[nova][neutron][cyborg] Bandwidth (and accel) providers are broken if CONF.host is set

Bence Romsics bence.romsics at gmail.com
Wed Nov 27 14:28:53 UTC 2019


Matt wrote:
> The point of the API change is to let neutron do everything
> automatically and avoid additional configuration, right? The
> configuration on the neutron side was meant to be a workaround.

I had the following reasons to propose using configuration:

(1) Using a nova API from neutron-server means we'd have a startup
ordering dependency. Neutron would have to wait for nova to start up.
So far it had to wait for placement - there we did not have an
alternative. Here we have.

(2) We cannot use a new nova API for fixing this bug in stein and
train. If the backportable fix works on master too, do we want to
create a different fix for master?

Sean wrote:
> > (1) Extend ovs and sriov agents config with
> > 'resource_provider_hypervisors' for example:
> >
> > ml2_conf.ini
> > [ovs]
> > bridge_mappings = physnet0:br-physnet0,...
> > resource_provider_bandwidths = br-physnet0:10000000:10000000,...
> > resource_provider_hypervisors = physnet0:hypervisor0,... # this is
> im guessing you are adding this for the ironic smart nic usecasue but i dont think this makes sense.

That's right. I made a mistake there. Thanks for catching it. I meant
keying resource_provider_hypervisors by the physdevs (the values in
mappings) not the physnets. For example:

resource_provider_hypervisors = br-physnet0:hypervisor0,...

I understand we don't have use for this right now (neutron with ironic
does not use placement), but if we don't allow the 1:N mapping here,
then we prohibit the future use of it too.


More information about the openstack-discuss mailing list