Posted on June 7, 2017 6:19 pm
 |  Asked by Jon Nicholson
Print Friendly, PDF & Email

We’re deploying Openstack Neutron (based on Redhat Openstack Platform 10) with arista switches. In this pass we’re only using the Layer 2 driver, allowing Neutron to take care of the L3.

With this configuration neutron creates a network for the routers to communicate VRRP over, we see this being created on our CVX controller, but as the network never gets any ports associated with it so CVX does not communicate this to the physical switches.

This means that the neutron routers go “split brain” as they have no l2 connection between them.

I’ve got a case open with support about this, has anyone else seen this behaviour or know how to fix it?

Posted by Anuraag
Answered on June 7, 2017 6:20 pm


The Arista ML2 driver doesn’t provision the switch ports connected to the Neutron network nodes. As those nodes are running at a fixed/pre-determined location,  configuring the switch port statically (with the range of allowed VLANs configured for Neutron to allocate) at the time of deployment should be sufficient.


The problem is that the leaf switches have no l2 connection between them and the controller nodes are on different leaf switches for redundancy.

I’ve created a work-around by using Software VXLan for the HA networks between Controllers, but this does seem to be a bit of a fudge….

(Jon Nicholson at June 26, 2017 8:44 am)

Post your Answer

You must be logged in to post an answer.