Open paulczar opened 10 years ago
We could do it by ComputeCapability ( http://docs.openstack.org/developer/nova/devref/filter_scheduler.html ) if there isn't some already built in method of expressing which HVs are capable of satisfying the needs of a given flavor. I would imagine flavors would be where one would determine ironic vs kvm vs docker. We may have to use the extra_specs data in the flavor to accomplish this.
doing it for Ironic+other could be interesting as Ironic has some different scheduling options:
http://docs.openstack.org/developer/ironic/deploy/install-guide.html#configure-compute-service-to-use-the-bare-metal-service
KVM + Baremetal + Docker.
To start with to support a homogenous cluster we should create a roll for each of:
compute-kvm, compute-ironic, compute-docker and genericize the compute role so that it can be called as a dependency of each ( perhaps it defaults to KVM for backwards compat? )
The we need to figure out what is needed to run more than one hypervisor class in a cluster, say KVM+Ironic. or KVM + Docker. used to require different AZs ... but is that still the case?