Planning Hardware for a Production-Like Kubernetes Homelab

Building a Kubernetes homelab that actually teaches you production patterns requires hardware that can handle multiple VMs, networking complexity, and storage I/O without falling over.

This post covers the hardware decisions I made before diving into Proxmox, Kubernetes, and networking — and why cheap mini PCs are often better than old enterprise servers for learning at home.

What You Actually Need for Kubernetes

Before buying anything, I set requirements based on what I wanted to learn, not just what could technically run:

CPU

Minimum: 4 cores (2 control plane + 2 workers)
Recommended: 6-8 cores for multiple worker nodes + headroom
Why: Modern CPUs handle virtualization efficiently; you don't need server-grade Xeons

Memory

Minimum: 16GB (barely usable for 3 VMs)
Recommended: 32GB+ for realistic workloads
Ideal: 64GB if running monitoring, databases, storage layers
Why: Kubernetes eats RAM. So does Proxmox. So does every pod you deploy.

Storage

Minimum: 512GB NVMe
Recommended: 1TB+ NVMe for OS + VMs + storage experiments
Bonus: Additional drives for Rook/Ceph or Longhorn testing
Why: ZFS benefits from fast storage, and you'll want space for snapshots

Networking

Minimum: 1Gbit NIC
Recommended: 2.5Gbit or dual NICs for VLAN segmentation
Why: Real networking forces you to learn ingress, load balancing, and firewall rules properly

Why I Chose a Mini PC Over Enterprise Hardware

When I started researching, the common advice was "Buy a used Dell R720 for $200!"

But after running the numbers, I realized:

Power consumption: Enterprise servers pull 150-300W idle. Mini PCs? 15-30W.
Noise: Rack servers sound like jet engines. Mini PCs are silent.
Modern CPUs: N100-class systems are great for entry-level labs, but I wanted more cores and memory headroom for multi-node clusters.
Cost over time: The electricity alone pays for a new mini PC within a year.

For learning Kubernetes, you don't need 40 cores. You need:

Enough RAM to run 3-5 VMs
Fast storage for responsive feedback loops
Low power draw so you can leave it running 24/7
Quiet operation so it can sit on your desk

My Hardware Choice

For this homelab, I wanted something that felt production-capable without the downsides of enterprise gear — noise, heat, power draw, and aging hardware.

Mini PC

I went with a Ryzen-based mini PC that hits a really solid sweet spot for specs vs cost.

Key specs:

CPU: AMD Ryzen 7 (8 cores / 16 threads)
Memory: 32GB RAM (user-upgradeable)
Storage (included): NVMe SSD
Expansion: Additional NVMe slot
Networking: 2.5Gb Ethernet
Form factor: Small, quiet, low-power

The main reason I chose this model was the balance it struck:

A modern Ryzen 7 CPU with plenty of cores for virtualization
32GB of RAM out of the box, which is realistically the minimum for a serious Kubernetes lab
Dual NVMe support, which is uncommon at this price point
Still one of the cheaper mini PCs with comparable hardware

At the time I bought it, anything else with similar CPU and memory was either:

More expensive
Locked down on upgrades
Or based on significantly weaker CPUs

This box gives me enough headroom to comfortably run:

Proxmox
Multiple Kubernetes control plane and worker VMs
Monitoring stacks (Prometheus, Grafana)
Storage layers without immediately running into resource pressure

Additional Storage Upgrade

In addition to the included NVMe drive, I added a second 1TB NVMe M.2 2280 SSD.

The specific brand isn’t critical here — any decent 1TB NVMe M.2 2280 drive will work — but the reason for adding it is important.

I added the second drive so I could:

Run ZFS mirror (RAID1) at the Proxmox layer
Tolerate a single-disk failure without taking the entire lab down
Make disk failures easy to recover from instead of catastrophic

With two NVMe drives mirrored:

The lab keeps running if a drive fails
The failed disk can be replaced
ZFS resilvers automatically without rebuilding everything

This is exactly the kind of real operational failure scenario that’s valuable to experience in a homelab:

Real failure
Real recovery
No data loss
Minimal downtime

It also makes snapshots, VM cloning, and storage experimentation far less stressful.

Why This Setup Works Well for Kubernetes

This hardware gives me:

Enough CPU cores to oversubscribe VMs realistically
Enough RAM to avoid artificial constraints
Fast, redundant storage for ZFS and Kubernetes workloads
Low enough power draw to leave running 24/7

Most importantly, it lets me practice production-like patterns:

Multiple nodes
Storage redundancy
Failure recovery
Resource contention

Without dealing with:

Enterprise noise levels
10-year-old hardware
Or massive power bills

Hardware Summary

Here’s the final setup I landed on:

Component	Choice	Reason
Mini PC	Ryzen 7-based mini PC	Best price/performance for cores + RAM
RAM	32GB	Minimum for comfortable Kubernetes learning
Storage	2× 1TB NVMe (ZFS mirror)	Redundancy + easy recovery
Networking	2.5Gb Ethernet	VLANs and higher throughput
Power	~20–30W idle	Cheap to run 24/7

Alternative Options (If You're Shopping)

Biggest thing is buy what you can afford and keep in mind trying to get something that will work long term.

