For the longest time, my "home lab" was less of a lab and more of a digital pile of laundry. I had multiple services running across all sorts of random hardware. If something went down, it was a scavenger hunt to figure out which device was hosting it, let alone why it had failed. I never really knew what was what, or where to look when things went wrong.
I decided it was time to stop treating my network like a science experiment gone wrong and actually build something reliable.
Phase 1: Cleaning Up the Network
The first step was stabilizing the backbone. You can't host services if your network is flaky. I picked up a Unifi Dream Router 7, along with a few Unifi managed switches and access points.
Suddenly, I had visibility. I could see clients, manage traffic, and actually trust the WiFi. Naturally, once the network was stable, the itch to host more things kicked in. I wanted a web server to test apps and pages I was building for work, but I didn't want to just clutter up my personal laptop.
Phase 2: Enter Proxmox (Thanks, Mathias)
A friend, Mathias, suggested I look into Proxmox. I started small with a simple mini-PC. Honestly? It just worked. Spinning up LXC containers and VMs was addictive. I named this node PVE1.
Phase 3: The GPU Hurdle & The Gaming Rig Resurrection
I tried to expand the capabilities of the mini-PC, but I hit a wall fast. I ran into some "fun" issues involving kernel panics, which turned out to be a conflict between specific Proxmox versions and the supported iGPUs.
I needed a different approach — something with more raw power. I looked over at my gaming PC, gathering dust in the corner. Plenty of RAM, decent storage, and a dedicated GPU.
I wiped it and reformatted it as PVE2. Now I had a cluster: a low-power mini-PC for 24/7 lightweight services, and a beastly gaming rig for the heavy lifting.
The Current Stack
With the hardware sorted, I've settled into a really comfortable software stack running across the cluster:
- Proxmox: PVE1 & PVE2 Hypervisor Layer
- Cockpit: Web-based system admin
- PiHole: Ad blocking & DNS
- NextCloud: Private cloud storage
- HomeAssistant: Smart home brain
- ESPHome: IoT device firmware
- Docker Test Server: Sandbox for work projects
- Ollama: Local AI inference
- Plex: Media Server
- Grafana: Data Visualization
- VictoriaMetrics & VictoriaLogs: The observability backend
Seeing What's Actually Happening
The real game-changer lately has been VictoriaLogs. I'm scraping logs from everything — the hypervisors, the containers, the apps — and importing them into VictoriaLogs.
It is making it genuinely fun to see what is happening on my network. Instead of guessing why a service is slow or why a connection dropped, I can just query the logs. It turns the invisible background noise of the network into actionable data.
What's Next?
The lab is never "done," but for the first time, I feel like I have a platform rather than a pile of parts. If you are on the fence about moving from bare metal or random Raspberry Pis to a hypervisor like Proxmox, take the leap. Just maybe keep a spare gaming PC handy, just in case.