Open Stack

On Tue, Jul 29, 2014 at 10:12 AM, Jonathan Proulx <jon at jonproulx.com> wrote:
> Hi All,
>
> Would making an nova-network mechanism driver for the ml2 plugin be possible?
>
This has been discussed a bit, and yes, in theory this is possible.
Nachi has started looking into this as far as I know. I also think
this may have come up during the nova mid-cycle which is going on this
week.

> I'm an operator not a developer so apologies if this has been
> discussed and is either planned or impossible, but a quick web search
> didn't hit anything.
>
> As an operator I would envision this a a transition mechanism, which
> AFAIK is still lacking, between nova network and neutron.
>
> If a DB transition scrip similar to the ovs->ml2 conversion could be
> created, operator could transition their controller/network-nodes to
> neutron while initially leaving the compute nodes with active
> nova-network configs active.  It's a much simpler matter for most
> operators I think to then do rolling upgrades of compute hosts to
> proper neutron agents either by live migrating existing VMs or simply
> through attrition.  And this would preserve continuity of VMs through
> the upgrade (these may be cattle but you still don't want to slaughter
> all  of them at once!)
>
> This is no longer my use case as I jumped into neutron with Grizzly,
> but having just transitioned to Icehouse and ML2, it got me to
> thinking.  If this sounds feasible from a development standpoint I'd
> recommend taking the discussion to the operators list to see if others
> share my opinion before doing major work in that direction.
>
This is a really good idea actually, thanks for sharing this. One
place where I feel like we haven't had enough input in the
nova-network/neutron parity discussion is around input from operators.

Thanks,
Kyle

> Just a thought,
> -Jon
>
> _______________________________________________
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev