Comment 1 for bug 973609

This should only happen if network is blocking for a long time or the network node is not connected for some reason. Are you running only one nova-network? If so it could be getting overloaded with requests. You might try using multi_host mode and running one nova-network on each compute host.
you might try:
rabbitmqctl list_queues

it should show you how many messages are waiting in the queue. If the number is greater than 0 either the worker is blocking doing some work or it can't keep up with the messages.

FYI you can also change the timeout using:
rpc_response_timeout=<timeout_seconds>

so you could set it much higher and see if that helps.