Homelab
Production-Grade Kubernetes Homelab with Talos Linux, GitOps, and Zero Trust Architecture.
Note: Video shows earlier iteration of the setup.
Overview
This project transforms a collection of recycled enterprise thin clients into a highly available, production-grade Kubernetes environment designed around the principles of immutable infrastructure and declarative configuration. At its core, the cluster runs on Talos Linux, a container-optimized operating system that eliminates configuration drift by removing the console, shell, and SSH in favor of a secure, API-driven management plane. The entire infrastructure lifecycle is managed as code: Terraform and Ansible handle the provisioning of Proxmox VMs and cluster resources, while ArgoCD enforces a strict GitOps workflow, continuously reconciling the cluster state against a central Git repository.
Networking is architected for a complete Zero Trust security model. Public ingress is routed exclusively through Cloudflare Tunnels, removing the need for open inbound ports, while private administrative access is secured via a NetBird WireGuard mesh overlay. Inside the cluster, Cilium leverages eBPF technology to replace standard kube-proxy functions, providing high-performance networking and L2 service announcements. State persistence is handled by Longhorn, which aggregates mixed NVMe and SATA media into a distributed block storage system, enabling the reliable hosting of a complex application stack that includes media automation, workflow orchestration, and a comprehensive Grafana/Mimir/Loki observability suite.
Highlights
- Engineered API-driven Talos Linux nodes with read-only filesystems and zero SSH.
- Automated cluster state reconciliation and atomic rollbacks via ArgoCD and Git.
- Deployed Cilium CNI with eBPF replacement for kube-proxy and L2 announcements.
- Secured public ingress via Cloudflare Tunnels and private access through NetBird VPN.
- Orchestrated container workloads with K8S while managing edge devices via Ansible.
- Aggregated NVMe and SATA tiers into a high-availability Longhorn block storage pool.
- Centralized metrics and logging using Grafana, Loki, Mimir, and Alloy.
- Deployed various other apps such as a media stack and workflow automation tools.