On time or it's free

The Proxmox Cluster

Three Proxmox servers joined into one cluster with working high availability, proved by deliberately killing a node in a window you choose, so you watch the platform recover before you trust anything to it.

Three servers each running their own virtual machines are three separate single points of failure. When one dies at three in the morning, everything on it stays down until somebody notices and drives in.

We join your three Proxmox nodes into one cluster where a dead node is an inconvenience instead of an outage. All three reach the same storage, and the cluster gets a network to itself so a busy migration cannot knock it over. Then we break it on purpose: one node hard-killed in a window you choose, with the time from failure to recovery written down.

The engineers who build these have been standing up remote infrastructure since 2003, and they build this cluster pattern most months. You get an empty, proven platform; your own machines go on afterwards at per-machine prices, so you never pay a cluster rate to spin up a container.

What's included

  • Your three Proxmox servers joined into one cluster you run from a single screen, with a network of its own for cluster traffic so a busy night cannot knock it over
  • Storage all three nodes can reach, so a running machine moves between them in seconds without its disk having to follow
  • Automatic recovery: when a node dies, the machines it was running come back on one of the other two without anybody being woken up
  • Fencing set up so a node that half-fails is stopped before it can fight the survivors over the same disks
  • A test workload moved live across all six node pairs, then one node deliberately killed and the recovery timed, with the timings handed to you
  • The honest limits of your build in writing: what your storage can and cannot snapshot, what happens if two nodes go at once, and the safe order for rebooting one
  • A plain-English handover of the whole cluster, plus a snapshot of its configuration you can roll back to

How it works

  1. 01

    Book and send access

    Book online with your ABN or NZBN. We tell you exactly what access and material we need; the clock starts once we have it.

  2. 02

    We do the work

    An experienced employee who specialises in this task works through it, and asks before anything outside the scope.

  3. 03

    Delivered with evidence

    You get everything listed on this page, with a written account of what was done.

Before you order

Have these ready so we can start the moment you order. The delivery clock only runs while we have them.

  • Root or sudo SSH access to all three nodes, each running the same major version of Proxmox VE with no machines on it yet
  • Three servers with the same CPU maker, at least two network ports each, on one network at one site
  • Shared storage all three may already reach: an NFS export, an empty iSCSI LUN, or a ZFS pool with the same name on every node
  • Whether your switch ports are set up for link aggregation, and any VLANs the cluster should carry
  • A two-hour window when we may hard-kill one node, and a way to reach you during it

Who it's for

A business with three servers of its own that wants one dying machine to stop being an outage, and wants to watch the failover happen before trusting anything to it.

Questions

How fast is delivery?

Standard delivery is within 43 weekday hours (≈ 3 working days). Express delivery — within 26 weekday hours (≈ 2 working days) — costs 25% more. The clock pauses while we wait on access or on the window you pick for the node-failure test, so scheduling that test to suit you never counts against the deadline.

Why exactly three nodes? Can we start with two and add more later?

Two nodes cannot hold a vote when one of them goes. The survivor has no way to tell whether its partner is dead or merely unreachable, so it refuses to take over, and the automatic recovery you are buying never happens. Three is the smallest number that works, which is why it is the unit here rather than a package size. Adding a fourth or fifth node to a working cluster is a far smaller job than the first build, so talk to us when you get there.

Do you build Ceph storage?

Not at this price. Ceph is a storage platform in its own right: how well it runs depends on your disk count, your disk type and how much network you can give it, and one built on thin networking degrades months later in ways that cost real money to unpick. This build attaches storage you already run, which means an NFS export, an iSCSI LUN, or each node keeping its own ZFS pool with scheduled replication between them. If Ceph is what you want, talk to us and we will scope it as its own build.

Do you put our virtual machines on it?

No, and that is deliberate. You get an empty platform that has been proven to fail over, and your own machines go on afterwards at per-machine prices, through The Per-VM Setup and The Per-LXC Setup. If you have an existing Proxmox environment to bring across instead, that is The Proxmox-to-Proxmox Migration.

Are backups part of this?

What you get here is a snapshot of the cluster configuration, which is a different thing from a backup of the machines that will run on it. Backups are The Proxmox Backup Server Setup: Proxmox Backup Server deployed, jobs and retention set, an off-site copy configured, and a restore performed in front of you before handover.

What do we need in place before you start?

Root access to all three nodes with Proxmox already installed on each, and shared storage already running that all three are permitted to reach. The three machines need the same Proxmox major version and the same CPU maker, since Intel and AMD cannot hand a running machine to each other, plus at least two network ports each and a low-latency link between them. We do not supply, specify, rack or cable hardware. All of those checks run before we start, and if something does not line up we tell you before you pay.

Often booked alongside