Comment 14 for bug 903199

Hi all,

I had to face the same problem on a setup running a VlanManager: high cpu and memory usage, service unable to contact rabbitmq, virtual machine not spawning correctly because not getting a fixed_ip.
I did try to upgrade eventlet to trunk but it did not solve the problem.
I found a workaround faking the result in the function fixed_ip_disassociate_all_by_timeout in nova/db/sqlalchemy/api.py
Basically, if that function does nothing, my nova network runs smoothly even during stress tests.
I have to deallocate fixed_ips running a sql query by cron.