OST uses lago to bootstrap the architecture to run its test on, using LagoInitFile.

Essentially it is a yml file with definitions, here is and example of definition of (virtual) host:

  vm-type: ovirt-host
  memory: 2047

This will create 1 VM with 2047 ram, with the assigned name 'host-0'. If we want 100 hosts or to have some of definitions parameterized, we use jinja to pre-process the yml file.

Using Jinja templates

run-suite.sh, the entrypoint to run a suite, will take LagoInitfile.in and consume it as a jinja template, and output an interpolated yml file. This means we can use all sort of jinja expressions, here's couple of examples.

Jinja variables and 'env'

Lets use a simple jinja expression to create a unique host name using environment variable. The trick is done by 'run-suite.sh' scripts that render jinja template and loads all the shell environment into global variable 'env', that is accessible to jinja expressions in the template.

Here's how to use it to give a special name to a host:

# export suite_name=integ-tests
# cat LagoInitFile.in
{{ env.suite_name }}-host-0:    # integ-tests-host-0:
  vm-type: ovirt-host

jinja Loops

If we want 100 hosts, use jinja loops expression:

{% for i in range(100) %}
  {{ env.suite_name }}-host-{{ i }}:
    vm-type: ovirt-host
{% endfor %}

Load vars into the template context

By default jinja will loads vars/main.yml and load it into rendering context, similar to how it is done with environment variables.

yml # cat vars/main.yml hostCount: 100

Now this expression loops 100 times: yml {% for i in range(hostCount) %} ...

How it is done

There is a mini common python script, that invokes the template engine on the LagoInitFile.in and outputs the final file with first loading the environment and vars/main.yml demonstrated before. It is recommended that all suites will do the same, it requires nothing more than pointing the control.sh script to load in the pre_suite section:

bash prep_suite () { render_jinja_templates }