[Openstack-operators] Multiple nova-computers per host?
azpekt at gmail.com
Tue Aug 6 09:38:11 UTC 2013
On 07/30, Michael Still wrote:
> I've just read a libvirt code review which implies to me that people
> are deploying multiple nova-compute processes per physical compute
> host to work around scaling concerns.
> I have some concern that this might be bad idea, because I can think
> of a variety of race conditions you'd experience if you did that. It
> _might_ be possible to work around some of these race conditions by
> playing with periodic task configuration, but it wouldn't help with
> some others.
> Is anyone here deploying in this way?
Frankly saying, I`m not deploying my copute nodes in that fashion; but can you pls elaborate, what you mean by scaling issues? I thought, that having one nova-compute process on host is sufficient, and it will scale up to all available h/w resources - or am I missing somehthing here?
More information about the OpenStack-operators