exec and idle broadcasters == bad for ffmpeg about nginx-rtmp-module HOT 2 CLOSED

If broadcast is still in publish mode (stream not closed) then it's ok to exec ffmpeg. Child process is killed when stream is closed. I don't think it's a good idea to schedule children until then using a byte counter. Ffmpeg complains are not that bad :) Child process is never restarted instantly but after respawn_timeout which keep your system from 100% CPU in any case. BTW I think you can tune ffmpeg network timeout.

from nginx-rtmp-module.

I understand and agree
The case I was facing used a proxy in front of nginx, like broadcaster > rtmp > proxy > rtmp > nginx-rtmp
Said proxy is not -yet- closing the connection to nginx-rtmp when the broadcaster stops publishing.
Therefore these ugly messages in the logs ;)

Note, to make things cleaner, I tend to think having just nginx-rtmp exec /bin/sh -c "touch /tmp/encode-my_stream.req" and a script running via crontab / daemontools checking for /tmp/encode-stream_name, and running the matching transcoding process ... might be a way to go (at a small scale at least)

from nginx-rtmp-module.