Home Operations

Infrastructure as Code for my home Kubernetes clusters

---

Overview

This repository contains the Infrastructure as Code (IaC) for my home Kubernetes clusters, managed using GitOps principles with Flux. The infrastructure runs on Talos Linux, a modern OS designed specifically for Kubernetes.

---

Kubernetes

Clusters

I run two Kubernetes clusters:

• Chongus (Primary) - Dell R730 servers with NVIDIA GPUs
• Bitty (Secondary) - Intel NUC cluster

Core Components

• cilium - eBPF-based CNI with native routing and Gateway API support
• cert-manager - Automated TLS certificate management
• external-dns - Automatic DNS management via Cloudflare
• external-secrets - Kubernetes External Secrets Operator integrated with Doppler
• envoy-gateway - Gateway API implementation for HTTP routing
• rook-ceph - Distributed storage with Ceph
• volsync - PVC backup and restore to B2 and MinIO
• cloudnative-pg - PostgreSQL operator with HA and backup

GitOps

Flux watches the cluster-apps/ directory and reconciles the cluster state automatically. The workflow:

graph LR
    A[Git Push] --> B[Flux Detects Change]
    B --> C[Reconcile Kustomization]
    C --> D[Deploy HelmRelease]
    D --> E[Update Cluster State]

Loading

Directory Structure

This repository follows a structured GitOps layout:

📁 cluster-apps/           # Application definitions (Flux source)
├── 📁 base/              # Shared applications across clusters
├── 📁 chongus/           # Chongus cluster applications (NEW PATTERN)
│   └── 📁 [namespace]/
│       └── 📁 [app]/
│           ├── 📁 app/   # HelmRelease + configs
│           └── ks.yaml   # Flux Kustomization
├── 📁 bitty/             # Bitty cluster (deprecated pattern)
└── 📁 components/        # Reusable Kustomize components

📁 clusters/              # Cluster bootstrap configurations
├── 📁 chongus/
│   ├── 📁 bootstrap/     # Helmfile-based bootstrap
│   ├── 📁 flux/          # Flux Kustomizations
│   └── 📁 talos/         # Talos configuration
└── 📁 bitty/

📁 .taskfiles/            # Operational automation

Workflow

Network Architecture

Container Networking:

• Cilium CNI with eBPF
• Native routing mode (10.244.0.0/16)
• KubeProxy replacement enabled
• Hubble for observability

Gateway API:

• envoy-external (172.22.12.2) - Internet-accessible via Cloudflare Tunnel
• envoy-internal (172.22.12.1) - Local network only (Tailscale)

Load Balancing:

• Cilium LBIPAM (172.22.12.0/24)
• Maglev algorithm with DSR mode

DNS:

• External-DNS with Cloudflare provider
• Automatic record creation from Gateway API HTTPRoutes

Certificates:

• Let's Encrypt via cert-manager
• Automatic TLS for all HTTPRoutes

---

☁️ Cloud Dependencies

While most infrastructure runs on-premises, some cloud services are used:

Service	Purpose	Cost
Cloudflare	DNS, Tunnel, CDN	~$0/month (free tier)
Doppler	Secret management	~$0/month (free tier)
Backblaze B2	Backup storage	~$5/month
GitHub	Git hosting, CI/CD	~$0/month (free tier)

---

🔧 Hardware

Chongus Cluster (Primary)

Device	CPU	RAM	Storage	Purpose
Dell R730 x3	Intel Xeon	256GB+	2x Samsung 870 EVO 2TB	Kubernetes nodes with NVIDIA GPUs

Storage:

• Rook-Ceph: 6x 2TB SSDs (2 per node)
• Storage Class: ceph-block (default)
• Replication: 3 replicas

Bitty Cluster (Secondary)

Device	CPU	RAM	Storage	Purpose
Intel NUC x3	Intel i5/i7	32GB+	NVMe	Kubernetes nodes with QuickSync

• Rook-Ceph: 3x 512GB SSDs (1 per node)
• Storage Class: ceph-block (default)
• Replication: 3 replicas

Supporting Infrastructure

Device	Purpose
True NAS	NFS storage for media and shared files
Raspberry Pi	Ansible-managed DNS, Tailscale, and mDNS repeater

---

Getting Started

Prerequisites

Tools are managed via mise:

# Install mise
curl https://mise.run | sh

# Install all tools
mise install

Bootstrap a New Cluster

# 1. Generate Talos configuration
task talos:generate-clusterconfig

# 2. Apply to nodes
task talos:apply-clusterconfig

# 3. Bootstrap cluster
task k8s-bootstrap:talos-cluster

# 4. Deploy core apps and CRDs
task k8s-bootstrap:apps

Flux will then automatically sync applications from cluster-apps/.

Common Operations

# Validate Flux resources locally
task flux:validate

# Force reconcile an application
flux reconcile helmrelease [app-name] -n [namespace]

# Check cluster status
kubectl get kustomization -A
kubectl get helmrelease -A

# View Ceph storage health
kubectl -n rook-ceph exec -it deploy/rook-ceph-tools -- ceph status

---

Documentation

For detailed information about the repository structure, patterns, and best practices, see:

• CLAUDE.md - Comprehensive architectural context and patterns
• Taskfile Reference - Task automation commands

---

Acknowledgments

This repository is inspired by the k8s-at-home community and draws patterns from:

• onedr0p/home-ops - Excellent reference implementation
• k8s-at-home - Amazing community and support
• kubesearch.dev - Discovery of Helm charts and deployment examples

Special thanks to the maintainers of all the open-source projects used in this cluster.