Why Build Lab Stack?
This site exists because I believe in a different approach to infrastructure. After 13 years working in cloud and DevOps, I’ve seen both the power and the limitations of “cloud by default” thinking. Lab Stack is my answer to a few questions I kept asking myself:
1. How do I keep track of what’s running where?
Every homelab eventually reaches a point where you forget what’s running where. Lab Stack solves this by pulling live data from my Proxmox clusters, TrueNAS, and Docker hosts to generate the network topology and service listings you see on the homepage. The infrastructure documents itself.
2. Does self-hosting actually save money?
If you don’t need 100% uptime guarantees, self-hosting makes economic sense. My entire homelab—243 vCPUs, 520GB RAM, 47TB storage—would cost approximately $15,000/month on AWS or GCP. I built it for a fraction of that by sourcing enterprise hardware from eBay.
The math is simple: even accounting for power, cooling, and occasional hardware replacements, I’ll save over $500,000 compared to cloud costs over 3 years.
3. Can AI actually help manage infrastructure?
This site is also an experiment in practical AI integration. The deployment pipeline includes Claude agents that:
- Analyze infrastructure changes and generate documentation
- Help write and validate configuration files
- Assist with troubleshooting and optimization
It’s not AI for AI’s sake—it’s about making a single person capable of managing enterprise-grade infrastructure.
4. Can I self-host the entire publishing stack?
I wanted to prove out a complete self-hosted publishing stack:
- Hugo generates static HTML from Markdown
- Forgejo (self-hosted Git) handles version control and CI/CD
- Cloudflare Tunnels provide secure public access without exposing ports
No external dependencies except Cloudflare for edge caching. The entire pipeline runs on my hardware.
5. Where do you even get enterprise hardware on a budget?
Perhaps the most important lesson: you can build serious infrastructure on a budget. My workstation runs dual Xeon 6430 processors (128 cores total) with 256GB DDR5 RAM and multiple GPUs—all sourced from eBay for a fraction of retail cost.
Enterprise hardware gets replaced on 3-year cycles. What’s “end of life” for a Fortune 500 company is perfectly capable of running AI workloads, video encoding, and development environments for individual use.
The Build Journey
Building a dual-socket Xeon workstation from eBay parts isn’t for everyone, but the results speak for themselves.
