Comment 4 for bug 1835958

Tom,

There are two things that _should_ help:

- Disable power state checks in nova.conf.
- Break up your ESXi clusters into smaller ones.

The problem is the distribution of the power state checks themselves, if there are ESXi clusters that are large (greater than say 500 VMs) the nova-compute instances, in a few cases, begin to experience service degradation due to the number of power state checks. I wrote about tangentially at http://www.revolutionlabs.net/vio_resource_schedulers_part2/.