If you're looking for similar hardware, here's what matters:

Budget Tier (~$200-300)

Intel N100 or N95 mini PCs
16GB RAM (expandable to 32GB preferred)
256GB NVMe
Good for: Learning Kubernetes basics
You’ll outgrow it when: You add monitoring + storage layers

Mid-Range (~$400-600)

Intel i5/i7 12th-gen or AMD Ryzen 5000-series
32GB+ RAM
512GB+ NVMe
2.5Gbit NIC or dual NICs
Good for: Full homelab with monitoring, storage layers, multiple projects

Enthusiast (~$800+)

Intel i7/i9 or Ryzen 7/9
64GB+ RAM
1TB+ NVMe + additional SSDs
10Gbit NIC or multiple 2.5Gbit ports
Good for: Running prod-like workloads, CI/CD, multiple clusters

Storage Considerations

This is where many guides get it wrong. You don't need a NAS or SAN to learn Kubernetes storage.

Here's what I recommend:

For Learning

Single NVMe: Totally fine. Use Proxmox ZFS for snapshots.
Local storage: Kubernetes can use local-path-provisioner or OpenEBS for simple cases.
What you'll learn: PVCs, storage classes, ephemeral storage

For Advanced Learning

Second drive: Add a SATA SSD or second NVMe
Storage layers: Practice Rook/Ceph, Longhorn, or MinIO
What you'll learn: Distributed storage, replication, real-world storage ops

Don't overbuild: You'll learn more from constrained resources than from infinite storage.

Networking Hardware

You'll also need networking gear that supports VLANs (covered in the next post about Proxmox and networking).

Minimum:

Managed switch with VLAN support (even a cheap TP-Link or Netgear)
Router that can handle inter-VLAN routing

Recommended:

UniFi switch (USW-Lite-8-PoE or similar)
Firewall appliance or pfSense/OPNsense box. I use UniFi Cloud Gateway myself.
Why: Proper VLAN segmentation teaches you production networking patterns

I covered my UniFi networking setup in detail in Why I Switched to UniFi for My Homelab, which explains the VLAN architecture and what it enables for Kubernetes.

Power and Physical Setup

Power Consumption Math

Mini PC: ~20W average × 24h × 365 days × $0.12/kWh = $21/year
Old server: ~200W average × 24h × 365 days × $0.12/kWh = $210/year

That $190 difference buys better hardware every single year.

Physical Placement

Desk setup: Most mini PCs are silent enough to sit next to your monitor
Rack setup: Some mini PCs have rack ears or 1U mounting options
Cooling: Ensure airflow; don't stack in enclosed spaces

What You Don't Need

Let me save you some money:

❌ ECC RAM — Helpful for prod, unnecessary for learning
❌ Redundant power supplies — This is a homelab, not a datacenter
❌ 10Gbit networking — 1Gbit is fine, 2.5Gbit is plenty
❌ Rack-mount servers — Unless noise/power aren't concerns
❌ Enterprise SSDs — Consumer NVMe is fast enough

Cost Breakdown

Here's what I spent (You don't have to get the same things):

Affiliate Disclosure: Some of the links below are Amazon affiliate links. If you purchase through these links, I may earn a small commission at no additional cost to you. This helps support the blog and allows me to continue creating content like this. I only recommend products I actually use in my homelab.

Core Hardware

Sometimes you can get this Mini PC on sale, I originally bought it for $375 during Black Friday. Prices are also going up I believe due to the RAM situation.

Item	Cost	Notes
Mini PC (base)	$529.99	GMKtec M5 Plus
RAM upgrade	N/A	32GB included
Additional storage	$136.99	BIWIN Black Opal 1TB
Hardware Subtotal	$596.98

Networking (Optional but Recommended)

Item	Cost	Notes
Network switch	$109.00	UniFi Lite 8 PoE
Router/Firewall	$199.00	UniFi Cloud Gateway
Networking Subtotal	$308.00	Enables VLANs and production-like networking

Grand Total: $904.98

The networking gear is optional if you already have a managed switch and router that supports VLANs. But for learning production Kubernetes patterns, proper network segmentation is worth the investment.

Compare this to:

Enterprise server: $200 (used) + $200/year power = $400 first year
Cloud equivalent: $50-100/month for similar resources = $600-1200/year

The mini PC pays for itself immediately and keeps paying. I did opt for a bit higher end mini PC but I didn't want to run out of resources.

Mistakes I Almost Made

Before settling on this hardware, I almost:

Bought a used R720 — Realized power costs and noise weren't worth it
Cheaped out on RAM — 16GB is technically usable but frustrating
Skipped the managed switch — VLANs are essential for realistic networking
Went with only 256GB storage — Snapshots and experiments eat space fast

If you're on the fence about any of these, spend the extra money upfront. You'll save time and frustration.

What's Next

Now that the hardware is sorted, the next post covers:

Installing Proxmox VE
Configuring ZFS storage
Setting up VLAN networking with UniFi
VM creation for Kubernetes nodes
Firewall gotchas that will waste hours if you miss them

See below for the Part 2 in this series on setting up the mini-pc with Proxmox.

Final Thoughts

You don't need enterprise hardware to learn Kubernetes properly. You need:

✅ Enough resources to run multiple VMs
✅ Fast storage for quick feedback loops
✅ Networking gear that forces you to learn production patterns
✅ Low power draw so you can leave it running

The hardware is the foundation. Get it right, and everything else becomes easier.

In the next post, we'll install Proxmox and configure networking that actually resembles what you'd see in a real datacenter — VLANs, firewall rules, and all.