Production-grade infrastructure practices on bare metal. Terraform provisioning, Kubernetes, CIS L1 hardening, GitOps with FluxCD, and full observability stack — documented as I build it.
Homelab Architecture: 3-VM Kubernetes Cluster on KVM
Why not just use EKS? Managed Kubernetes is great for production. It’s terrible for learning what Kubernetes actually needs from the underlying OS. When kubespray says “ensure net.bridge.bridge-nf-call-iptables = 1”, it’s helpful. Understanding why — that kube-proxy depends on this to intercept traffic through the bridge — is the kind of knowledge you only get by running it yourself on bare metal. The goal: build a production-grade cluster at home, document every decision and tradeoff, then mirror the pattern to AWS with confidence. ...