To this day, Docker does not support mapping user/group ids from the
host to the container. Hence, all files created on the mounted volumes
are owned by git:git (1000:1000). On the host that uid/gid possibly
belongs to a different user/group.
With this change the container will adapt the uid/gid of git in
/etc/passwd and /etc/group to the environment variables USERMAP_UID and
USERMAP_GID if specified.
In addition all files in /home/git and /var/log/gitlab have to be
adapted to reflect those changes. This can be a time consuming process
which can be improved by committing the container after the first run.
Idea from: http://chapeau.freevariable.com/2014/08/docker-uid.html
In the forthcoming commit all daemons will be auto started using
supervisord and the execution of the init script will end at the launch
of supervisord. This does not allow room for the `gitlab:check` task to
be executed in such a way that its output.
User's will need to login to the container and manually start the
`gitlab:check` task if the wish to check the status.
When the database connection parameters are not specified, the
image will try to spin up a redis container if it is able to
communicate with the host docker.
This is only possible if the following two options are specified in the
docker run command:
`-v /var/run/docker.sock:/run/docker.sock`
`-v $(which docker):/bin/docker`
This is primarily added to get the quick start guide to work without
much of a hassel.
When the redis connection parameters are not specified, the image
will try to spin up a redis container if it is able to communicate with
the host docker.
This is only possible if the following two options are specified in the
docker run command:
`-v /var/run/docker.sock:/run/docker.sock`
`-v $(which docker):/bin/docker`
This is primarily added to get the quick start guide to work without
much of a hassel.
