I'm using tripleo to deploy openstack 16, which supports NovaReservedHostMemory to reserve memory by role, it's also said that NovaComputeCpuSharedSet and NovaComputeCpuDedicatedSet can be used to set where vms are scheduled. But all the cpu set seems to require that you know number of cpus, so you can say
NovaComputeCpuSharedSet: "8-111" # reserve first 8 out of 112 cpus
What if I don't know how many cpus or it's not the same for each compute nodes?
I did try the per-node.yaml way of using NodeDataLookup to set hieradata (nova::compute::cpu_shared_set) but it's totally ignored by deployment.
