Showcase
My homelab
This is my personal lab: a small collection of repurposed hardware, storage and networking equipment used to learn, host services and keep important data under my control. It is intentionally practical rather than flashy, with low power use, recoverability and simple maintenance as the priorities.
The lab at a glance
A logical map of the lab, not a rack photograph or a physical wiring plan. The everyday services share the home network; backup destinations are shown separately so they are not mistaken for machines that must all stay on.
Everyday use
PC / phone / tablet / TV
|
HOME NETWORK
|
+--------------+--------------+
| |
GEEKOM N100 MAIN TRUENAS
Always-on Docker HP desktop
| |
+-- Caddy: local HTTPS +-- Files
+-- CoreDNS: local names +-- Photos
+-- Hosted services +-- Live storage
+-- Monitoring
The N100 runs services; the main NAS holds the primary storage. They are peers on the network, not a chain where all NAS traffic passes through the N100.
Recovery paths
MAIN TRUENAS
|
+-- Local replication
| |
| v
| SECOND TRUENAS
| Fractal Node 804
| Usually off / suspended
|
+-- Encrypted cloud copy
|
v
OFF-SITE STORAGE
SELECTED FILES + CONFIG COPIES
|
+-- Encrypted Restic backup
|
v
RASPBERRY PI 4
Separate backup destination
The second NAS is a local recovery copy, not an off-site backup. The Pi is physically at a separate location, reachable over Tailscale, and has completed scheduled backups since relocation. Its repository integrity check and earlier local sample restores are useful but separate evidence; no full replacement-host application-and-database recovery has been verified. The Restic selection is not a full copy of every service, database or media library.
On-demand experiments
DELL MICRO PC
+-- Proxmox / virtual machines
+-- Powered off until needed
GL.iNet ROUTER
+-- Isolated / mobile lab work
+-- Separate from home-network edge
Design goal: keep daily services quiet and low-power, wake the experiment hardware only when needed, and keep recovery copies separate from the systems they protect.
Hardware
Fractal Node 804
The Node 804 is my TrueNAS Voyager. It is the second NAS and normally stays powered off or suspended, receiving local replication from Normandy for recovery and protected copies.
GEEKOM N100
The N100 is the quiet, efficient always-on Docker host. It runs the lightweight services that do not need a full server, including the home proxy, DNS and monitoring tools.
HP Enterprise desktop — TrueNAS Normandy
The HP Enterprise desktop is my main TrueNAS server, Normandy. It is housed in a two-bay enclosure and runs the live storage, photo library and primary file workloads.
Dell Micro PC — Proxmox
The Dell Micro PC runs Proxmox for virtual machines and experiments. It is currently powered off, so it is available when needed without adding to the always-on power draw.
GL.iNet router
The GL.iNet router is the network edge for isolated or mobile lab work. It lets me test network changes without exposing the rest of the home network.
Raspberry Pi 4 — additional backup destination
The Pi holds an encrypted Restic repository for selected important files and configuration snapshots. It is now running at a separate location, reachable over Tailscale MagicDNS from the backup host. Scheduled post-relocation backups have completed successfully; the latest documented run on 12 October 2026 reported 46 snapshots, and a repository integrity check found no errors. A picture and Docker-project files had previously been restored and verified from it locally. This confirms the remote backup path and selected-file recovery, not a full replacement-host recovery.
Read the implementation and restore evidence.
How it fits together
The main PC and lab machines connect through the home network. The N100 publishes friendly local HTTPS names through Caddy and provides DNS for the services. The NAS stores the important data, while the second NAS, removable storage and encrypted off-site copies provide recovery paths.
This is a living setup. Hardware and services change as I learn, but the design goal stays the same: useful services, clear failure boundaries and backups that can actually be restored.
Coming next
I will add photographs of each machine, power measurements and a service-by-service inventory as the lab documentation grows.