Roles are sets of Ansible defaults, files, tasks, templates, variables, and other Ansible components that work together. As you saw on Run Your First Command and Playbook, moving from a command to a playbook makes it easy to run multiple tasks and repeat the same tasks in the same order. Moving from a playbook to a role makes it even easier to reuse and share your ordered tasks. You can look at Ansible Galaxy, which lets you share your roles and use others’ roles, either directly or as inspiration.
So what exactly is a role, and why should you care? Ansible roles are basically playbooks broken up into a known file structure. Moving to roles from a playbook makes sharing, reading, and updating your Ansible workflow easier. Users can write their own roles. So for example, you don’t have to write your own DNS playbook. Instead, you specify a DNS server and a role to configure it for you.
To simplify your workflow even further, the Ansible Network team has written a series of roles for common network use cases. Using these roles means you don’t have to reinvent the wheel. Instead of writing and maintaining your own create_vlan playbooks or roles, you can concentrate on designing, codifying and maintaining the parser templates that describe your network topologies and inventory, and let Ansible’s network roles do the work. See the network-related roles on Ansible Galaxy.
To demonstrate the concept of what a role is, the example playbook.yml below is a single YAML file containing a two-task playbook. This Ansible Playbook configures the hostname on a Cisco IOS XE device, then it configures the DNS (domain name system) servers.
---
- name: configure cisco routers
hosts: routers
connection: ansible.netcommon.network_cli
gather_facts: no
vars:
dns: "8.8.8.8 8.8.4.4"
tasks:
- name: configure hostname
cisco.ios.ios_config:
lines: hostname {{ inventory_hostname }}
- name: configure DNS
cisco.ios.ios_config:
lines: ip name-server {{dns}}
If you run this playbook using the ansible-playbook command, you’ll see the output below. This example used -l option to limit the playbook to only executing on the rtr1 node.
[user@ansible ~]$ ansible-playbook playbook.yml -l rtr1 PLAY [configure cisco routers] ************************************************* TASK [configure hostname] ****************************************************** changed: [rtr1] TASK [configure DNS] *********************************************************** changed: [rtr1] PLAY RECAP ********************************************************************* rtr1 : ok=2 changed=2 unreachable=0 failed=0
This playbook configured the hostname and DNS servers. You can verify that configuration on the Cisco IOS XE rtr1 router:
rtr1#sh run | i name hostname rtr1 ip name-server 8.8.8.8 8.8.4.4
The next step is to convert this playbook into a reusable role. You can create the directory structure manually, or you can use ansible-galaxy init to create the standard framework for a role.
[user@ansible ~]$ ansible-galaxy init system-demo [user@ansible ~]$ cd system-demo/ [user@ansible system-demo]$ tree . ├── defaults │ └── main.yml ├── files ├── handlers │ └── main.yml ├── meta │ └── main.yml ├── README.md ├── tasks │ └── main.yml ├── templates ├── tests │ ├── inventory │ └── test.yml └── vars └── main.yml
This first demonstration uses only the tasks and vars directories. The directory structure would look as follows:
[user@ansible system-demo]$ tree
.
├── tasks
│ └── main.yml
└── vars
└── main.yml
Next, move the content of the vars and tasks sections from the original Ansible Playbook into the role. First, move the two tasks into the tasks/main.yml file:
[user@ansible system-demo]$ cat tasks/main.yml
---
- name: configure hostname
cisco.ios.ios_config:
lines: hostname {{ inventory_hostname }}
- name: configure DNS
cisco.ios.ios_config:
lines: ip name-server {{dns}}
Next, move the variables into the vars/main.yml file:
[user@ansible system-demo]$ cat vars/main.yml --- dns: "8.8.8.8 8.8.4.4"
Finally, modify the original Ansible Playbook to remove the tasks and vars sections and add the keyword roles with the name of the role, in this case system-demo. You’ll have this playbook:
---
- name: configure cisco routers
hosts: routers
connection: ansible.netcommon.network_cli
gather_facts: no
roles:
- system-demo
To summarize, this demonstration now has a total of three directories and three YAML files. There is the system-demo folder, which represents the role. This system-demo contains two folders, tasks and vars. There is a main.yml is each respective folder. The vars/main.yml contains the variables from playbook.yml. The tasks/main.yml contains the tasks from playbook.yml. The playbook.yml file has been modified to call the role rather than specifying vars and tasks directly. Here is a tree of the current working directory:
[user@ansible ~]$ tree
.
