Skip to content
Snippets Groups Projects
Select Git revision
  • master default protected
1 result

live

  • Clone with SSH
  • Clone with HTTPS
    • ale's avatar
    Implement tombstones for the autoscaler
    ale authored 
    Since VMs take a while to shut down due to their lame-duck period, we
want to stop considering them available as soon as we send the
shutdown command, so we maintain an internal list of "tombstones" used
to filter the list of available hosts.
    605bf40e
    History
    ale's avatar 605bf40e
    History
    Name Last commit Last update
    agent
    autoscaler
    director
    proto
    proxy/ansible
    vendor
    README.md
    go.mod
    go.sum
    README.md

    ai3/tools/live

    A toolkit to manage a RTMP/HLS-based video streaming service, with few streams and many listeners. The toolkit will autoscale a pool of reverse HTTP proxies, which will relay content from a single streaming server which takes care of the RTMP-to-HLS conversion.

    Auto-scaling

    Auto-scaling works by managing a pool of cloud instances via an API. Instances are scaled up and down based on the global utilization of the entire pool (compared to thresholds, with hystheresis).

    Instances report their utilization to a centralized service. This service:

    • serves DNS requests for specific instances
    • serves HTTP requests by redirecting clients to a specific instance based on the output of load-balancing algorithm