Twelve servers became three, with no maintenance window

The challenge

Twelve ageing physical servers, three of them out of warranty, running everything from the ERP to the shop-floor file share. Any patch meant a Sunday, and the last two Sundays had overrun.

What we did

We audited the real load — it fitted comfortably on three nodes — and built a Proxmox cluster with a Ceph pool at three replicas. Workloads were converted and migrated in four batches over three weekends, each verified before the next began.

The result

Patching now happens during working hours by moving workloads between nodes. The hardware refresh that had been budgeted for the following year was cancelled.