This repository build docker image for SRS.
You can checkout the docker image of wushaobo/srs-docker. The tag used below is 1.0 .
The version of SRS inside is v2.0-r2.
Still go ahead with most of 3rd party libs in the srs source, like Nginx(1.5.7), except
- Using the latest stable version of Ffmpeg rather than playing with the old version(2.1.1) inside srs source.
./configure --with-hls --with-nginx --with-transcode --with-ingest --with-stat --with-http-callback --with-http-server --with-http-api --log-trace \
--without-hds --without-ssl --without-dvr --without-ffmpeg --without-stream-caster --without-librtmp --without-research --without-utest --without-gperf --without-gmc --without-gmp --without-gcp --without-gprof --without-arm-ubuntu12 --without-mips-ubuntu12 \
--jobs=4 --x86-x64 --log-trace
There are many samples explained in SRS repository. I pick one of them to demo how we play it with docker.
In this sample, the SRS accepts RTMP stream and generates http flv stream as the output served by a built-in http server. The doc in SRS repository is here.
I suppose you would setup some other services integrating with srs in a compose file, like samples/compose.srs-http-flv.yml as follows.
version: '3'
services:
srs:
image: wushaobo/srs-docker:1.0
command: /opt/srs/objs/srs -c /opt/srs/conf/http.flv.live.conf
ports:
- "28080:8080"
- "21935:1935"
Run the command as follows.
docker-compose -f samples/compose.srs-http-flv.yml up -d srs
Then, your inbound RTMP stream could be pushed to the 1935 port of srs container (or 21935 port of the docker host), and the outbound http flv stream could be pulled from the 8080 port of srs container (or 28080 port of the docker host).
The SRS source includes many config files as samples, but you usually need to customize one config for your exclusive requirement.
The common idea is to pass your own config file in, but I insist on a graceful way. Using config file generation from environment variables, it could help to work better with configuration management.
In this sample, the SRS accepts RTMP stream and generates HLS as the output served by the nginx server. Additionally, the SRS is expected to send callback requests when any inbound stream is published or unpublished.
This image includes a config file generator based on a template engine. The template file is in jinja2 and compiled by python.
The callback address should be configured as hooks when running an SRS container. The template file config_generator/templates/srs_hls_hooks.conf.j2 has these lines
...
http_hooks {
enabled on;
on_publish {{ SRS_CALLBACK_URL }};
on_unpublish {{ SRS_CALLBACK_URL }};
}
...
You pass the environment variable SRS_CALLBACK_URL in when creating the container,
SRS_CALLBACK_URL=http://example.com/api/callback
the config file will be generated with the value before being run with.
...
http_hooks {
enabled on;
on_publish http://example.com/api/callback;
on_unpublish http://example.com/api/callback;
}
...
I suppose you would setup some other services integrating with srs in a compose file, like samples/compose.srs-hls-with-hooks.yml as follows.
version: '3'
services:
srs:
image: wushaobo/srs-docker:1.0
command: /tmp/srs/gen_conf_and_run.sh
ports:
- "21935:1935"
- "20080:80"
environment:
- config.env
And you will pass in a config.env file including environment variables when creating container,
SRS_CALLBACK_URL=http://example.com/api/callback
Run the command as follows,
docker-compose -f samples/compose.srs-hls-with-hooks.yml up -d srs
Then, your inbound RTMP stream could be pushed to the 1935 port of srs container (or 21935 port of the docker host), and the outbound HLS could be pulled from the 80 port of srs container (or 20080 port of the docker host).
And when the RTMP stream is successfully pushing/disconnecting to/from SRS, the callback request to SRS_CALLBACK_URL will be sent.