What's in the rack, and why

Here’s the whole thing at a glance, top to bottom, and what each part is for.

The network

Everything hangs off a custom pfSense router built on a small mini PC with a four-port network adapter. Running my own router (rather than the box the ISP hands you) means I control the firewall rules, the network segments, and exactly what can talk to what — which matters when you’re hosting something that touches money.

The compute

Three machines run Proxmox, clustered together and talking over 10Gb SFP+ networking. Clustering means a virtual machine can move between hosts, so I can take one machine down without taking my services down with it. Storage across the three is pooled with Ceph, so the data survives a disk — or a whole node — failing.

The Raspberry Pis

A separate 10-node Raspberry Pi cluster, powered over the network cable itself (PoE), runs Kubernetes via k3s. This is where I learn the orchestration layer that modern deployments sit on. A dedicated node handles bulk storage on RAIDZ2, which tolerates two drives failing at once.

Each of these gets its own post. Next up: the Proxmox and Ceph cluster.

(Draft — a photo of the actual rack would carry this whole page.)