Comment 5 for bug 999698

Can you show me the exact command(s) you are using to restart all services at the same time? It's starting to sound like:

1) When nova-compute gets restarted, it's making a call out to nova-network.

2) nova-network is getting restarted before this call is completed, resulting in the timeout error that occurred.

3) This code path in nova-compute is not properly handling the timeout exception, resulting in the service stopping.

So if it's just a bash for loop restarting all of the services, try restarting network before compute.