No sound when running Chrome in Docker container on Ubuntu 14.04

[asemt]

Hi,

@jfrazelle:
first of all: Thank you very much for these Dockerfiles - they're fantastic! 👏

My problem is related to running the Chrome Docker container without having sound inside the container. I run the container with:

docker run -it \
            --rm \
            --net host \
            --cpuset-cpus 0 \
            --memory 512mb \
            -v /tmp/.X11-unix:/tmp/.X11-unix \
            --device /dev/snd \
            -e DISPLAY=unix$DISPLAY \
            --name chrome \
            my/jfrazelle_chrome:2015-11-28

I get following errors as soon as Chrome tries to access the sound device (browsing to youtube.com, for example):

ALSA lib confmisc.c:768:(parse_card) cannot find card '0'
            ALSA lib conf.c:4260:(_snd_config_evaluate) function snd_func_card_driver returned error: No such file or directory
            ALSA lib confmisc.c:392:(snd_func_concat) error evaluating strings
            ALSA lib conf.c:4260:(_snd_config_evaluate) function snd_func_concat returned error: No such file or directory
            ALSA lib confmisc.c:1251:(snd_func_refer) error evaluating name
            ALSA lib conf.c:4260:(_snd_config_evaluate) function snd_func_refer returned error: No such file or directory
            ALSA lib conf.c:4739:(snd_config_expand) Evaluate error: No such file or directory
            ALSA lib pcm.c:2267:(snd_pcm_open_noupdate) Unknown PCM default
            ALSA lib confmisc.c:768:(parse_card) cannot find card '0'
            ALSA lib conf.c:4260:(_snd_config_evaluate) function snd_func_card_driver returned error: No such file or directory
            ALSA lib confmisc.c:392:(snd_func_concat) error evaluating strings
            ALSA lib conf.c:4260:(_snd_config_evaluate) function snd_func_concat returned error: No such file or directory
            ALSA lib confmisc.c:1251:(snd_func_refer) error evaluating name
            ALSA lib conf.c:4260:(_snd_config_evaluate) function snd_func_refer returned error: No such file or directory
            ALSA lib conf.c:4739:(snd_config_expand) Evaluate error: No such file or directory
            ALSA lib pcm.c:2267:(snd_pcm_open_noupdate) Unknown PCM default

I run Docker v1.9.1 build a34a1d5 on Ubuntu 14.04.3 LTS.

Some more information gathered inside the Chrome container:

root@godock:/# cat /proc/asound/cards 
 1 [PCH            ]: HDA-Intel - HDA Intel PCH
                      HDA Intel PCH at 0xf7418000 irq 65
root@godock:/# cat /proc/asound/modules 
 1 snd_hda_intel
root@godock:/# cat /proc/asound/devices 
  1:        : sequencer
  2: [ 1- 0]: digital audio playback
  3: [ 1- 0]: digital audio capture
  4: [ 1- 0]: hardware dependent
  5: [ 1]   : control
 33:        : timer

Any help would be much appreciated.

Best,
Andreas

[jessfraz]

The problem is probably pulseaudio on your host interfering 

[jessfraz]

I would look at the other issues in this repo wrt to sound and see if any of them fix it

[pwFoo]

I get sound working with chromium executed as a "normal" user (uid 1000).

1. Maybe you need dbus and shm inside the container

-v /run/dbus/:/run/dbus/ -v /dev/shm:/dev/shm

2. The user have to be member of the audio group inside the container (isn't needed outside...)

3. audio group gid have to be the same inside and outside the container (something like "groupmod --gid $AUDIO_GID audio" inside...)

Maybe not all is necessary, but I need all together to get it work...

docker run --rm --name chromium --device /dev/snd -e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix -v /run/dbus/:/run/dbus/:rw -v /dev/shm:/dev/shm <myCustomChromiumImage>

But success is also docker host dependent (arch missing USER_NS, alpine fails because of security features like grsec I think,...). All works fine with Voidlinux for me...

