ajwest3d/jellyfin-gpu-exporter repository overview

⁠Jellyfin GPU Exporter

A lightweight Prometheus exporter that monitors the NVIDIA GPU usage inside a running Jellyfin Docker container.

This exporter runs independently, calls nvidia-smi inside the Jellyfin container using docker exec, and exposes metrics on /metrics for Prometheus scraping.

⁠📦 Features

• GPU presence check
• Memory usage (used/total)
• Utilization percentage
• Temperature (°C)
• Power draw (watts)
• Number of active GPU processes
• Error counter (for failed scrapes)

⁠🚀 Getting Started

⁠1. Build the image (optional)

git clone https://github.com/youruser/jellyfin-gpu-exporter.git
cd jellyfin-gpu-exporter
docker build -t ajwest3d/jellyfin-gpu-exporter:latest .

⁠2. Run the exporter

docker run -d \
  --name jellyfin-gpu-exporter \
  -e TARGET_CONTAINER=jellyfin \
  -e EXPORTER_PORT=9109 \
  -v /var/run/docker.sock:/var/run/docker.sock \
  -p 9109:9109 \
  ajwest3d/jellyfin-gpu-exporter:latest

• TARGET_CONTAINER: Name of your Jellyfin container (default: jellyfin)
• EXPORTER_PORT: Port to expose metrics on (default: 9109)

⁠📈 Metrics exposed

⁠📡 Prometheus config

- job_name: 'jellyfin-gpu'
  static_configs:
    - targets: ['localhost:9109']

⁠🧪 Local test

Visit the metrics endpoint in a browser:

http://localhost:9109/metrics

You should see all metrics in Prometheus format.

⁠🛠 Future Ideas

• Multi-container support
• Multi-GPU support
• Label per GPU index
• Optional metrics filtering

⁠📄 License

MIT

Built with ❤️ by @ajwest3d⁠