bad network format error happen sometimes with no reason

[li,chen]

I'm doing a instance provision test based on Grizzly.

While some times the first instance I created always met the error:
 Bad networks format (HTTP 400)

I checked the code in /usr/lib/python2.7/dist-packages/nova/api/openstack/compute/servers.py, looks like it happens in function _get_requested_networks:
                # For quantumv2, requestd_networks
                # should be tuple of (network_uuid, fixed_ip, port_id)
                if self._is_quantum_v2():
                    networks.append((network_uuid, address, port_id))
                else:

And it is reported as a TypeErr exception.

But, all later instances I have used the same command to boot up just works fine.
And, I add some debug output in the code, restart nova-api several times, the issue just gone.

Anyone know why this happen ?

Thanks.
-chen

[Christopher Yeoh (cyeoh-0)]

Could you add the full backtrace from the log file (as well as some of the log file before be useful) for the error you are seeing? Also if you restart all the nova services rather than just nova-api does it occur again?