Comment by darragh-oreilly for Hi all, I'm running a three-node Havana installation; two compute nodes and a controller. After initial installation, the setup worked just fine for a few weeks. Then I started having networking problems where previously responsive instances were unable to communicate with the outside world, and new instances couldn't get DHCP addresses. Much Googling yielded no answers that actually worked. I did manage to get it back up and running, but I'm not sure how to be honest. Again, after a while, networking just died. This time, I'm noticing that OVS doesn't have any of the flows that it used to. Restarting the Neutron services and/or OVS makes no difference. Last time, I rebooted the compute and controller nodes and everything came back up fine (and this shouldn't be necessary for that matter). This time, I've rebooted repeatedly to no effect. I'm not really sure what you need from me in order to troubleshoot the situation, but ask and I will deliver. Thanks!! Controller/Networking Node: ovs-vsctl show (note that I have 9 running instances, and they're not here) ovs-vswitchd.log (See all the weird entries that occur on restart) neutron.conf plugin.ini UPDATE: I just upgraded my installation from Havana to IceHouse, and I am no longer running into this issue, or several of the other issues that I experienced previously.

The bridge br-tun has no ports - I guess they were cleaned out when the l2 agent started on the reboot. Check the rows in the ovs_tunnel_endpoints table in the ovs database are correct. Is the L2 agent RPCing the server? Really need the l2 agent logs here.