Today Karbon deploy K8s cluster (as a service) without extra features.
We are using GPU pass-through, so when we get bare k8s cluster there are need for extra deployments (manually) just to make GPU accessible as a resource to K8s-cluster.
anything in road map for dealing with this?
desired state is to mark node as GPU-Node and all driver and plugins installation will be performed by Karbon (including different cuda driver versions)