Comment 4 for bug 1017795

ttx: Both of the attached patches replace the compute_api.get calls with a single compute_api.get_all call which should be limited to the context and only contain servers for each user/tenant.

Like I mentioned above an alternate solution might be leave the compute_api.get call as is and simply limit the number of UUIDs we allow to be specified in the scheduler_hints.

We'll see what other think....