project-management-only/home-tv-station#14: Publish only mode

The advantage of doing it this way, rather than allowing the pod to run several publishers, is that the load can be spread across nodes rather than having a single pod demand a large number of resources

Introducing a new env var: CONTAINER_MODE

Values:

• combined (default)
• production: only run the publisher process
• server: only run Nginx

With this change, then it's possible to stand up a pod to act as a streaming server and then several pods to stream different channels in:

Pod 1 env vars

CONTAINER_MODE="server"
RTMP_APPLICATION="mytv"

Assuming that TCP port 1935 on pod 1 is fronted by a service called rtmp-service, we'd then create publishers using something like:

Pod 2 env vars

CONTAINER_MODE="publisher"
RTMP_SERVER="rtmp-service"
RTMP_APPLICATION="mytv"
RTMP_STREAMNAME="one"

Pod 3 env vars

CONTAINER_MODE="publisher"
RTMP_SERVER="rtmp-service"
RTMP_APPLICATION="mytv"
RTMP_STREAMNAME="two"

This would publish two streams

• rtmp://[server]/mytv/one
• rtmp://[server]/mytv/two

Both would also be available using HLS. However, it wouldn't be possible to:

• Use the next api
• Use dynamic publishing start/stop

Things like broadcast windows and the test card could still be used - the timings just need to be set in each pod.