Open Stack

I am using icehouse on Ubuntu 14.04 from the canonical repository.

I'm using nova/neutron(ovs+vxlan). My system has 6 blades in a single enclosure.

if i get a power-blip or something like that, the blades start @
different speeds (different generations, bios settings, amount of ram,
that sort of thing).

It never returns to normal without logging in to each blade and
manually restarting services in the right order.

Has anyone else this issue? If so, how do you deal with it?

For example, the upstart config for neutron-plugin-openvswitch-agent
doesn't have a 'wait' or 'prereq' for openvswitch-switch to be up and
running.

The worst part about this is that it usually silently fails. I end up
with all my services running, but not working. E.g. I will end up with
2 interfaces instead of 4 on the host (qvo/qve/qbr/tap... usually
missing the qve). So then the DHCP in my instance will fail.

On the controller, the dnsmasq might come up, but the underling
interface is not there to bind to, and it silently just does nothing
but is running.

So, i can't be the only one to find this very brittle. Is there some
clever solution? Something I am missing? some 'make-it-all-good-now'
script i can run?