Virtualisation & compute

1. What I walked into

Twenty-two physical servers, the oldest nine years old, each running a single application. Most warranties had lapsed, every machine was backed up with a different tool, and in summer the machine-room cooling ran at its limit.

• no shared storage and no spare capacity
• the inventory was a spreadsheet nobody had updated in three years
• one failed motherboard could have stopped warehouse management for days

2. What I changed or built

A cluster of four identical hosts running Proxmox, with shared Ceph storage and a separate migration network. The machines moved one at a time, in weekend maintenance windows, always with a rollback plan ready.

• templates for the Windows and Linux servers
• nightly backups and a weekly restore test
• a capacity plan that survives losing a whole host

3. The numbers

The 22 machines became 36 virtual ones, because the big applications were split by role along the way. The machine room draws 58% less power, the cooling has not raised an alarm since, and the move caused not a single unplanned outage.

4. The tools, and why these

Proxmox brought clustering and live migration without licence fees, and built on what the team already knew about Linux. Ceph meant there is no single storage box that could stop everything on its own. The critical Windows services run on it just as they did on bare metal.

5. What it meant for the business

A hardware fault now costs minutes, not days: the machines move to another host and production never notices. A new server comes out of a template in half an hour — no purchase order, no waiting.