Running motioneye docker as non-root user, can't access /var/run

[towerofpower256]

If I installed the motioneye docker and ran as root, no issues there.

However I run into an issue when I use the motioneye docker image as another user that is not root (e.g. --user 1003:1003). I started getting errors that are visible in the log saying that motioneye was unable to create the .pid file in the /var/run directory. Motioneye isn't able to start either, and I can't get to the web interface.

The log:

CRITICAL: pid directory "/etc/motioneye/run" does not exist or is not writable
    INFO: hello! this is motionEye server 0.42
CRITICAL: pid directory "/etc/motioneye/run" does not exist or is not writable
    INFO: hello! this is motionEye server 0.42
CRITICAL: pid directory "/etc/motioneye/run" does not exist or is not writable
    INFO: hello! this is motionEye server 0.42

Here's my docker run command, including --user and RUN_UID and RUN_GID arguments:

docker run --name="motioneye" \
    -p 8765:8765 \
    --hostname="motioneye" \
    -v /etc/localtime:/etc/localtime:ro \
    -v /srv/motioneye/etc:/etc/motioneye \
    -v /srv/motioneye/lib:/var/lib/motioneye \
    --restart="always" \
    --detach=true \
    --user 1003:1006 \
    -e "RUN_UID=1003" \
    -e "RUN_GID=1006" \
    ccrisan/motioneye:master-amd64

I was able to get around this by updating motioneye.conf to store the pid file in another directory that motioneye should have access to. This fix appears to be working so far.

# path to the directory where pid files go (must be writable by motionEye)
#run_path /var/run
run_path /etc/motioneye/run # Fix for no access to /var/run within container if non-root

I suspect this is due to the Docker image that the motioneye docker image is utilizing, debian:buster-slim. https://github.com/debuerreotype/docker-debian-artifacts/blob/337f494fae12a1db13a003cea38e74f43d312ee6/buster/slim/rootfs.tar.xz

Doing some poking around:

• /run exists, with permissions of 755 (writable by root, read-only to everything else).
• /var/run exists appears to be a link to /run.

Console output:

I have no name!@motioneye:/run$ ls -l /var/ | grep run
lrwxrwxrwx 1 root root     9 Jul 19  2019 lock -> /run/lock
lrwxrwxrwx 1 root root     4 Jul 19  2019 run -> /run

I have no name!@motioneye:/run$ ls -l / | grep run
drwxr-xr-x   1 root root 4096 Jun 22 10:35 run

I have no name!@motioneye:/run$ ls -l /run        
total 16
drwxrwxrwt 2 root root 4096 Jul 19  2019 lock
drwxr-xr-x 2 root root 4096 Jun 22 10:35 motion
drwxr-xr-x 2 root root 4096 Jul 19  2019 mount
drwxr-xr-x 2 root root 4096 Jul 24  2019 systemd
-rw-rw-r-- 1 root utmp    0 Jul 19  2019 utmp

In the motioneye docker file, I can see that it's trying to own the /var/run folder, but I believe it's only chowning the link to the /run directory, not the /run directory itself, which is causing the issue.

Does that mean the fix is just to update the docker file to include /run in the directories that it's chowning?

chown motion:motion /var/run /var/log /run ....

[towerofpower256]

Possibly related to: #2407 #2388

[towerofpower256]

This article here suggests that it may be better to create a directory within run to hold app-specific pid files. https://serverfault.com/questions/159334/what-permissions-are-needed-to-write-a-pid-file-in-var-run

E.g. a new directory /var/run/motioneye, grant full access for the motion user to that directory, and update the motioneye.conf to store the pid files in there, instead of just /var/run.

[tobschndr]

I am facing the same issues. Via Portainer the log is showing the following after reboot (as assumed with regard to the article):

INFO: hello! this is motionEye server 0.42.1 CRITICAL: pid directory "/var/run/motioneye" does not exist or is not writable

So I wrote a startup script as linked in the article: (thanks @towerofpower256 )

mkdir /var/run/motioneye chown motioneye:motioneye /var/run/motioneye

My questions:

• Will/can this task (adding the folders and permissions after reboots) be part of the docker container in future?
• How "dangerous" is it to run the container as root? (without adding the UID and GID to docker run and modifying the /var/run)

[PhatHub]

For those who haven't figured out the systemd-onic (sorry didn't know how to say "systemd" in the same way of "pythonic") way of doing it, I just found out... utilize tmpfiles.d

It took me a while because the first time I tried to google for the answer, it took me weeks, and I couldn't find anything besides "it'll be complex."

A few weeks later I found this answer within minutes, and it all clicked.

I did this on Ubuntu Server, so YMMV if you're on something else, like RedHat. tmpfiles.d is an installable package, so if it doesn't work, make sure it's installed but I doubt it'll be missing since basically everyone is running systemd.

1. Create a motioneye file sudo touch /usr/lib/tmpfiles.d/motioneye.conf
2. Edit the file to create a temp file sudo vim /usr/lib/tmpfiles.d/motioneye.com (or use your preferred editor instead of vim)
3. ...and add a single line, d /var/run/motioneye 0744 motion motion (I put tabs instead of spaces to kind of line up columns in the future)
4. Exit your editor, stop motioneye systemctl stop motioneye to force it to delete the previous pid directory.
5. Restart it and see if it works. systemctl start motioneye

Hopefully this helps.

[MichaIng]

@PhatHub It's about the Docker image here, which has no systemd or native tmpfiles support.

Best solution is to allow not having any PID file at all. It is nonsense in this Docker image but currently not possible. I'll see if I find time to implement this during the weekend. But otherwise using /etc/motioneye indeed is a good idea and shouldn't cause any issues.