Tooling as a service via executable images

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?

• We use Docker for running an ephemeral container with dcos-cli contained. This means Docker is your only dependency no matter what you want to run.

DCOS_CONFIG=/dcos-cli/.dcos/dcos.toml
SOURCES=${SOURCES:-'[ "https://github.com/mesosphere/multiverse/archive/master.zip", "https://github.com/mesosphere/universe/archive/version-1.x.zip",]'}
MESOS_MASTER_URL=${MESOS_MASTER_URL:-'http://127.0.0.1:5050'}
MARATHON_URL=${MARATHON_URL:-'http://127.0.0.1:8080'}
TOKEN=${TOKEN:-'1234'}
EMAIL=${EMAIL:-'apollo@capgemini.com'}

• 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='http://172.31.1.11:5050' \
           -e MARATHON_URL='http://172.31.1.11:8080' \
           -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",]'
frameworks_list:
  - cassandra
  - chronos
  

- name: "install dcos-cli package"
  when: "frameworks_{{ item }}_enabled | bool"
  run_once: true
  docker:
    name: "{{ item }}"
    image: "{{ frameworks_dcos_cli_image }}"
    state: started
    command: "package install --options=/config --yes {{ item }}"
    volumes:
    - "/tmp/{{ item }}-config:/config"
    env:
      MESOS_MASTER_URL: "{{ frameworks_mesos_master_url }}"
      MARATHON_URL: "{{ frameworks_marathon_url }}"
      SOURCES: "{{ frameworks_sources }}"
  tags:
    - "{{ item }}"
  with_items:
    - "{{ 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