Open Stack

Hi! I would like some help with a weird behavior on l3-agent of a network
controller node. My scenario is on the figure attached.

1) Why l3-agent is assigning ipv6 addresses from internal tenant network on
qg_ interfaces of virtual routers?

When I built the environment these qg_ interfaces only had fe80::
addresses. Now, I think maybe due some upgrade on neutron, I'm seeing these
addresses. I didn't understand that.

I've just cleaned up openvswitch database with neutron-ovs-cleanup, checked
that no qr or qg interface were created. After that, with no interface,
I've started neutron-l3-agent and the interfaces were created and the
addresses assigned.

2) What are the recommendation to setup an ipv6 environment like this,
related to routes?

For example, in my firewall/router I have a static route for a range of
networks (those managed by prefix delegation server). Every packet to one
of these networks goes via the address of br_ex (next hop). Arriving at
br_ex, this bridge has routes to each network already delegated using LLA
of the tenant virtual router as next hop. These routes on br_ex are created
by dibbler notify script.

Is this the right way?

I've to keep a cron always checking if these routes are ok.

Tks!

- JLC
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20161207/ab610938/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: esquema_vrouter.png
Type: image/png
Size: 78625 bytes
Desc: not available
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20161207/ab610938/attachment.png>