9 nodes, 2 networks, 3 DNS layers, and a VPN mesh that ties it all together. Here's how my homelab actually connects.
After the fifth DNS outage in my homelab, I stopped guessing and built a checklist. Here's the framework I use every time.
Pi-hole v6, a Docker reboot, and a DNS chicken-and-egg problem that took down my entire mesh network for two hours.
I configured Headscale split DNS for my staging environment. Then my production domains stopped resolving. Here's the debugging story.
I replaced Tailscale's cloud control plane with Headscale on a 5 EUR VPS. Here's what I gained, what broke, and why I'd do it again.
My cluster went down. Not because of Kubernetes. Not because of a bug. Because the power flickered for 200 milliseconds.
Helm charts are somebody else's opinions wrapped in Go templates. I wanted my own opinions. Here's what happened.
You don't need Terraform for three mini PCs. I heard that a lot. After rebuilding my cluster four times, I stopped listening.
Traefik's docs make it look simple. It is not. A week of lessons from replacing K3s's built-in ingress controller.
How I turned a Raspberry Pi 4 into the DNS brain of my homelab -- and every way it broke along the way.
I could SSH into each VM and run the K3s install script. I could also eat soup with a fork. Here's how I automated a 3-node K3s cluster with Ansible.
I have 12GB of RAM per node. Full Kubernetes would eat half of it before I deploy a single pod. Here's why K3s is the only sane choice for a homelab.
Installing Proxmox on an Acemagic mini PC should be straightforward. The installer worked fine. Everything after that was where the problems started.
Everyone says 'just use an old PC.' Nobody tells you why that's bad advice for a K3s cluster that needs to run 24/7. Here's what actually matters — and what I bought.
Cloud abstractions are comfortable. That's exactly the problem. Why I'm spending real money on bare metal to build an internal developer platform from scratch.