The Architecture That Makes a Home Lab Feel

There's a threshold in home lab development that nobody talks about enough. You cross it when your setup stops being a collection of interesting services you've spun up and starts being actual infrastructure—the kind where you can patch hosts without taking anything down, where deployments happen through code commits instead of SSH sessions, where you have the same operational capabilities as production environments that cost six figures to build.

Brandon Lee, who runs VirtualizationHowto, recently walked through the exact stack that got him across that threshold. After two decades with VMware, he rebuilt his home lab on Proxmox VE, Ceph storage, Talos Linux for Kubernetes, and a GitOps workflow that would look familiar to any platform engineering team. What's interesting isn't just the tools—it's how they compose into something that operates like enterprise infrastructure while running on five mini PCs.

The VMware Refugee Problem

Lee's pivot from VMware is worth examining because it's happening across the community right now. "I have been a VMware guy for a long time going all the way back to circa 2004," he explains in the video. "I made the switch recently to Proxmox VE server in the home lab. And honestly, I've not regretted it at all."

This isn't just one person's preference shift. Broadcom's VMware licensing changes have created an entire generation of home lab operators looking for new hypervisor foundations. Proxmox has momentum right now—not because it's objectively superior to VMware for all use cases, but because the community support and trajectory feel sustainable in ways that commercial software increasingly doesn't.

The foundation of Lee's setup is a five-node Proxmox cluster running on Minisforum MS01 mini PCs with dual 10-gigabit networking. That's the hardware substrate, but the real infrastructure starts with Ceph.

Why Shared Storage Changes Everything

Ceph is where Lee's setup stops being a toy and starts being production-grade. He's running both Rados Block Device (RBD) for VM storage and CephFS for shared file storage across services. "Moving to Ceph was probably the single biggest leap forward in my home lab," he notes. "You can migrate workloads between nodes, you can perform maintenance, you can patch systems without taking anything down."

This is the operational capability that separates hobby infrastructure from the real thing. When you can do rolling updates across your cluster without service interruption, you're operating at a different level. The architecture supports the workflows that enterprise teams take for granted.

But shared storage is just the foundation. The question is what you build on top of it.

The Immutable Kubernetes Layer

Lee's Kubernetes setup runs on Talos Linux, which represents a different philosophy than traditional distributions. Talos is immutable and locked down by design—you can't SSH into nodes. Everything is API-driven. "It can feel a little bit different at first, a little bit limiting," Lee acknowledges, "but once you understand it and start learning how to interact with it, it feels way more secure and it feels more like the cattle in the cattle versus pets analogy."

He pairs Talos with Omni, which provides the management layer for lifecycle operations. This combination removes much of the manual work that typically comes with Kubernetes cluster management. The appeal isn't just technical—it's operational. When your infrastructure is immutable and API-driven, you can't accidentally break things through one-off SSH sessions. The constraints create reliability.

For ingress and routing, Lee uses Traefik, which handles service discovery, routing, and automatic TLS certificates through Let's Encrypt. The key advantage, as he describes it: "What I really like about it is how well it fits with infrastructure as code. You define your rules and Traefik basically handles the rest."

GitOps as the Control Plane

The operational model that ties this together is GitOps through Argo CD. Lee stores his Docker Compose configurations and Kubernetes manifests in GitLab, and Argo CD continuously reconciles the cluster state with what's defined in the repository. "Instead of manually deploying applications, Argo watches that git repository and makes sure that my Kubernetes cluster running in Talos matches the desired state," he explains. "If something drifts or it changes manually... Argo CD automatically brings that back into sync."

This is infrastructure as code taken seriously. Every deployment happens through a code commit. Manual changes get automatically reverted. It's the same operational model that production engineering teams use, just running on mini PCs in someone's basement.

The monitoring layer is Netdata, which Lee highlights for requiring almost no configuration. "You just simply install it and within a few seconds you're automatically getting metrics." For anyone who's spent hours configuring Prometheus and Grafana, this matters. Sometimes the best tool is the one that works immediately.

The Backup Reality

Lee runs both Veeam and Proxmox Backup Server. "I keep Veeam just because that's mostly what's in the enterprise," he notes. "Proxmox backup server is purpose-built for Proxmox... Having both gives me confidence that I can recover from just about anything."

The dual backup strategy reveals something about how serious home lab operators think about their infrastructure. These aren't just experiments anymore—they're running services and storing data they actually care about. The backup architecture reflects that reality.

For remote access, Lee no longer exposes anything directly to the internet. Instead, he uses Twingate as a zero-trust solution, supplemented by Apache Guacamole for web-based access and Kasm for ephemeral workspaces. It's a defense-in-depth approach that treats home infrastructure like it matters—because increasingly, it does.

The Out-of-Band Problem

One detail that stands out: Lee specifically called out Intel vPro for out-of-band management as transformative. When he switched from traditional servers with IPMI to mini PCs, he lost that capability temporarily. Getting it back with vPro "changed everything," he says, particularly when working on the lab remotely.

This is the kind of operational concern that only makes sense once you're treating your home lab like production infrastructure. If you're just running some Docker containers, you don't care about lights-out management. If you're running a Kubernetes cluster with production-like workflows and you're 100 miles from home, you absolutely do.

What This Setup Actually Represents

Lee's stack isn't unique—many of these tools are popular in the home lab community. What's notable is how they compose into something that operates like enterprise infrastructure. The architecture supports operational patterns that used to require commercial software and significant hardware investment.

Proxmox and Ceph provide the infrastructure layer. Talos and Argo CD handle workload deployment and lifecycle. Traefik manages ingress and TLS. GitLab stores the source of truth. Netdata provides visibility. Dual backup systems protect data. Zero-trust networking secures access. vPro enables remote management.

Individually, these are just tools. Together, they create infrastructure that supports enterprise-grade operational patterns. "Your home lab stops feeling like a very amateur home lab and it feels more like what you would find in enterprise datacenter," as Lee describes it.

The interesting question isn't whether this is better than commercial alternatives—that's context-dependent. The interesting question is what it means when this level of infrastructure capability is accessible to anyone with a few mini PCs and some weekend time. The gap between hobby infrastructure and production infrastructure keeps shrinking. The tools are available. The knowledge is documented. The only question is whether you want to run it.

Dev Kapoor covers open source software and developer communities for Buzzrag.