Homelab / Personal computing
A personal infrastructure.
I started finding media online as a teenager, first through IRC and then through BitTorrent sometime in the mid-2000s. There was a fair amount of waiting and sorting involved. By the time streaming took off, I was already used to keeping a collection of files and deciding how to organize it.
Streaming took most of that work out of watching something. You could sit down, open an app, and start an episode without waiting for it to download or finding a player that could open the file. I kept using torrents, but I understood why opening an app became the default.
Pressing play is still easy once you’ve found the right app. The part before that can involve checking which service has a film this month, whether it’s still there when you come back, and what the subscription costs now. Amazon added ads to Prime Video in 2024. Netflix raised prices across most U.S. plans in January 2025, and a title can still leave the service when its license expires.
After years of keeping my own files, I still liked having a collection I could come back to. I’d also grown used to picking something on the TV and having the app remember where I’d stopped. I wanted that same ease with the films and shows I’d already kept. Jellyfin gave me a way to try that at home.
Once I had a player, there was more to try. I’m learning Spanish, so I set up Bazarr to look for Spanish subtitles for films and shows I already had. That gives me the option to read along in Spanish when I put something on, without having to look for subtitles each time.
Seerr gives friends and family a place to request films and shows without having to ask me to add them. I limit requests to 1080p to keep the library from filling up with larger 4K files. Drive space goes quickly enough as it is.
Sonarr and Radarr manage the library, and Prowlarr manages the indexers they search. They’re separate projects, but together they can take a request through to something people can watch in Jellyfin without my moving a file by hand. Fig. 01 shows how those pieces connect.
Media workflow
- RequestsSeerr
Sends requests to the library managers.
- AutomationSonarr / Radarr
Queue downloads and organize completed files.
- Host storageMedia library
Holds the files the library managers import.
- StreamingJellyfin
Reads the library and serves it to clients.
The downloader shares Gluetun’s network path.
Maintenance & recovery
- Service definitionsDocker Compose
The operator manages services from modular Compose files.
- Separate host storageApplication state
Running services keep settings and databases outside Git.
- Selected application dataEncrypted archive
The backup script checks extracted files and SQLite integrity.
Selected relationships in the public reference. Supporting services and external access are omitted.
Watching away from home
I wanted friends and family to be able to put on a film at their place without needing to know anything about the server at mine.
Plex and Emby have both made it possible to watch a personal library away from home for years. Plex now requires a paid pass for remote video on affected apps, which the viewer or the server owner can cover. Emby still offers full playback in a browser for free, with paid unlocks or Premiere for full playback in some apps. Plex playback requirements · Emby playback by app
Jellyfin has no paid playback tier. The connection is mine to set up and look after. I use Cloudflare Tunnel to make Jellyfin and Seerr available over HTTPS through my domain. Friends and family can request something in Seerr and watch what’s already in the library in Jellyfin. I manage the accounts and request limits. Fig. 02 shows how both apps are reached from outside the house.
Jellyfin’s library view, with titles hidden and artwork blurred.
- Friends and family
Use a browser or Jellyfin client
- HTTPS
- Cloudflare
Custom domain
- Encrypted Cloudflare Tunnel
Home server
cloudflared- Local HTTPJellyfin
Watch the library
- Seerr
Request films and shows
The downloader’s network
I wanted the downloader to use the VPN without routing the rest of the media stack through it. The Compose file makes that boundary easy to find when I come back to the setup later. A small systemd service also keeps qBittorrent in sync when Gluetun reconnects with a different forwarded port.
qbittorrent:
depends_on: [gluetun]
network_mode: service:gluetunSelected lines from the public Compose example.
Keeping it manageable
I wrote a small menu for the things I do most often. I type homelab in a terminal and pick what I need. I can check which containers are running, update one of them, see how much disk space is left, run a backup, or open LazyDocker. It saves me from having to remember a different command for each routine job.
homelab
Homelab Management
------------------
1) Check for updates
2) Update all media containers
3) Update one container
4) Show container status
5) Restart one container
6) Run encrypted backup
7) Sync configurations to local Git mirror
8) View recent update logs
9) Check disk space
P) Prune unused Docker images
L) Open LazyDocker
0) Exit
Select:homelab menu, transcribed from the terminal capture.Before an update, I like to know what I’d have to put back if it went wrong. The Compose files live in Git, and the apps keep their settings and databases separately on the host. I can check what’s running and back up the data I might need before changing a container.
The backup script makes an encrypted archive, then opens it again to verify the files, including the SQLite databases. I’d still test a restore separately before relying on it.
- Copy selected application data
- Package and encrypt with GnuPG
- Decrypt to temporary storage and check file hashes and SQLite integrity
I spent more time with Linux because there were things I wanted the server to do. Once it settled into regular use, I got less interested in adding services and more interested in being able to tell what had changed when something stopped working. Most of the time, of course, I’m just using it to watch something.
There’s a big difference between what I could put together on a home computer when I was finding files on IRC and what I can put together now. A machine I already had could run a media server and a handful of tools written by people working on different problems. There’s also a couple of decades’ worth of people writing down what they tried. A manual, a forum thread, or someone else’s Compose file often gave me enough to try something for myself.
That leaves me with choices I didn’t used to have. I can keep the player I like, add the subtitle tool because it’s useful to me, and decide for myself when a service needs an update. Those decisions can stay small, which is part of why I enjoy making them.
View the source on GitHub
A reference with example Compose files and scripts adapted from my setup, available under the MIT license.