Home Lab Kubernetes

GitOps-managed K3s cluster running on Talos-based nodes with Longhorn storage.

This site showcases a subset of my home lab infrastructure — production-ready manifests for common self-hosted applications.

ArgoCD CI/CD Pipeline

All deployments are managed via ArgoCD — GitOps-style continuous delivery.

Pipeline Flow

Developer → Git Commit → ArgoCD Detects → Sync → K3s Cluster

1. Commit Changes

# Edit manifests
vim Apps/pairdrop/base/deployment.yaml

# Commit & push
git add -A && git commit -m "Update PairDrop config"
git push origin main

2. ArgoCD Auto-Sync

• Polls repo every 3 minutes
• Detects changes in main branch
• Applies manifests: kubectl apply -k Apps/pairdrop/base/
• Rolls out deployment

3. Verify Deploy

# Watch rollout
kubectl rollout status deployment/pairdrop -n pairdrop-space

# Or via ArgoCD UI
https://argocd.k3s.wagmilabs.fun

Manual Deploy (Testing)

kubectl apply -k Apps/pairdrop/base/
kubectl rollout restart deployment/pairdrop -n pairdrop-space

Featured Apps

Storage & Infrastructure

• Longhorn — Cloud-native distributed block storage
• Windows VM — Windows 11 VM via KubeVirt

Applications

• PairDrop — Local file sharing (AirDrop alternative)
• Minecraft — Minecraft server with K8s management

Cluster Architecture

3-Node K3s Cluster (Talos Linux)

┌─────────────────┐         ┌─────────────────────────┐
│  Control Plane  │         │     Worker Nodes        │
│                 │         │                         │
│  ┌───────────┐  │         │  ┌───────┐  ┌───────┐   │
│  │   etcd    │  │         │  │Worker │  │Worker │   │
│  │ API Server│  │         │  │ Node 1│  │ Node 2│   │
│  │ Scheduler │  │         │  │       │  │       │   │
│  └───────────┘  │         │  │Longhorn│ │Longhorn│  │
│                 │         │  │Replica │ │Replica│   │
│                 │         │  │   1    │ │   2   │   │
│                 │         │  └───────┘  └───────┘   │
└─────────────────┘         └─────────────────────────┘
        │                              │
        └─────────── Workloads ────────┘
              (Pods scheduled on workers only)

Storage: Longhorn (2 Replicas on Workers)

• Control Plane → Runs etcd, API server, scheduler (no workloads)
• Worker Node 1 → Runs workloads + Longhorn Replica 1
• Worker Node 2 → Runs workloads + Longhorn Replica 2

Workloads are scheduled only on worker nodes. Control plane remains untouched for cluster stability.

If a worker fails, pods reschedule to the remaining worker. Longhorn maintains replica count automatically.

Tech Stack

Why This Architecture?

Design Decisions

1. GitOps Workflow — All changes via Git commits, auditable and versioned
2. Immutable Nodes — Talos Linux reduces attack surface, no SSH
3. Replicated Storage — Longhorn provides 3 copies across nodes
4. Declarative Config — Kustomize for DRY manifests

Trade-offs

Contact

Repo: github.com/mentholmike/k3s
Domain: k3s.wagmilabs.fun
LinkedIn: Michael Wyatt

This is a personal home lab — not for production use without review.