An Introduction to Configuration Management with Ansible

Ansible keeps track of the state of resources in managed systems in order to avoid repeating tasks that were executed before. If a package was already installed, it won’t try to install it again. The objective is that after each provisioning execution the system reaches (or keeps) the desired state, even if you run it multiple times. This is what characterizes Ansible and other configuration management tools as having an idempotent behavior. When running a playbook, you’ll see the status of each task being executed and whether or not the task performed a change in the system.

When writing Ansible automation scripts, you can use variables, conditionals, and loops in order to make your automation more versatile and efficient.

Ansible collects a series of detailed information about the managed nodes, such as network interfaces and operating system, and provides it as global variables called system facts. Facts can be used within playbooks to make your automation more versatile and adaptive, behaving differently depending on the system being provisioned.

Ansible uses the Jinja2 Python templating system to allow for dynamic expressions and access to variables. Templates can be used to facilitate setting up configuration files and services. For instance, you can use a template to set up a new virtual host within Apache, while reusing the same template for multiple server installations.

Ansible comes with hundreds of built-in modules to facilitate writing automation for common systems administration tasks, such as installing packages with apt and synchronizing files with rsync, and also for dealing with popular software such as database systems (like MySQL, PostgreSQL, MongoDB, and others) and dependency management tools (such as PHP’s composer, Ruby’s gem, Node’s npm, and others). Apart from that, there are various ways in which you can extend Ansible: plugins and modules are good options when you need a custom functionality that is not present by default.

You can also find third-party modules and plugins in the Ansible Galaxy portal.

A control node is a system where Ansible is installed and set up to connect to your server. You can have multiple control nodes, and any system capable of running Ansible can be set up as a control node, including personal computers or laptops running a Linux or Unix based operating system. For the time being, Ansible can’t be installed on Windows hosts, but you can circumvent this limitation by setting up a virtual machine that runs Linux and running Ansible from there.

The systems you control using Ansible are called managed nodes. Ansible requires that managed nodes are reachable via SSH, and have Python 2 (version 2.6 or higher) or Python 3 (version 3.5 or higher) installed.

Ansible supports a variety of operating systems including Windows servers as managed nodes.

An inventory file contains a list of the hosts you’ll manage using Ansible. Although Ansible typically creates a default inventory file when installed, you can use per-project inventories to have a better separation of your infrastructure and avoid running commands or playbooks on the wrong server by mistake. Static inventories are usually created as .ini files, but you can also use dynamically generated inventories written in any programming language able to return JSON.

In Ansible, a task is an individual unit of work to execute on a managed node. Each action to perform is defined as a task. Tasks can be executed as a one-off action via ad-hoc commands, or included in a playbook as part of an automation script.

A playbook contains an ordered list of tasks, and a few other directives to indicate which hosts are the target of that automation, whether or not to use a privilege escalation system to run those tasks, and optional sections to define variables or include files. Ansible executes tasks sequentially, and a full playbook execution is called a play. Playbooks are written in YAML format.

Handlers are used to perform actions on a service, such as restarting or stopping a service that is actively running on the managed node’s system. Handlers are typically triggered by tasks, and their execution happens at the end of a play, after all tasks are finished. This way, if more than one task triggers a restart to a service, for instance, the service will only be restarted once and after all tasks are executed. Although the default handler behavior is more efficient and overall a better practice, it is also possible to force immediate handler execution if that is required by a task.

A role is a set of playbooks and related files organized into a predefined structure that is known by Ansible. Roles facilitate reusing and repurposing playbooks into shareable packages of granular automation for specific goals, such as installing a web server, installing a PHP environment, or setting up a MySQL server.