Skip to content

Clustering, Lightly

A cluster is several Proxmox nodes managed as one: one web UI shows all machines, and VMs can move between physical hosts. This chapter is deliberately light — most home labs never need it — but you should know what it gives and what it costs, so you recognize the day you do need it.

What a cluster gives you

  • One pane of glass: the resource tree shows every node and every guest. Manage everything from one login.
  • Migration: move a running VM from one host to another (qm migrate — verified against the official man page). Patch a host without downtime.
  • High availability (optional): if a node dies, its VMs restart on surviving nodes. This is the enterprise headline feature; at home it's usually overkill.

What it demands

This is the part people skip, so say it plainly:

  • Three nodes (or two plus a quorum device) for reliable decisions. Two nodes alone can't safely decide who's in charge after a network split — this is the quorum problem, and Proxmox will refuse to do unsafe things rather than guess.
  • Shared storage for migration to be useful. Moving a VM between hosts is only seamless if both hosts see the same disks (NFS, Ceph) — otherwise the disk has to be copied too.
  • A fast, reliable interconnect. Cluster chatter is latency-sensitive; Wi-Fi between nodes is asking for pain.

From the docs: these requirements follow the official Proxmox cluster documentation. The short version: clustering rewards wired, three-node, shared-storage setups and punishes everything else.

Joining, in outline

From the docs — the UI path: on the first node, Datacenter → Cluster → Create Cluster. On each additional node, Datacenter → Cluster → Join Cluster, pasting the join information from the first node. Nodes must be on the same LAN with reliable networking, running the same Proxmox version.

The honest home-lab take

One good Proxmox box with scheduled backups to a NAS covers most home needs. Add a second node when you want to tinker with migration; think about three when uptime genuinely matters to you. Don't build a cluster because it sounds cool — build it when you can name the problem it solves.

The one idea to keep

Clusters trade simplicity for mobility and resilience, and they demand three nodes, shared storage, and good networking to do it safely. For one box, you lose nothing by skipping this chapter. Next: when things go wrong, in Troubleshooting.