docker-compose-cookbooks/jellyfin/README.md

Jellyfin

Overview

Jellyfin is a free, open-source media server. It organizes your movies, shows, music and books and streams them to web, mobile and TV clients. In this setup Sonarr and Radarr fill the library; Jellyfin just reads it.

Project Details

• Project Repository: https://jellyfin.org/ (Docs)
• Container Image: jellyfin/jellyfin
• Reverse Proxy Port: 8096

Getting Started

Prepare the shared media folder (once per host)

All media apps run as user 1000 and share /data/media. If Docker creates the folder itself it's owned by root, and the apps fail with "Folder '/media/' is not writable by user 'hotio'". Create it with the right owner before first start:

sudo mkdir -p /data/media/{downloads/prowlarr,movies,shows,music,books}
sudo chown -R 1000:1000 /data/media
sudo chmod -R u+rwX,g+rwX /data/media

Check with ls -ln /data/media — the owner should be 1000 1000.

Open http://localhost:8096 in your browser and complete the setup wizard:

1. Create your admin account

2. Add media libraries — point each one at the matching container folder:

   Library type	Folder
   Movies	/movies
   Shows	/shows
   Music	/music
   Books	/books

3. Finish the wizard. Jellyfin scans the folders and pulls metadata.

For Jellyseerr, create an API key under Dashboard → API Keys.

GPU transcoding (optional)

Deploy compose-gpu.yaml instead of compose.yaml. It uses the same env vars. The file defaults to NVIDIA.

1. Install the NVIDIA Container Toolkit on the host (once per host; skip it if another GPU stack such as Ollama already works):

   sudo pacman -S nvidia-container-toolkit        # Arch
   sudo nvidia-ctk runtime configure --runtime=docker
   sudo systemctl restart docker
   

2. Check the container can see the GPU: docker exec jellyfin nvidia-smi

3. Enable it in Jellyfin: go to Dashboard → Playback → Transcoding:

   • Hardware acceleration: NVENC
   • Hardware decoding: tick every codec your card supports. An RTX 30-series card decodes H.264, HEVC (incl. 10bit), VP9 (incl. 10bit) and AV1.
   • AV1 encoding: RTX 40-series and newer only. Leave it off on older cards.
   • Tone mapping: turn it on for HDR content. It runs on the GPU.

4. Confirm it's working: play something that needs transcoding. nvidia-smi on the host should list a jellyfin-ffmpeg process.

Intel/AMD: in compose-gpu.yaml, remove the deploy block and uncomment the /dev/dri:/dev/dri device. Then choose VAAPI (AMD) or QSV (Intel) as the hardware acceleration.

Environment Variable Notes

Variable	Description
JELLYFIN_IMAGE	Container image (default: jellyfin/jellyfin:latest)
JELLYFIN_RESTART	Restart policy (default: unless-stopped)
JELLYFIN_PORT	Published port for the web UI (default: 8096)
JELLYFIN_URL	Public URL clients use to reach the server (default: http://localhost:8096) — set this to your reverse proxy URL
JELLYFIN_MEDIA	Shared media root on the host (default: /data/media) — the same folder qBittorrent, Sonarr, Radarr and Bazarr use
VOL_PATH	Base path for persistent data volumes (default: /data)

Volume Notes

Volume	Purpose
${VOL_PATH:-/data}/jellyfin/config	Configuration, database and metadata
${VOL_PATH:-/data}/jellyfin/cache	Transcode and image cache
${JELLYFIN_MEDIA:-/data/media}/movies → /movies	Movies (Radarr root folder /media/movies)
${JELLYFIN_MEDIA:-/data/media}/shows → /shows	TV shows (Sonarr root folder /media/shows)
${JELLYFIN_MEDIA:-/data/media}/music → /music	Music
${JELLYFIN_MEDIA:-/data/media}/books → /books	Books

Jellyfin mounts only the library subfolders, not /data/media/downloads, so half-finished downloads never show up in the library.

Network Notes

Requires an external proxy network. Create it once with:

docker network create proxy

Jellyseerr reaches Jellyfin at http://jellyfin:8096 over the proxy network.

Docker Run

docker run -d \
  --name=jellyfin \
  -e JELLYFIN_PublishedServerUrl=http://localhost:8096 \
  -p 8096:8096 \
  -v ${VOL_PATH:-/data}/jellyfin/cache:/cache \
  -v ${VOL_PATH:-/data}/jellyfin/config:/config \
  -v ${JELLYFIN_MEDIA:-/data/media}/books:/books \
  -v ${JELLYFIN_MEDIA:-/data/media}/movies:/movies \
  -v ${JELLYFIN_MEDIA:-/data/media}/music:/music \
  -v ${JELLYFIN_MEDIA:-/data/media}/shows:/shows \
  --network proxy \
  --restart unless-stopped \
  jellyfin/jellyfin:latest

Additional Notes / Gotchas

• Only port 8096 is published. HTTPS (8920) is handled by Nginx Proxy Manager. Local network auto-discovery (UDP 1900/7359) won't work through the bridge network, so add the server to clients by URL.
• Photos are left out on purpose — use the Immich stack for photo libraries.
• Hardware transcoding is not in compose.yaml. See GPU transcoding under Getting Started.

Dockhand Stack, Deploy from Git

• Cookbooks Repository
• stackname: jellyfin
• Compose file path: jellyfin/compose.yaml (or jellyfin/compose-gpu.yaml for hardware transcoding)
• Additional env file (optional): jellyfin/sample.env

Load jellyfin/sample.env into the Environment variables in Dockhand, then create the Stack.