Comment 6 for bug 1353962

Even if tempest leaked every server it created (which is definitely not happening) if the network allocates a /24 there should still be plenty of ips available. From the logs above tempest only allocated a total of 171 servers:

http://logs.openstack.org/23/112523/1/check/check-tempest-dsvm-postgres-full/acac6d9/logs/qemu/

But the exception is a quota one, the default tenant quota for a fixed ips is definitely more than 2. (otherwise we'd have much larger issues) I'm pretty sure we set it to -1, but I'm not sure. Looking at the test class in question:

http://git.openstack.org/cgit/openstack/tempest/tree/tempest/api/compute/admin/test_servers_negative.py

The the tenant for that class will only have 2 server create calls. One in setUpClass and one in this test. I have a hard time believing 2 fixed ips exceeded the quota limit. Looking at the tests running around the same time, starting at:

http://logs.openstack.org/23/112523/1/check/check-tempest-dsvm-postgres-full/acac6d9/console.html.gz#_2014-08-07_09_42_03_634

which was the last test from a previous class on the same thread, until the same thing after the failed test:

http://logs.openstack.org/23/112523/1/check/check-tempest-dsvm-postgres-full/acac6d9/console.html.gz#_2014-08-07_09_42_27_778

I don't see anything that should have accidentally changed the quota to be < 2. I guess there could be a more hidden cross interaction because there are a bunch of admin api tests running around the same time (the joys of a default alphabetical test order), but I don't see one yet.