Installing and Managing DC/OS using Ansible

Creating and managing your infrastructure and installation using Ansible modules.

The Mesosphere DC/OS Ansible roles are now a supported life cycle management method for DC/OS alongside the Mesosphere Universal Installer. For users that are already using Ansible to manage their infrastructure and applications, the Ansible roles will be usable with only a few changes to their existing stack, as they have been developed with isolations and strict name spacing in mind. While they have been developed to work with infrastructure provisioned with the Mesosphere Universal Installer, they can be used in combination with any cloud and on-premise setups.

We have broken down the lifecycle management of DC/OS into 4 roles to handle the different management aspects of DC/OS such as cluster prerequisites, bootstrap tasks, master tasks, private agent tasks and public agent tasks. Each role can be read about more below:

  • Prerequisites - This role handles all of the requirements to run DC/OS.
  • GPU - This Role is a specialized addition to the Prerequisites role. It detects Nvidia GPUs and tries to install the Drives on CentOS systems.
  • Bootstrap - The Boostrap role handles all tasks associated with downloading, generating and serving the DC/OS install and upgrade scripts to all nodes in the cluster.
  • Master - These tasks include downloading install and upgrade files from the Bootstrap node as well as handling some checks to ensure that upgrade has gone accordingly. It will back out upgrades if there problems keeping the cluster from going into an undesired state.
  • Agents - These task handle all upgrade and installation tasks for ALL agent types.

About the Mesosphere DC/OS Ansibles roles

Use the Mesosphere DC/OS Ansible rolesthe official Ansible Galaxy.

NOTE: DC/OS Ansible roles are only available for CentOS and RHEL platforms at this time.

Creating test instances on AWS with terraform

If you would like to test out our Ansible solutions using pre-configured instances on AWS and are familiar with terraform, please download and use this terraform script to deploy a sample infrastructure so that you can deploy DC/OS on it using the ansible script above. The script is not officially supported by Mesosphere and is provided only for testing purposes.

To deploy the infra using the script:

  1. Create a new folder
  2. Copy the above script to a file named main.tf
  3. Initialize terraform: terraform init
  4. Deploy the infrastructure using terraform apply

Final output should be something like this:

  1. Apply complete! Resources: 32 added, 0 changed, 0 destroyed.
  3. Outputs:
  5. bootstrap_private_ip = 172.12.15.101
  6. bootstraps = 3.87.63.8
  7. cluster-address = dcosansible-660064571.us-east-1.elb.amazonaws.com
  8. masters = 100.27.19.199
  9. 54.196.221.181
  10. 35.168.16.40
  11. masters_private_ips = 172.12.6.95
  12. 172.12.29.65
  13. 172.12.42.160
  14. private_agents = 34.207.192.11
  15. 3.80.226.211
  16. 3.85.31.136
  17. public-agents-loadbalancer = ext-dcosansible-1616099901.us-east-1.elb.amazonaws.com
  18. public_agents = 3.86.34.141

To learn more about using Terraform as your deployment manager visit the Universal Installer page.

Using the Mesosphere DC/OS Ansible roles in combination with the Mesosphere Universal Installer

Mesosphere supports the use of a combination of the Universal InstallerTerraform-Ansible-Bridge-module and Ansible to manage the life cycle of the DC/OS software.

  1. module "dcos-ansible-bridge" {
  2.   source  = "dcos-terraform/dcos-ansible-bridge/local_file"
  3.   version = "~> 0.1.0"
  5.   bootstrap_ip         = "${module.dcos-infrastructure.bootstrap.public_ip}"
  6.   master_ips           = ["${module.dcos-infrastructure.masters.public_ips}"]
  7.   private_agent_ips    = ["${module.dcos-infrastructure.private_agents.public_ips}"]
  8.   public_agent_ips     = ["${module.dcos-infrastructure.public_agents.public_ips}"]
  10.   bootstrap_private_ip = "${module.dcos-infrastructure.bootstrap.private_ip}"
  11.   master_private_ips   = ["${module.dcos-infrastructure.masters.private_ips}"]
  12. }
  14. module "dcos-infrastructure" {
  15.   source  = "dcos-terraform/infrastructure/aws"
  16.   version = "~> 0.1.0"
  18.   [...]
  20. }

This will generate a local hosts file, which is an Ansbile compatible inventory file, that can be used to make Ansible aware of the cluster nodes as they are created by the Universal Installer. It will also generate a Ansbile compatible host vars file dcos.yml that will be populated by the clusters bootstrap and master nodes addresses. These two files can then be used to invoke any Ansible playbook against the cluster, such as the provided example one, to roll out the roles to their respective nodes.

Using the Mesosphere DC/OS Ansible roles for on-premise setups

Using Ansible to automate DC/OS installation, upgrades and configuration on on-premise setups is supported by Mesosphere. The Mesosphere provided Ansible rolesMesosphere DC/OS System Requirements and runs with CentOS/RHEL.

GPU Support

When using the provided example playbook, it will automatically detect NVIDIA GPU hardware and try to install the appropriate drivers. This methods tries to work without any restart - the GPU role identifies the currently used kernel and tries to get its headers either from the current base repository or from the Centos vault repositories. This might fail if you have a custom Kernel in use. In that case we kindly ask you to not use this module by simply commenting it out in your Playbook and handling the driver installation by yourself. Once the driver is installed and CUDA Libraries are available the DC/OS agent will automatically offer GPU resources. HINT: make sure to install the drivers before the DC/OS install happens.

DC/OS Ansible Quickstart Guide

Guided walkthrough of installing DC/OS using Ansible.

DC/OS Ansible FAQs and Troubleshooting

Frequently asked questions and common issues when working with Ansible for DC/OS