# How Trellis uses Ansible
Since Trellis is powered by Ansible, the best way to understand Trellis is to understand Ansible itself. Even knowing a just key Ansible concepts will help you learn Trellis and how to customize it to fit your needs.
Ansible's own documentation (opens new window) is very comprehensive and should be considered as an extension of Trellis' documentation.
However, since Ansible itself is unopinionated, this will explore some key concepts and how they apply to Trellis.
At the highest level, Trellis provides a few playbooks which execute tasks organized into roles.
Trellis' playbooks are found in the root of Trellis itself:
dev.yml(opens new window) - provisions a development server. This playbook assumes that your local Trellis project files have been synced to a virtual machine and automatically installs WordPress.
server.yml(opens new window) - provisions a remote (non-dev) server. This playbook assumes you will be deploying sites separately and does not attempt to install WordPress.
deploy.yml(opens new window) - deploys a single site to an environment
rollback.yml(opens new window) - rolls back a previously deployed release
xdebug-tunnel.yml(opens new window) - opens or closes the PHP Xdebug tunnel
Each playbook listed above contains a list of roles to run. A role's main
purpose is to group a collection of tasks to run within the
All of Trellis' roles are found under the top-level
roles (opens new window) directory. Additionally, there are some 3rd party community roles used from Ansible Galaxy which are specified in the
galaxy.yml (opens new window) file.
Roles in Trellis usually contain one of more of these subfolders:
defaults- variables defined with low precedence
tasks- tasks to be executed - the main functional part of roles
templates- templates in Jinja format which are used in tasks
In Ansible, inventory (opens new window) is a list of defined hosts in your infrastructure.
For most Trellis projects, this list of hosts is usually one development virtual machine, one staging server (optional), and one production server.
If you look at the default inventory files in
hosts (opens new window) directory, you'll see three files named after the standard environments:
Here's what an inventory file in Trellis looks like:
[production] your_server_hostname [web] your_server_hostname
Each host is under two groups:
web. These groups can be used
for any semantic grouping you want, but in Trellis you at least need those two
# Group variables
The "group" naming isn't the most clear, but as shown above, these refer to Ansible's concept of "inventory groups". And since Trellis' inventory hosts are named for environments, "group vars" are really just environment specific variables. Though they can also be used for any semantic grouping of inventory hosts for more advanced use cases.
all group (in
group_vars/all) is special and applies to all groups.
All variables in Ansible can be considered global. Even if a variable is
defined within a role (eg:
roles/nginx/defaults/main.yml), it can be
referenced or re-defined in a
group_vars file. Once a role is included in a
playbook, their variables (in
vars) are available globally.
As an example, let's say you wanted to change PHP's max execution time in development to be higher than in production.
We can apply two things we learned above:
- variables are global
- group vars can be used to define environment specific values
Taking advantage of Ansible's variable
precendence (opens new window), we'll just override the variable by re-defining it in