Hardware
Physical hardware overview: cluster nodes, networking, and GPU.
This section covers the physical hardware running the homelab Kubernetes cluster.
Cluster Overview
The cluster runs on 7 virtual machines provisioned by Pulumi on a single Proxmox host:
| Role | Count | Hostnames | IPs |
|---|---|---|---|
| Control Plane | 3 | k8s-controller1–3 | 192.168.1.211–213 |
| Worker | 3 | k8s-worker1–3 | 192.168.1.221–223 |
| GPU Worker | 1 | k8s-worker4 | 192.168.1.224 |
VIP (API server): 192.168.1.210:6443
Load Balancer pool: 192.168.1.220–230 (Cilium L2 announcements)
Key Specs Per VM
| Node | vCPUs | RAM | Disk | Special |
|---|---|---|---|---|
| k8s-controller1–3 | 4 | 8 GiB | 50 GiB | Talos built-in VIP |
| k8s-worker1–3 | 4 | 14 GiB | 200 GiB | Longhorn storage |
| k8s-worker4 | 8 | 16 GiB | 250 GiB | RTX 5070 Ti (PCIe passthrough), dedicated=ai taint |
Totals: 32 vCPU, 82 GiB RAM, 1000 GiB disk. Sized in
core/platform/talos.go
.