Comment 23 for bug 1818239

I'm at a tad late (5+ years after the last comment), but in case it helps, we just ran into a situation where a user (not cloud admin) exhausted their IP allocations in a user-defined subnet, which caused an error in the nova-compute.log:

NoMoreFixedIps: No fixed IP addresses available for network: 62c7b2e2-4406-4dda-9a01-f0ec623335d4

which caused the BuildFailureWeigher to offset the weight by -1000000.

This was in a Stein cluster, so quite old. A newer version may have this disabled by default, but we had to manually disable it. Regardless, it caused all other tenants' VM deployments to fail since every compute node that tried to deploy the above VM resulted in the above error, causing each compute node to be assigned a large negative weight.

Just wanted to mention our experience in case it helped someone else who was searching for the problem.