MistServer Bootstrap Stack

Cloneable example that bundles MistServer, Prometheus, and Grafana with ready-to-use VOD assets and a live ingest helper. The default setup compiles MistServer from source with -DWITH_AV=true for hardware-accelerated transcoding support, provisions Prometheus scraping, and loads a starter Grafana dashboard.

What you get

• Custom MistServer image built from the official development branch with AV acceleration enabled.
• Pre-baked configuration at configs/mistserver.conf mounted read/write for easy updates through the Mist UI.
• Sample VOD assets & simulated livestream preconfigured in Mist.
• Prometheus configured to scrape MistServer metrics and Grafana auto-provisioned with a baseline dashboard and Prometheus datasource.
• Default --shm-size=1gb to avoid the 64 MB Docker shared-memory limit.

Requirements

• Docker Engine 24+ and Docker Compose Plugin 2.20+.
• At least ~4 GB free RAM (MistServer build plus Grafana/Prometheus).

Quick start

macOS / Windows (published ports)

docker compose up --build

Services become available at:

• MistController UI: http://localhost:4242 (default credentials admin/admin).
• MistPlayer: http://localhost:8080/{stream}.html
• RTMP ingest: rtmp://localhost:1935/live/{stream}
• Prometheus: http://localhost:9090
• Grafana (admin/admin): http://localhost:3000

Optional GPU passthrough (Linux only)

docker compose -f docker-compose.yml -f docker-compose.gpu.yml up --build

docker-compose.gpu.yml grants the container access to NVIDIA GPUs (gpus: all) when the NVIDIA Container Toolkit is installed, and exposes /dev/dri for Intel Quick Sync / VA-API. If your distribution restricts /dev/dri to the render group, edit the commented group_add line with your render group GID (getent group render). Docker Desktop on macOS/Windows does not expose host GPUs to Linux containers, so this override has no effect there.

Working with MistServer

• Default streams:
  • vod: on-demand playback from assets in assets/vod/.
  • live: simulated livestream using ffmpeg
  • push: stream ready to accept RTMP inputs
• Metrics are exposed at http://localhost:4242/telemetry. Prometheus scrapes this endpoint every 10 seconds.

Simulating live sources

• To emit a synthetic FFmpeg test signal to the stream named 'push':

  ./scripts/videogen.sh -f flv -ac 2 rtmp://localhost:1935/live/push

• To playback the result (substitute your favourite player):

  ffplay http://localhost:8080/hls/push/index.m3u8
  ffplay http://localhost:8080/cmaf/push/index.m3u8
  ffplay http://localhost:8080/webrtc/push

• On the MistController UI (http://localhost:4242) you can view the stream and some basic QoE metrics
• The embedded MistPlayer allows you to seamlessly switch between protocols, like WebRTC: http://localhost:8080/push.html?dev=1

Grafana provisioning

Grafana loads the MistServer Overview dashboard automatically. Start with the example panels (target status, scrape duration, latest metrics) and extend the queries as you explore the Mist metrics namespace. To import community dashboards, drop the JSON in grafana/dashboards and restart Grafana.

Cleanup

docker compose down

Add --volumes to drop Prometheus/Grafana data directories.