (mgmt1 & mgmt2) with keystone, nova, glance etc.
> the access is through a VIP and a HAProxy balanced  by round robin.
> When I use the nova CLI for example “nova --debug list”, I can see that the initial keystone request is spread around the different mgmt nodes.
> But the answer from nova is always from the first node mgmt1.
> I’m assuming it just takes the first nova-compute endpoint in the list. Is there a work around?
> Should the nova-compute endpoint also be the VIP?

Yes, you should put all endpoints and keystone itself behind the HAProxy VIPs.
There is no magic on the client or server side that would randomly choose a endpoint for you from the list.

Cheers,
Robert van Leeuwen
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20150331/c7dd2d42/attachment.html>