[openstack-dev] [Neutron][kuryr] mutihost networking with nova vm as docker host

Antoni Segura Puimedon toni+openstackml at midokura.com
Thu Nov 5 12:18:36 UTC 2015


On Thu, Nov 5, 2015 at 10:38 AM, Vikas Choudhary <choudharyvikas16 at gmail.com
> wrote:

> ++[Neutron] tag
>
>
> On Thu, Nov 5, 2015 at 11:33 AM, Vikas Choudhary <
> choudharyvikas16 at gmail.com> wrote:
>
>> Hi All,
>>
>> I would appreciate inputs on following queries:
>> 1. Are we assuming nova bm nodes to be docker host for now?
>>
>
Yes. That's the assumption for deployments as of now before we tackle
containers running on
more complicated deployment topologies (like containers running inside
tenant VMs).


>
>> If Not:
>>
>
When we go for other kinds of deployments,


>          - Assuming nova vm as docker host and ovs as networking plugin:
>>             This line is from the etherpad[1], "Eachdriver would have an
>> executable that receives the name of the veth pair that has to be bound to
>> the overlay" .
>>
>
The binding will obviously have to change for such deployments


>             Query 1:  As per current ovs binding proposals by Feisky[2]
>> and Diga[3], vif seems to be binding with br-int on vm. I am unable to
>> understand how overlay will work. AFAICT , neutron will configure br-tun of
>> compute machines ovs only. How overlay(br-tun) configuration will happen
>> inside vm ?
>>
>>              Query 2: Are we having double encapsulation(both at vm and
>> compute)? Is not it possible to bind vif into compute host br-int?
>>
>>              Query3: I did not see subnet tags for network plugin being
>> passed in any of the binding patches[2][3][4]. Dont we need that?
>>
>
The spec for containers on VMs has not yet been drafted and we are open for
proposals and discussion. I would like to have more than one spec proposal
for it and to try to achieve community consensus before the new year for
what's the best way to go.

Currently it seems that the approaches that will be proposed are:
- ovn-like solution with vlan tag per port [5]
- routed solution with port per VM as explained by Brenden Blanco [6]

I'm hoping that we will arrive to something in between or perhaps more
complete than either of those options.

[5] http://docs.openstack.org/developer/networking-ovn/containers.html
[6] https://gist.github.com/drzaeus77/89aa3db154c688a15ee6

Regards,

Toni

>
>>
>> [1]  https://etherpad.openstack.org/p/Kuryr_vif_binding_unbinding
>> [2]  https://review.openstack.org/#/c/241558/
>> [3]  https://review.openstack.org/#/c/232948/1
>> [4]  https://review.openstack.org/#/c/227972/
>>
>>
>> -Vikas Choudhary
>>
>
>
> __________________________________________________________________________
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20151105/5eacba6b/attachment.html>


More information about the OpenStack-dev mailing list