The honest motivation here is simple: I wanted to stop consuming cloud-native content passively and start practicing it on hardware I actually control. OpenShift and Kubernetes are everywhere in the industry, but there’s a difference between running docker on a laptop and wrestling with a real cluster on real iron. This homelab is my sandbox for that gap - a place where I can break things, rebuild them, and start executing Infrastructure as Code vs clickOps like I was doing in my previous labs with Proxmox and Vmware ESXI (RIP Homelab license)
What I built (so far)
The current setup is modest but purposeful:
- Nodes: A fleet of Lenovo m720q tiny pc, compact enough to live quietly in a corner and dense enough to be genuinely useful.
- Rack: A custom 3D-printed “lab rax” rack that holds the nodes in a tidy, serviceable stack. (More on the design in a future post I could derail all day on 3D printing)
- Networking: A PCIe dual-port 10Gig NIC paired with a sodola switch. Yes, 10G in a home lab feels like overkill and that’s kinda the point. Fast networking changes what’s possible, and I want headroom for when storage gets involved.
Right now the cluster is running OpenShift Native, which has been a fantastic way to get comfortable with the platform’s full lifecycle - install, operators, upgrades, the works.
Where this is going
This is only chapter one. The roadmap from here:
- Move the nodes to Proxmox. Running a hypervisor layer gives me flexibility to slice the hardware across multiple workloads instead of committing each node to a single OS. I also found out the hardway that running openshift on baremetal has it’s gotchas if you don’t have the supporting infrastructure like you would in a datacenter (more on this)
- Put that 10G NIC to work. The network upgrade was always meant to pay off later, with the proxmox abstraction I am planning to slice up the nvme drives on the tiny pc’s to boot+distributed storage.
- Ceph + NVMe. Distributed storage is the piece that turns a “cool cluster” into a real platform. Pairing Ceph with NVMe over 10G should give me performance I’d normally only see in a datacenter.
Each step builds on the last, and each one is an excuse to learn something new.
Stay tuned
This blog will track the whole journey - the wins, the outages, the 2am debugging sessions, and the moments where it all finally clicks. If you’re thinking about building your own lab, or just curious about what happens when someone tries to run a datacenter in their living room, stick around.