} Tooling as a service via executable images | Capgemini Engineering

This is a recap about how we use the isolation of containers and the power, portability and simplicity of Docker for providing tooling as a service via executable images using DCOS-CLI as an example. True story!

The problem:

  • We want to install Mesos Frameworks.
  • We want to deploy apps with Marathon.
  • We want to automate the installation of Mesos Frameworks.
  • We want to automate the deployment of apps.

The pain!

  • DCOS-CLI from Mesosphere is the best solution for the first two problems but we need to follow all the steps to install it.
  • We need to make sure our environment is suitable to install and run it so it has all the required dependencies.
  • We need to make sure every environment where we want to run it is suitable to install and run it so every environment has all the required dependencies.
  • We need to ensure all the dependencies and any executables comply with security requirements, as the attack vector can be quite extensive.

The solution:

  • To offer DCOS-CLI as a service that can be executed with zero dependencies with the specific environment.
  • To consume this service from a given automation tool like Ansible.

But how?

SOURCES=${SOURCES:-'[ "https://github.com/mesosphere/multiverse/archive/master.zip", "https://github.com/mesosphere/universe/archive/version-1.x.zip",]'}
  • The container creates a temporary config file from your environment variables, executes your command and dies. Check out the source code here.
docker run -v `pwd`/chronos_config:/config \
           -e MESOS_MASTER_URL='' \
           -e MARATHON_URL='' \
           -it capgemini:dcos-cli package install --options=config --yes chronos
  • Now it’s up to you how to automate it. We use ansible and the Frameworks role for Apollo which looks something like this:
# defaults file for frameworks
frameworks_dcos_cli_image: capgemini/dcos-cli
frameworks_zk_master_peers: "zk://{{ zookeeper_peers_nodes }}/mesos"
frameworks_mesos_master_url: "http://{{ ansible_ssh_host }}:5050"
frameworks_marathon_url: "http://{{ ansible_ssh_host }}:8080"
frameworks_sources: '["https://github.com/Capgemini/universe/archive/version-1.x.zip",]'
  - cassandra
  - chronos
- name: "install dcos-cli package"
  when: "frameworks_{{ item }}_enabled | bool"
  run_once: true
    name: "{{ item }}"
    image: "{{ frameworks_dcos_cli_image }}"
    state: started
    command: "package install --options=/config --yes {{ item }}"
    - "/tmp/{{ item }}-config:/config"
      MESOS_MASTER_URL: "{{ frameworks_mesos_master_url }}"
      MARATHON_URL: "{{ frameworks_marathon_url }}"
      SOURCES: "{{ frameworks_sources }}"
    - "{{ item }}"
    - "{{ frameworks_list }}"

The benefits:

  • Reduced impacting surface.
  • Only one dependency with the run time container, i.e Docker.
  • Crazily easy to integrate with other automation tools!
  • Reusable.
  • Build once, Run everywhere!

At the time of writing this post, Mesosphere, the creators of DCOS-CLI among such other amazing tools around Mesos, are working on their own image for DCOS-CLI

Join our team

If you like the sound of what you've read and would like to join our team, we're hiring!

Find out more about working with Capgemini

Posts by Alberto