Before diving into software and services, let's talk hardware. My homelab runs on three mini PCs, chosen for their balance of performance, power efficiency, and price.
The Firewall and Control Plane: Glovary
This Glovary appliance serves double duty. It has four Intel i226-V 2.5GbE NICs, making it perfect for running OPNsense as my firewall. On the same box, I run Proxmox to host a VM that acts as my Kubernetes control plane (intel-c-firewall).
Specs:
- CPU: Intel N100 (4 cores)
- RAM: 8 GB
- Storage: 128 GB NVMe
- Network: 4x 2.5GbE Intel i226-V
- Cost: 42,000 KSH
The N100 is surprisingly capable for this workload. It handles firewall duties and the Kubernetes API server without breaking a sweat.
The Primary Worker: AMD Mini PC
This AMD-based mini PC (amd-w-minipc) is the workhorse of the cluster. With 16 cores, it handles most of the application workloads.
Specs:
- CPU: AMD Ryzen (16 cores)
- RAM: 13 GB
- Storage: ~1 TB
- Network: 2.5GbE
- Cost: 40,000 KSH
The 2.5GbE port is essential for keeping up with storage traffic between nodes.
The Storage Worker: ACEMAGIC
The ACEMAGIC (intel-w-acemagic) has a neat built-in display showing power consumption and temperatures. It currently handles storage-heavy workloads.
Specs:
- CPU: Intel (4 cores)
- RAM: 16 GB
- Storage: ~1 TB
- Network: 2.5GbE
- Cost: 25,000 KSH
The higher RAM makes it good for caching and storage operations with Longhorn.
Why Mini PCs?
A few reasons:
Power efficiency. The entire cluster runs on about 50W idle. That's less than a single traditional server and means I can run 24/7 without worrying about the electricity bill.
Quiet operation. These sit in my kitchen. Silence matters.
Compact footprint. Three mini PCs take up less space than a single rack-mount server.
Affordable. The entire cluster cost around 174,000 KSH. Compare that to enterprise hardware.
Network Topology
All three nodes connect to a 5-port 2.5GbE unmanaged switch. The Glovary firewall handles routing between VLANs and provides internet access. For WiFi, I use 2 Google Nest WiFi Pro units in bridge mode, connected to the switch.
Internet → Glovary (OPNsense) → Managed Switch → Worker Nodes
↓
Proxmox VM (Control Plane)
Power Consumption
I measured with a smart plug:
- Idle: ~50W total
- Under load: ~80-100W
That's roughly 36 kWh per month at idle. At Kenya's electricity rate of 26 KSH per unit, that's about 936 KSH per month to run the entire cluster 24/7.
Total Cost
Here's the full breakdown:
| Component | Cost (KSH) |
|---|---|
| Glovary (Firewall + Control Plane) | 42,000 |
| AMD Mini PC (Primary Worker) | 40,000 |
| ACEMAGIC (Storage Worker) | 25,000 |
| 2x Samsung 980 NVMe (1TB each) | 26,000 |
| 5-port 2.5GbE unmanaged switch | 6,500 |
| 40Gbps patch ethernet cables | 2,210 |
| 2x 75ft Cat 8 ethernet cables | 5,200 |
| 2x Google Nest WiFi Pro (second hand) | 15,000 |
| Shipping (all items) | 12,000 |
| Hardware Total | 173,910 |
| Software (all open source) | 0 |
For about 174,000 KSH, I have a fully functional Kubernetes cluster with distributed storage, running 24/7. The entire software stack - Talos Linux, Kubernetes, Flux, Longhorn, Prometheus, and all the self-hosted apps - is open source and free.
Lesson learned: Shipping from overseas is cheaper than buying locally. The only downside is the wait - but the savings are worth it.
What I'd Change
Looking back, a few things I'd do differently:
More RAM on the control plane. 8 GB works but 16 GB would be more comfortable for running additional services.
Matching NICs. Having consistent 2.5GbE across all nodes simplifies networking.
NVMe everywhere. Some nodes still have SATA SSDs. NVMe makes a noticeable difference for etcd and database workloads.
Next Steps
I have an old Dell laptop with an Intel Core i7 (8th gen) that would be great for Jellyfin transcoding. Planning to add it to the cluster and repurpose the ACEMAGIC for something else.
Future Expansion
Each node has room for vertical scaling through additional storage:
- AMD Mini PC: Has an empty SSD slot - could add a 4TB SSD
- Dell Laptop: Has an empty SSD slot - could add a 4TB SSD
- Glovary Firewall: Can support an additional SSD - could add a 4TB SSD
That's up to 12TB of potential storage expansion without adding new nodes. This is vertical scaling - increasing capacity on existing hardware rather than horizontal scaling (adding more machines).
The challenge is cost. 2.5" SSDs at 4TB capacity are expensive. For now, the 2x Samsung 980 NVMe drives handle my storage needs.
The plan is to max out vertical scaling before considering a dedicated NAS. With 12TB of potential expansion across existing nodes, that should cover my needs for a while.
Check out My Home Lab for the full software stack and the GitOps repository running on this hardware.
Last updated on January 14th, 2026