Comment 27 for bug 1160442

Saw this report just now, but we had the same issue and were caused by slow quantum server. In addition to port created by scheduler retrying request on a different compute, we also have a bug in the quantum client as reported here https://bugs.launchpad.net/neutron/+bug/1267649. We were able to fix this temporarily by increasing the timeout value for the quantum_url_timeout property in nova compute to larger values, 2 minutes in our case.