I recently moved all my stuff to a homelab cluster — here’s how that went and what I learned along the way.
Why Homelab?
After years of running services across various cloud providers, I decided it was time to bring everything home. The benefits were clear:
- Cost savings — no more monthly cloud bills for personal projects
- Learning — nothing teaches you Kubernetes like running it yourself
- Full control — my hardware, my rules, my data
The Hardware
I went with a modest but capable setup:
- 3x Intel NUCs (i5, 32GB RAM each)
- A managed switch with VLAN support
- A Firewalla Gold Pro for network security (yes, I unboxed it on YouTube)
Software Stack
# The core components
cluster:
orchestration: k3s
gitops: ArgoCD
ingress: Traefik
storage: Longhorn
monitoring: Prometheus + Grafana
secrets: SOPS + Age
Lessons Learned
-
Storage is the hardest part — distributed storage on bare metal is no joke. Longhorn saved me, but I had to learn the hard way that not all SSDs are created equal.
-
DNS matters more than you think — split-horizon DNS with Pi-hole took a few iterations to get right.
-
Backups are non‑negotiable — a firmware update bricked one of my NUCs. Velero turned that from a panic into a restore job.
What’s Next
I want to add GPU passthrough for some local LLM work and a bit of fun “overkill homelab” experimentation.
If you’re thinking about building your own homelab cluster, honestly: go for it. The learning experience alone is worth it.