Out of curiosity, why are you running 2 k8s nodes on a single physical machine? Why not just run k8s on the physical machine? K8s should be able to manage all the processes correctly for you, and in addition you won't have the resource overheads of 2 VMs.

The only reason I can see having a VM as an abstraction layer is if your hypervisor runs in a cluster and it's able to move volumes across nodes, but even then, it feels like having a single VM running on the nodes that take up the full physical resources makes more sense.