Proxmox and Ceph: building a virtualisation cluster that never stops

When a client asks for a server infrastructure that “does not go down”, the answer is no longer a big server with a SAN. It is a cluster of identical machines that share their storage and take over from each other automatically. Here is the architecture we deploy most often, and the choices that make it hold.

The principle: no single part

In a Proxmox VE cluster with Ceph, every node brings compute and disks. Ceph replicates every block of data across several nodes. If a server stops, its virtual machines restart elsewhere in under a minute and the data is still there. There is no SAN to buy, no single controller everything depends on.

The sizing that matters

• Three nodes minimum, five for performance. Quorum requires an odd number.
• A dedicated Ceph network, 10 GbE at minimum, 25 GbE once volumes exceed a few tens of terabytes. The most common mistake is running replication over the user network.
• NVMe SSDs for OSDs or, on a tight budget, enterprise SSDs with an NVMe journal.
• 3× replication by default. Erasure coding saves space but costs latency; we keep it for archives.

Migrating from VMware

We import virtual machines one by one, outside production hours, converting drivers to VirtIO. The most critical servers go last, once the cluster has been proven by the others. No global outage, ever.

Backup is still essential

Ceph protects against hardware failure, not human error or ransomware. Proxmox Backup Server or Veeam, with an immutable off-site copy, completes the architecture.

What it costs

No per-socket licence, no mandatory annual renewal. An optional Proxmox Enterprise subscription for the stable update repository and vendor support. For most SMBs, the saving versus VMware pays for the hardware in under two years.

See our virtualisation and server infrastructure service and our Proxmox partner page.