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:

RoleCountHostnamesIPs
Control Plane3k8s-controller1–3192.168.1.211–213
Worker3k8s-worker1–3192.168.1.221–223
GPU Worker1k8s-worker4192.168.1.224

VIP (API server): 192.168.1.210:6443 Load Balancer pool: 192.168.1.220–230 (Cilium L2 announcements)

Key Specs Per VM

NodevCPUsRAMDiskSpecial
k8s-controller1–348 GiB50 GiBTalos built-in VIP
k8s-worker1–3414 GiB200 GiBLonghorn storage
k8s-worker4816 GiB250 GiBRTX 5070 Ti (PCIe passthrough), dedicated=ai taint

Totals: 32 vCPU, 82 GiB RAM, 1000 GiB disk. Sized in core/platform/talos.go .