Controlled handling of an outage is always your first best choice. I would shut down or save all virtual machines in this scenario before performing any operations.
If the nodes lose connectivity to the network but retain connectivity to storage, they will gracefully shut down or save their VMs, depending on the VM's configuration. If they lose network and storage, they will pause the virtual machines for about a minute, then decide that the storage has failed and hard power off the virtual machines.