Late 2025 Infrastructure Snapshot

2025/12/07

I wanted to spend some time writing about my lab infrastructure, as I think it’s a key part of my learning, development, and professional growth. More importantly, I like tinkering with stuff, and this is a fairly cheap and useful way to do exactly that.

At this point, in all its incarnations, my homelab is almost a decade old. Like a phoenix, it has burned to the ground more than once, only to be reborn. This site itself is on its 4th iteration, check the wayback machine.

this site

This site is hosted on a single node k8s (k3s) running on a proxmox VM. Traffic is sent to the VM directly via portforward where traefik picks it up directly. The container is just a caddy server serving hugo generated assets. traefik does TLS as well via a HTTP-01 challenge, which is way more convient than the DNS challenges I used to do.

this is probably the most simplistic k8s setup I could imagine, and it works really well for me. The rollouts are painless, the gitops flow works really well for me, and I honestly appreciate the k8s workflow.

future plans

general lab

VMs and LXCs

I use Proxmox for the main hypervisor. I’ve used other stuff, but I am deeply familiar with the main opensource components Proxmox is built on, so the learning curve is very low. I have used other stuff (xcp-ng, nutanix CE, microcloud) but for my lab use, it’s great. I have had some serious hiccups, but nothing that cannot be solved. I like ZFS and use it extensively on my NAS and as the root storage for Proxmox.

Nodes are old Optiplexes with a mix of hardware, one has an older Nvidia card for hardware decoding/encoding for Jellyfin, and all will soon have dedicated 10G SFP links. Storage is all flash, surplus SSDs from eBay.

routing

I used OPNsense for routing, it runs on it’s own baremetal Optiplex with a dedicated NIC. I use an older Brocade ICX-7250 for a primary switch, it is the most/cheapest 10G switch you can buy as of early 2025 IMO.

applications

jellyfin

devbox

k8s

homeassistant