Proxmox Cluster: Running High Availability at Home
Proxmox clustering lets VMs migrate between nodes automatically if one goes down. Here's how it works, what you actually need to set it up, and whether it's worth the complexity.
This post contains affiliate links. If you buy through them, I earn a small commission at no extra cost to you.
Most homelabs don’t need high availability. A single Proxmox node running a dozen VMs is fine — if it goes down, you lose access until it comes back up, which is usually acceptable for a home setup.
But if you want to understand how HA actually works, or if you’re running services that people in your household depend on, a Proxmox cluster changes the equation. When a node fails, its VMs restart automatically on another node. No manual intervention.
Here’s what it takes to build that, and an honest assessment of whether it’s worth it.
What Proxmox Clustering Actually Does
A Proxmox cluster is a group of Proxmox nodes that share management state. From the web UI, you see all nodes and all VMs in one place. VMs can be live-migrated between nodes while running. And with High Availability enabled, the cluster monitors node health and restarts VMs on healthy nodes if a node fails.
The minimum viable cluster is three nodes. This isn’t arbitrary — the cluster uses a quorum system to prevent split-brain scenarios. With two nodes, if they lose connectivity with each other, neither can be sure the other is truly down (versus just network partitioned). Three nodes means a majority vote is always possible.
You can technically run a two-node cluster with a QDevice (quorum device) — a lightweight third participant that only votes but doesn’t host VMs. A Raspberry Pi or old thin client works for this.
Hardware Requirements
For a three-node home cluster:
Three machines with matching or compatible hardware. They don’t need to be identical, but CPU architecture should match (all Intel or all AMD) and the VMs need to run on all nodes. If you’re live-migrating VMs, the destination CPU needs to support at least the same instruction set as the source.
Shared storage, or a way around it. VMs that live-migrate need their disk images accessible from both the source and destination node. In enterprise setups, this is a SAN or Ceph cluster. In a homelab, you have a few options:
- Ceph — Proxmox has built-in Ceph support; it distributes storage across nodes. Needs three nodes and dedicated drives. Gets complex fast.
- NFS shared storage — a NAS that all nodes mount via NFS. Simpler than Ceph, adds a single point of failure (the NAS).
- Replication without shared storage — Proxmox’s built-in VM replication syncs disk images to another node on a schedule. Not instant failover, but close enough for many use cases.
A dedicated cluster/sync network. Don’t run cluster traffic on your main LAN. A second NIC on each node, connected via a cheap switch, keeps cluster communication isolated. The TP-Link TL-SG105 is $20 and fine for this.
Setting Up the Cluster
On your first node:
pvecm create my-homelab-clusterOn each additional node (after ensuring they can reach the first node):
pvecm add <first-node-ip>That’s the basic cluster formation. The web UI will show all nodes under a single panel.
Enable HA for a VM in the Proxmox UI: select the VM, go to More > Manage HA, and add it to an HA group. Assign it a priority if you want to prefer certain nodes.
The Proxmox HA Watchdog
For HA to work, Proxmox needs to be sure a failed node is actually dead before restarting its VMs on another node. This is the “fencing” problem — you don’t want two nodes both running the same VM simultaneously (that causes data corruption).
Proxmox handles this with a watchdog timer. If a node stops sending its keepalive, the watchdog expires and the node fences itself (reboots). Only then does the cluster restart the VMs on healthy nodes.
This means a node failure takes 1-2 minutes to resolve — not instant, but automatic.
Three Mini PCs as a Cluster
A realistic home HA cluster:
Three Beelink EQ12 Pro mini PCs (~$180 each), a 5-port switch for the cluster network, and a NAS or Ceph for shared storage. Total cost around $600-700 depending on storage.
For shared storage without Ceph complexity, a 2-bay NAS with NFS export works. Mount the NFS share on all three nodes as a Proxmox storage pool, store VM disks there, and you can live-migrate VMs between nodes.
What HA Actually Gets You
With a three-node cluster and NFS shared storage:
- A node goes down at 3am → its VMs restart on another node automatically within 2-3 minutes
- You wake up and check Proxmox → the cluster shows the node offline, VMs moved
- No manual intervention required
What it doesn’t protect against:
- Your NAS going down (now all VMs lose their disks)
- Network failure that partitions the cluster
- Correlated failures (power outage hits all three nodes)
For true home HA, you want the NAS on a UPS, the cluster nodes on a UPS, and a network switch that’s reliable. That’s more infrastructure to maintain.
Is It Worth It for Home Use?
Honestly, for most home use cases: no. A single Proxmox node is simpler to maintain, cheaper to run, and sufficient for personal services. If it goes down, you restart it — and since it’s at home, that’s a 30-second walk to the server.
HA is worth it if:
- You’re running services that other household members depend on and would notice an outage
- You want the experience for professional development (it’s genuinely useful to understand)
- You’re building toward a setup that warrants the complexity
The learning value is high regardless. Running a Proxmox cluster at home — even for a few months — teaches you things about quorum, fencing, live migration, and shared storage that you can’t get from reading documentation.
Start with a single node. When you hit a reason to need HA, you’ll know.