Kubernetes Homelab
From installing K3s on a Debian server to fully operating the cluster: observability, GitOps, and disaster recovery.
Chapter 1 · September 17, 2026
From a Debian server to a Kubernetes platform at home
How a Debian server became a single-node K3s cluster with Traefik, Cloudflare Tunnel, storage, and firewall — without a big-bang migration.
Chapter 2 · September 17, 2026
Real observability in a small Kubernetes cluster
Metrics, logs, and traces on a single-node cluster: the Prometheus/Loki/Tempo stack, trace_id correlation, and the real cost of observability.
Chapter 3 · September 17, 2026
GitOps, secrets, and reproducible infrastructure
GitOps with Flux, secrets with SOPS + age, and reproducible infrastructure: how the repository became the owner of each piece of cluster state.
Chapter 4 · September 20, 2026
Backup does not mean Disaster Recovery
Having backups is not the same as having Disaster Recovery: what real restores revealed about the control plane, PersistentVolumes, OCI images, and offline recovery.
Chapter 5 · September 20, 2026
Measuring Disaster Recovery: reducing RTO by 29.9%
How measurable rehearsals and deterministic automation reduced the observed RTO of my K3s cluster from 1h07m09s to 47m03s.
Chapter 6 · September 20, 2026
The problem of consistent backups in Kubernetes
How independent control-plane and PV backups became recovery sets with exact identity and a measured 64-second consistency window.