How to handle environment specific role vars?

[Bo Vial]

I am trying to work out the best way to handle multiple environments (dev/stage/prod) in Ansible?

I have read the best practices outlined in the Ansible docs here: http://docs.ansible.com/ansible/latest/playbooks_best_practices.html

Having an entirely different inventory for dev/stage/prod makes sense but I am unsure how to achieve a similar separation for role vars.

For example, I have split the inventory per environment as follows:

inventories/

  production/

    hosts/

      groups_vars/

        group1

        group2

      host_vars/

        hostname1

        hostname2

  staging/

    hosts/

      group_vars/

        group1

      host_vars/

        stagehost1

The above works great for group vars and host vars, but I can't figure out a nice way to do something similar for role vars.

Currently, I have two possible candidate solutions

Option 1.

Set an environment variable 'ENVIRONMENT' to either 'production' or 'staging'. The first task in a role then includes a different vars file based on this environment variable. The directory structure would be:

roles/

  myrole/

    tasks/

      main.yml

    vars/

      main.yml <- common role vars

      production.yml <- production sepcific role vars

      staging.yml <- staging specific role vars

The first task in roles/myrole/tasks/main.yml would be

---

- name: Include environment specific role vars

  include_vars: "{{ lookup('env','ENVIRONMENT') }}.yml"

Option 2.

All my Ansible files will be stored in git. I could just keep a single roles/myroles/vars/main.yml file that is different between git branches "production" and "staging". I could then use git attributes to prevent this file from being modified during a merge so that they always remain independant. This solution would then result in a different main.yml depending on the git branch I am deploying.

I am currently leaning towards option 1, or are there better ways to manage this?