Building a Production-Grade Kubernetes Cluster at Home

After years of running Docker Compose for my homelab services, I decided it was time to level up. Here's my journey building a fully automated Kubernetes cluster on bare metal.

The Hardware Setup

I started with three Intel NUCs (10th gen i5, 32GB RAM each) as my control plane and worker nodes. Nothing fancy, but plenty of power for a homelab environment.

Network Architecture

• VLAN 10: Management network (192.168.10.0/24)
• VLAN 20: Kubernetes pod network (10.244.0.0/16)
• VLAN 30: Service network (10.96.0.0/12)

Installing Kubernetes

I went with kubeadm for installation because it gives me full control over the cluster configuration. Here's the basic setup:

# Initialize the control plane
sudo kubeadm init \
  --pod-network-cidr=10.244.0.0/16 \
  --service-cidr=10.96.0.0/12 \
  --control-plane-endpoint=k8s.homelab.local:6443

# Install Calico for networking
kubectl apply -f calico.yaml

GitOps with ArgoCD

Everything in my cluster is managed through GitOps using ArgoCD. I have separate repositories for:

• Infrastructure components (ingress, storage, monitoring)
• Application deployments
• Cluster configuration

This means I can destroy and rebuild my entire cluster from Git in under an hour.

Storage Solution

I'm using Longhorn for distributed block storage across the nodes. It provides:

• Automatic replication across nodes
• Snapshot and backup capabilities
• Easy volume expansion

Next Steps

In the next post, I'll cover setting up automated backups and disaster recovery for the Kubernetes cluster.