DebOps for Ansible users

To make integration of DebOps roles with your own infrastructure easier, DebOps playbooks introduce several new concepts to Ansible best practices.

Roles and playbooks are read-only

The current Ansible ecosystem relies on Ansible Galaxy for role distribution. The usual use case for Galaxy roles is either as examples to write your roles, or to heavily modify them according to your own needs.

DebOps roles and provided playbooks are designed to be used in a "read-only" mode. This allows easy upgrades of the project codebase by using the git pull command, as well as distributed development model where changes and new features can be shared between DebOps users through the central repository. You are not expected to have to modify roles directly; any and all changes necessary should be supported by the role default variables which can then be modified as needed through Ansible inventory.

Many roles support usage as "role dependencies" - other roles, either those in DebOps or written by its users, can request certain configuration to be defined on their behalf. This design principle allows for the DebOps roles to be focused on specific services or applications and encourages re-use of existing roles when writing custom ones by users.

For example, when you write a role that deploys a web application, you don't need to worry about required firewall configuration, web server or system package upgrades; existing roles can handle that for you.

Roles and playbooks are shared between multiple inventories

The custom debops* scripts provided with DebOps supports the creation and management of multiple "project directories". Each project directory contains its own Ansible inventory, as well as custom playbooks, roles and other resources written for that environment. This allows easy creation of development, test and production environments that use the same set of playbooks and differ only via inventory variables.

The official playbooks and roles are installed in central, fixed location (~/.local/share/debops/debops-playbooks/ on Linux systems), and the debops script generates ansible.cfg configuration file to provide correct paths for the ansible-playbook command to use them indirectly from the project directory.

The common playbook

In many Ansible environments a popular practice is to have a "common role" that contains tasks that are expected to be run on any and all hosts managed by Ansible.

In DebOps, there is an entire playbook dedicated to this, located in playbooks/common.yml. It includes multiple roles that prepare a host from an unknown to a known state - for example, a ferm-based firewall will be installed and configured on a given host, unless disabled, some common, useful packages will be installed, and so on.

Other DebOps roles not included in the common.yml playbook are designed for hosts that were configured by it - they might work outside of that environment, but it's not guaranteed.

File, template and task hooks

DebOps project introduces a set of Ansible lookup plugins which allow you to override certain aspects of public Ansible roles without modifying them directly. This allows for easier updates or customization of the files and templates according to your specific needs.

Certain roles use debops.debops.file_src or debops.debops.template_src to calculate path to files or templates used by a role. You can override these paths using .debops.cfg configuration file and provide your own versions of files and templates stored in DebOps project directory.

Some roles provide "task hooks" at the beginning and end of task lists, which are empty files in a specific subdirectories. Using debops.debops.task_src lookup plugin and settings defined in .debops.cfg configuration file you can "inject" your own tasks at the beginning or end of these roles, which gives you more control over the configuration.

By combining above techniques, you can very easily extend DebOps roles without losing the ability to update them, using git without having merge conflicts.

Ansible inventory is a source of truth

DebOps roles don't modify the Ansible inventory directly, it is treated as the "source of truth" defined by the user. This means that users can provide any inventory they want to use, be it a set of static YAML files, or a dynamic inventory based on scripts and a database.

Users don't even need to specifically configure things, as many roles use sensible defaults to configure the host according to its environment (DNS domain name, IP addresses, number of CPU cores, amount of RAM, network interfaces, etc.).

The aim is to have roles that work fine with the default configuration in a typical Debian installation, automatically implementing best practices for security where necessary.

Flattened lists in inventory

Some DebOps roles use sets of default variables (usually lists) to allow you to define different settings for all hosts in inventory, a group of hosts, or even specific hosts.

For example, using the debops.sshd role you can whitelist a certain subnet for all hosts in your inventory, add another subnet for a particular group of hosts, and so on.

You can also override the more general list on specific hosts if needed.

The [debops_*] host group namespace

To make host configuration in Ansible inventory more explicit, DebOps uses a set of Ansible host groups. All of the official groups are set in the [debops_*] namespace, so you are free to use other names without any possibility of a collision.

Common DebOps playbook, as well as some other service playbooks that are included in it, use [debops_all_hosts] group. This is a base group of the project and all hosts managed by DebOps should be included in it.

Service playbooks use the [debops_service_*] group namespace in Ansible inventory (for example, the debops.nginx role is activated on hosts in [debops_service_nginx] group). Some service playbooks use additional groups for various purposes; you are advised to check the role documentation to see what is their intended use case.

LDAP integration

Certain DebOps roles can access LDAP server to create or update data as needed. Custom modules are provided for LDAP entry and attribute management, deeper integration is planned in the future.