Architecture

Previously

• Parse HLS generated by nginx-rtmp, copy fragments
  • Relatively complex
• Had to run on CDN master

• "hls" ​mode

Now

• icedist mode
• Spawns ffmpeg
  • Connects to icedist
  • Dumps HLS
• Can run anywhere
  • Typically still runs on CDN master

1. Create configuration

• There's a README!
• Configs live in ~/relive/git/configs/
• Name file after your event acronym
  • Not mandatory, but I'll be sad if you don't
• Either
  • Take a recent config as a sample
  • Use gen_event_config.py to autogenerate from streaming WS

1. Create configuration

Note:

• Cronjobs check that directory
  • Ideally, only put the file there once you want stuff to happen
• Clean up old configs (move to old/) while you're at it
  • Won't delete data, but reduces job churn

1. Create configuration

Note:

• Media conference ID is used to poll "released" status for talks
  • This is a number​

2. Validate cronjobs

• schedule.xml and (optionally) media-events should pop up in data/project-name/ after a bit
• An empty array should appear in /video/relive/project-name/index.json

• Running jobs manually to debug is safe, see crontab -l -u relive
  • Also works if you're impatient

3. Launch scheduler

./launcher.sh project-name

• Should list room mappings and next events
• Check for plausibility

4. Set up steaming WS

• Just pointing RELIVE_JSON to the appropriate place should be enough
• Website will force-refresh indices on rollout

\o/