When a node becomes unresponsive for various reasons, Proxmox HA waits for 60 seconds before fencing the faulty node. Fencing prevents cluster services from coming online during that time. Then, HA moves the VM to the next available node in the HA member group. As of Proxmox VE 5.0, LXC containers cannot be live-migrated. So, HA will stop all LXC containers and then move them to the next node. Even if the node with VMs is still powered up but loses network connectivity, Proxmox VA will try to move all VMs out of the node to a different node.

Once the faulty node comes back online, however, HA will not automatically move the VMs back to the original node. This must be done manually. But a VM can only be moved manually if HA is disabled for that VM. So we have to disable HA first and then move to the original node and enable HA on the VM again. As we can see, Proxmox HA likes to manage everything on its own, although it adds little annoyances to manually performing certain functions. HA is focused on maintaining uptime, which it does suitably. Later in this chapter, we will see how to configure HA for virtual machines.