[jessfraz]

Or try --group-add audio on run

[pwFoo]

Will it map the hosts group or just add the user to the group existing inside the container?
If host (Voidlinux) and container (Debian) are different also the gid could be different.

void

audio:x:12:pulse

debian

audio:x:29:

[lidel]

Bit late to the party, but if anyone still wants to get it working on desktop with PulseAudio:

1. Add load-module module-native-protocol-tcp to /etc/pulse/default.pa & restart PA

2. Make sure PulseAudio cookie is present in X11 root

   # Set up PulseAudio Cookie if missing
   if [ x"$(pax11publish -d)" = x ]; then
       start-pulseaudio-x11
   fi

3. Pass PA address to the container as PULSE_SERVER (tcp:localhost:4713 is a default for module-native-protocol-tcp):

    -e PULSE_SERVER=tcp:localhost:4713

4. Extract & pass PA cookie:

   -e PULSE_COOKIE_DATA=`pax11publish -d | grep --color=never -Po '(?<=^Cookie: ).*'` 

Personally, I use a wrapper script that takes care of 2-4.

[mosajjal]

my problem with sound in Chrome is, when I play a sound (e.g. music) on my host, Chrome's sound is off and vice versa. I couldn't make them both have sound at the same time.

[ssebastianj]

@lidel: Nice, that solution gave my containers the power of sound 😄

Tested on:

• Ubuntu 16.04.1 LTS x86_64
• Docker v1.11.2
• PulseAudio v8.0
• Build Code: bvb5hx648ezythwbrehtzia

[briceburg]

jess/chrome is built on debian:sid, and the gid mappings may not line up with the host (e.g. on archlinux the audio group is 91, on debian it's 29). @pwFoo brought this up above.

I use a wrapper script with something like:

for group in audio video; do
  gid=$(getent group $group | cut -d: -f3)
  [ -z "$gid" ] || __docker_flags+=" --group-add=$gid"
done

Visiting http://hpr.dogphilosophy.net/test/ should help test. Beware around 2:35 in the first audio clip, things get CRAZY!

I'm attempting to use this technique in dex -- in the v1 runtime here -- which is working well for the chrome image based on jess's work.

[atilaromero]

to keep it short:
--group-add $(getent group audio | cut -d: -f3)

[denysvitali]

@atilaromero solution worked for me!

[fsmunoz]

Just to add something since this is one of the top results and the most informative: after some testing and multiple approaches the following should work for those in a RHEL7 host (or similar) running Chrome or similar:

## Sound options for RHEL7 OpenClient
snd_opts="--device /dev/snd \
	-e PULSE_SERVER=unix:${XDG_RUNTIME_DIR}/pulse/native \
	-v ${XDG_RUNTIME_DIR}/pulse/native:${XDG_RUNTIME_DIR}/pulse/native \
	--group-add $(getent group audio | cut -d: -f3)"

Note that this implies running pulseaudio on the host (which is the default). It also uses the default module-native-protocol-unixmodule instead of the TCP one (avoiding having to enable it just for this).

This works for me for the ones I've tried, based on Ubuntu and Debian; I suppose that the paths inside the container should be updated to reflect each distribution, I remember that dbus had different default locations some time ago.

[guiambros]

Thanks @fsmunoz! Much appreciated.

This worked for me, host on ubuntu 14.04 using Pulse Audio.

[dadreggors]

@fsmunoz I got jess/spotify audo working on Fedora 26 by adding this to end of run command but you seem to be adding this to a script or env file to set a variable. Where would you add that variable definition and how would you use this variable to start docker containers needing audo support?

Sorry, new to docker

[fsmunoz]

@ddreggors take a look at https://gist.github.com/fsmunoz/d83dbcf6c60b651491e0815b12158d1c , this is how I do it. IIRC @jessfraz does something similar but at this stage I'm not sure exactly from whom I stole it, the idea is that I have dockerised applications as alias so I can launch them "normally". Some settings there could be over the top in terms of limiting access but my own usage is more about convenience.YMMV, and hope this helps.