Flagship projectActive

Reproducible Kubernetes Homelab

A three-node Kubernetes platform on Proxmox engineered as a real operations environment. The goal is not simply to host applications—it is to practice provisioning, hardening, observability, recovery, and continuous reconciliation as one cohesive system.

03Proxmox nodes
IaCEnd-to-end rebuild
GitDesired state
HAFailure-tested design

System design

One delivery chain, clear ownership.

Each tool has a bounded responsibility. That separation keeps the rebuild path understandable and makes failures easier to localize.

  1. 01Provision

    Terraform

    Defines virtual machines, compute, storage, networking, and placement across the Proxmox cluster.

  2. 02Configure

    Ansible

    Builds a repeatable Fedora node baseline, installs the container runtime, and applies system hardening.

  3. 03Reconcile

    Flux

    Continuously reconciles controllers, configuration, and applications from version-controlled desired state.

  4. 04Observe

    Prometheus

    Exposes cluster health and workload behavior through metrics, dashboards, and actionable signals.

What the project demonstrates

Operational outcomes over tool collecting.

Repeatable rebuilds

The cluster can be destroyed and reconstructed without undocumented manual intervention.

Measurable behavior

Platform health is evaluated through telemetry and explicit verification—not visual assumptions.

Tested resilience

Node failures, recovery paths, and backup assumptions are tested as engineering requirements.

Durable knowledge

Procedures, architecture decisions, and incident lessons live alongside the implementation.

Application delivery

This website runs through the platform.

The site is statically built with Astro, packaged in a hardened unprivileged container, published to GHCR, pinned by digest, and reconciled into Kubernetes by Flux. Traefik and cert-manager complete the public delivery path.

AstroGHCRFluxKubernetes

Technology

Current platform stack.