├── playbook.yml
└── system-demo
├── tasks
│ └── main.yml
└── vars
└── main.yml
Running the playbook results in identical behavior with slightly different output:
[user@ansible ~]$ ansible-playbook playbook.yml -l rtr1 PLAY [configure cisco routers] ************************************************* TASK [system-demo : configure hostname] **************************************** ok: [rtr1] TASK [system-demo : configure DNS] ********************************************* ok: [rtr1] PLAY RECAP ********************************************************************* rtr1 : ok=2 changed=0 unreachable=0 failed=0
As seen above each task is now prepended with the role name, in this case system-demo. When running a playbook that contains several roles, this will help pinpoint where a task is being called from. This playbook returned ok instead of changed because it has identical behavior for the single file playbook we started from.
As before, the playbook will generate the following configuration on a Cisco IOS-XE router:
rtr1#sh run | i name hostname rtr1 ip name-server 8.8.8.8 8.8.4.4
This is why Ansible roles can be simply thought of as deconstructed playbooks. They are simple, effective and reusable. Now another user can simply include the system-demo role instead of having to create a custom “hard coded” playbook.
What if you want to change the DNS servers? You aren’t expected to change the vars/main.yml within the role structure. Ansible has many places where you can specify variables for a given play. See Using Variables for details on variables and precedence. There are actually 21 places to put variables. While this list can seem overwhelming at first glance, the vast majority of use cases only involve knowing the spot for variables of least precedence and how to pass variables with most precedence. See Variable precedence: Where should I put a variable? for more guidance on where you should put variables.
The lowest precedence is the defaults directory within a role. This means all the other 20 locations you could potentially specify the variable will all take higher precedence than defaults, no matter what. To immediately give the vars from the system-demo role the least precedence, rename the vars directory to defaults.
[user@ansible system-demo]$ mv vars defaults [user@ansible system-demo]$ tree . ├── defaults │ └── main.yml ├── tasks │ └── main.yml
Add a new vars section to the playbook to override the default behavior (where the variable dns is set to 8.8.8.8 and 8.8.4.4). For this demonstration, set dns to 1.1.1.1, so playbook.yml becomes:
---
- name: configure cisco routers
hosts: routers
connection: ansible.netcommon.network_cli
gather_facts: no
vars:
dns: 1.1.1.1
roles:
- system-demo
Run this updated playbook on rtr2:
[user@ansible ~]$ ansible-playbook playbook.yml -l rtr2
The configuration on the rtr2 Cisco router will look as follows:
rtr2#sh run | i name-server ip name-server 1.1.1.1
The variable configured in the playbook now has precedence over the defaults directory. In fact, any other spot you configure variables would win over the values in the defaults directory.
Specifying variables in the defaults directory within a role will always take the lowest precedence, while specifying vars as extra vars with the -e or --extra-vars= will always take the highest precedence, no matter what. Re-running the playbook with the -e option overrides both the defaults directory (8.8.4.4 and 8.8.8.8) as well as the newly created vars within the playbook that contains the 1.1.1.1 dns server.
[user@ansible ~]$ ansible-playbook playbook.yml -e "dns=192.168.1.1" -l rtr3
The result on the Cisco IOS XE router will only contain the highest precedence setting of 192.168.1.1:
rtr3#sh run | i name-server ip name-server 192.168.1.1
How is this useful? Why should you care? Extra vars are commonly used by network operators to override defaults. A powerful example of this is with the Job Template Survey feature on AWX or the Red Hat Ansible Automation Platform. It is possible through the web UI to prompt a network operator to fill out parameters with a Web form. This can be really simple for non-technical playbook writers to execute a playbook using their Web browser.
The Ansible Galaxy page for a role lists all available versions. To update a locally installed role to a new or different version, use the ansible-galaxy install command with the version and --force option. You may also need to manually update any dependent roles to support this version. See the role Read Me tab in Galaxy for dependent role minimum version requirements.
[user@ansible]$ ansible-galaxy install mynamespace.my_role,v2.7.1 --force
See also
Ansible Galaxy user guide
© 2012–2018 Michael DeHaan
© 2018–2021 Red Hat, Inc.
Licensed under the GNU General Public License version 3.
https://docs.ansible.com/ansible/latest/network/getting_started/network_roles.html