Why screen folder in /var/run disappears in Docker container after running Ansible playbook?

Files

I've got the following Dockerfile:

FROM ansible/ansible:ubuntu1604py3
ENV DEBIAN_FRONTEND noninteractive
RUN pip3 install ansible
RUN printf '[local]\r\nlocalhost ansible_python_interpreter="env python3"\r\n' > /etc/ansible/hosts
ADD . /opt
RUN ansible-playbook /opt/install_screen.yml -c local -v
RUN useradd -d /home/ubuntu -ms /bin/bash -g root -G sudo -p ubuntu ubuntu
USER ubuntu
WORKDIR /home/ubuntu
CMD ["/usr/bin/env", "screen"]

and the following install_screen.yml playbook file:

- hosts: localhost
  tasks:
    - name: Install screen (a full-screen window manager)
      apt: name=screen state=present
      register: screen_installed
    - name: Configure screen
      shell: bash -x /etc/init.d/screen-cleanup start
      register: screen_configured
      when: screen_installed is success
    - shell: ls -la /var/run/

Build

Now, I build the Docker container in a shell by:

$ docker build -t screen .

and the Configure screen task shows the following output:


changed: [localhost] => {"changed": true, "cmd": "bash -x /etc/init.d/screen-cleanup start", "delta": "0:00:00.009613", "end": "2018-05-01 23:09:49.159514", "rc": 0, "start": "2018-05-01 23:09:49.149901", "stderr": "+ set -e\n+ test -f /usr/bin/screen\n+ SCREENDIR=/var/run/screen\n+ case \"$1\" in\n+ test -L /var/run/screen\n+ test -d /var/run/screen\n+ find /var/run/screen -type p -delete\n++ stat -c%a /usr/bin/screen\n+ BINARYPERM=2755\n+ '[' 2755 -ge 4000 ']'\n+ '[' 2755 -ge 2000 ']'\n+ chmod 0775 /var/run/screen\n+ exit 0", "stderr_lines": ["+ set -e", "+ test -f /usr/bin/screen", "+ SCREENDIR=/var/run/screen", "+ case \"$1\" in", "+ test -L /var/run/screen", "+ test -d /var/run/screen", "+ find /var/run/screen -type p -delete", "++ stat -c%a /usr/bin/screen", "+ BINARYPERM=2755", "+ '[' 2755 -ge 4000 ']'", "+ '[' 2755 -ge 2000 ']'", "+ chmod 0775 /var/run/screen", "+ exit 0"], "stdout": "", "stdout_lines": []}

then my ls -la /var/run/ shell command shows:

Rather than solving the immediate problem you're having, let me suggest an alternative solution to your actual problem:


I need screen, so I can use multiple interactive shells in one instance of Docker. If I run multiple Docker instances, each takes 1G of RAM, so 10+ instances will kill my RAM.

Run screen (or tmux, but screen if you insist) locally on your workstation. Start up your docker container, and in the windows where you want to do something on the container, use docker exec -it bash to get a new interactive shell on the container. This will be much simpler than running screen inside the container, and allows you to also have windows running commands on your workstation.

You may also want to look into using Vagrant with the Docker provider to simplify the process of doing development on a Docker container.

StackExchange DevOps Q#4041, answer score: 2