Kubernetes Without the Overwhelm: k3s, Helm, Ingress, Storage and GitOps for a Homelab Cluster You Can Actually Run

Kubernetes Without the Overwhelm builds one small, honest Kubernetes cluster for a homelab and adds one layer at a time, so nothing arrives all at once. It starts with the question most guides skip: when Kubernetes is worth it at home and when Docker Compose is enough, and which distribution to use (k3s, Talos Linux, k0s, MicroK8s or kubeadm). Then it builds three Debian 13 VMs on Proxmox into a k3s cluster with embedded etcd, pinned to an exact release, configured from /etc/rancher/k3s/config.yaml and fronted by a kube-vip address that survives losing any one server. MetalLB replaces ServiceLB so services get real LAN addresses, Traefik routes hostnames, and cert-manager issues a wildcard certificate through a DNS-01 challenge, so internal names get trusted HTTPS without opening a port. Storage comes in three kinds, chosen per app: local-path for scratch, Longhorn replicas with backups to your NAS, and the NFS CSI driver for the shares you already have, with a clear line on what never belongs on NFS. Apps arrive as pinned Helm releases with probes and resource limits, secrets live in Git through Sealed Secrets or SOPS with age, and Flux reconciles the whole cluster from a Forgejo or GitHub repository while Renovate opens the version bumps as pull requests. Day 2 covers upgrades one server at a time with the system-upgrade-controller, etcd snapshots and a rehearsed restore, Velero app backups, kube-prometheus-stack and evictions, then a troubleshooting field guide and a rebuild drill. Every version, flag and default was checked against the official k3s, Kubernetes, MetalLB, Traefik, cert-manager, Longhorn, Helm, Flux and Velero documentation in October 2026. No prices or speed claims.

What you get

• Before you start: Kubernetes vs Compose at home, and k3s vs Talos Linux, k0s, MicroK8s and kubeadm in one table
• The parts that matter: control plane and nodes, Deployments, Services, Ingress, PVCs, namespaces and labels; kubectl get, describe, logs, diff and apply; kubeconfig contexts
• Building the cluster: three Debian 13 VMs from a cloud-init template, k3s v1.36.5+k3s1 pinned with INSTALL_K3S_VERSION, HA embedded etcd with cluster-init, the token, tls-san, kube-vip for a fixed API address, joining servers and agents, and a dead-node test
• Networking: flannel VXLAN, ClusterIP/NodePort/LoadBalancer, ServiceLB vs MetalLB layer 2 with an IP pool and pinned addresses, CoreDNS, LAN DNS on OPNsense, default-deny NetworkPolicy, a cluster VLAN
• Ingress and certificates: the Traefik bundled with k3s, Ingress vs IngressRoute vs Gateway API, HTTP to HTTPS redirect, cert-manager with a Cloudflare DNS-01 issuer, one wildcard served as Traefik's default, and the ingress-nginx retirement
• Storage: local-path and its limits, access modes and reclaim policies, Longhorn replicas, recurring snapshots and NFS backups, csi-driver-nfs to your NAS, StatefulSets, and why SQLite stays off NFS
• Real apps with Helm: a worked app with requests, limits and three probes, Helm 4 releases pinned with values files, helm diff, your own small chart, and Sealed Secrets or SOPS with age
• GitOps with Flux: a clusters/infrastructure/apps repository, flux bootstrap for Forgejo or GitHub, Kustomizations with dependsOn and SOPS decryption, OCI and Helm repositories, suspending and resuming, Renovate, and Argo CD compared
• Day 2: upgrade order and version skew, system-upgrade-controller Plans, etcd snapshots and --cluster-reset restores, Velero with file-system backups, kube-prometheus-stack on k3s, evictions and QoS, certificate rotation
• When things break and the reference build: Pending, ImagePullBackOff and CrashLoopBackOff, walking the request path, DNS from a debug pod, the k3s journal, the 'works on one node' checklist, the final config and a rebuild drill

82 pages, 24 diagrams plus a chapter-opener graphic for every chapter, a cluster checklist and a cheat sheet. PDF and EPUB. Homelab the Right Way, Book 12.

Not sure yet? Read the free sample first: the cover, the full table of contents and all of Chapter 2, "Building the Cluster" (why three servers, an etcd majority explained, three VMs from a cloud-init template, pinning the k3s version, the config file instead of flags, the first server with cluster-init, a fixed API address with kube-vip, joining the other servers and agents, a kubeconfig on your workstation, what k3s installs for you, and a test that the cluster survives a dead node), at https://payhip.com/b/XGIMw

Also in a bundle: Everything Bundle: every ebook in the store, $89 · Homelab Bundle, $59.