I created a complex state for API service, it involves git checkouts, python venv, uwsgi, nginx, etc etc. It works fine.
Now I would like to turn it into a template and execute it several times per minion, with variables supplied from pillar - i.e something like.
{% for apiserver in pillar.apiservers %}
include apiserver_template.sls, locals: apiserver.config
{% endfor %}
where apiserver_template will work with context supplied to it, with apiserver.config having all config data for each API instance. I know syntax is wrong but hopefully I am communicating the idea - ideally, something like executing ruby partials with supplying local variables.
How is it done properly in saltland?
It sounds to me like Jinja Macro is something you want to use for this. You can find more information about usage here: https://docs.saltstack.com/en/2015.8/topics/development/conventions/formulas.html#jinja-macros
In short what you will have in your case may look like:
{% macro api_server(git_repo, python_venv_path, python_venv_requirements) %}
{{python_venv_path}}:
virtualenv.managed:
- system_site_packages: False
- requirements: salt://{{python_venv_requirements}}
{{git_repo}}:
git.latest:
- name: {{git_repo}}
{% endmacro %}
Assuming you have a pillar apiservers where each api server has git_repo, python_venv_path and python_venv_requirements values, you can use the macro like this:
{% for server in salt.pillar.get('apiservers', []) %}
{{ api_server(server['git_repo'], server['python_venv_path'], server['python_venv_requirements']) }}
{% endfor %}
If you want - you can also put a macro in a separate state file and then import a marco as a regular salt resource.
Please also not that instead of pillar.apiservers I used salt.pillar.get('apiservers', []). This is a safer way to get data from pillar. If for some reason a pillar is unavailable - the later code will result in empty dict instead of failure in first case.