Pulseaudio server in Docker container

The full title should actually read Pulseaudio server in docker container on Raspberry PI 2, but this doesn't really matter.

In short here it is:

	docker run --rm -ti \
	--device=/dev/snd/controlC0 --device=/dev/snd/pcmC0D0p --device=/dev/sndpcmC0D1p --device=/dev/snd/timer \
	-p 4713:4713 \
	rpi-deb-pulse \
	/bin/bash -c "groupadd --gid $(ls -l /dev/snd/controlC0 | cut -d' ' -f4 | xargs getent group | cut -d: -f3) alsadev \
	&& gpasswd -a pulse alsadev \
	&& pulseaudio --system --high-priority --no-cpu-limit -v \
	-L 'module-alsa-sink device=hw:0,0' \
	-L 'module-native-protocol-tcp auth-ip-acl=10.2.0.0/24'"

view raw docker-pulse-server.sh hosted with ❤ by GitHub

Now everything one by one:

• docker run --rm -ti - run non-persistent docker and prepare it as interactive (shell) run
• --device=.... make local sound devices available in container. This feature is quite new in docker, so make sure you have recent version (mine is 1.8)
• -p 4713:4713 set up port forwarding between host and container to make pulse server available to clients 
• rpi-deb-pulse is my docker image built from armbuild/debian by running system upgrade and install pulseaudio
• /bin/bash -c "..." is command to run inside container, more on this below
• groupadd --gid $(...) alsadev create a group inside container named alsadev and having same id as group of /dev/snd/controlC0 device. We need this to allow pulse to access native alsa devices.
• $(ls -l /dev/snd/controlC0 | cut -d' ' -f4 | xargs getent group | cut -d: -f3) this subshell is executed on host, not in container and prints group id of /dev/snd/controlC0
• gpasswd -a pulse alsadev adds user pulse to our new group alsadev
• pulseaudio ... run pulseaudio
• -L 'module-alsa-sink device=hw:0,0' explicitly load alsa sink and specify which output to use
• -L 'module-native-protocol-tcp auth-ip-acl=10.2.0.0/24' load tcp protocol listener and (optionally) configure ip-acl to only accept connections from given subnet

How to then use it to play sound remotely (client):

1. pactl load-module module-tunnel-sink server=10.2.0.1
2. in pavucontrol on Playback tab select desired sink for a stream

Things to improve:

• list of devices may be replaced with another subshell to not specify each dev manually
• add systemd unit to start container automatically
• tune pulseaudio since sound over tcp tunnel is worse then playing locally

Why was this done :)

I have a Raspberry Pi 2 acting as a wireless router and i didn't wont to install pulse into host system. And i was just curious about whether this will work :)