Comment 22 for bug 1160442

It's not clear that multiple fixed ips being allocated for a VM when Nova and Neutron are experiencing heavy load is solved by the merged patch or whether it was merely masked. At the least, there needs to be documentation added that makes it clear how to set max_pool_size, max_overflow and pool_timeout according to expected load.

Suggestions via irc suggest max_pool_size should be set to x+1 where x is the expected maximum concurrent expected requests, with pool_timeout = 2.