Someone reported some problems running through: http://openstack.redhat.com/Neutron_with_existing_external_network I thought I would walk through it myself first to make sure that (a) it was correct and (b) that I remembered all the steps, but I've run into a puzzling problem: My target system is itself an OpenStack instance, which means that once br-ex is configured it really need to have the MAC address that was previously exposed by eth0, because otherwise traffic will be blocked by the MAC filtering rules attached to the instance's tap device: -A neutron-openvswi-s55439d7d-a -s 10.0.0.8/32 -m mac --mac-source FA:16:3E:EF:91:EC -j RETURN -A neutron-openvswi-s55439d7d-a -j DROP I have set MACADDR in br-ex, which works just fine until I restart neutron-openvswitch-agent (or, you know, reboot the instance), at which point the MAC address on br-ex changes any everything stops working. I've been poking through the code for a bit and I can't find either the source or an explanation for this behavior. It would be great if a wiser set of eyes could shed some light on this. Cheers, -- Lars Kellogg-Stedman <lars(a)redhat.com&gt; | larsks @ irc Cloud Engineering / OpenStack | " " @ twitter