Creating new integration tests

This section covers the following cases:

  • There are no integration tests for a collection or group of modules in a collection at all.

  • You are adding a new module and you want to include integration tests.

  • You want to add integration tests for a module that already exists without integration tests.

In other words, there are currently no tests for a module regardless of whether the module exists or not.

If the module already has tests, see Adding to an existing integration test.

Simplified example

Here is a simplified abstract example.

Let’s say we are going to add integration tests to a new module in the community.abstract collection which interacts with some service.

We checked and determined that there are no integration tests at all.

We should basically do the following:

  1. Install and run the service with a setup target.

  2. Create a test target.

  3. Add integration tests for the module.

  4. Run the tests.

  5. Fix the code and tests as needed, run the tests again, and repeat the cycle until they pass.

Note

You can reuse the setup target when implementing other targets that also use the same service.

  1. Clone the collection to the ~/ansible_collections/community.abstract directory on your local machine.

  2. From the ~/ansible_collections/community.abstract directory, create directories for the setup target:

mkdir -p tests/integration/targets/setup_abstract_service/tasks
  1. Write all the tasks needed to prepare the environment, install, and run the service.

For simplicity, let’s imagine that the service is available in the native distribution repositories and no sophisticated environment configuration is required.

Add the following tasks to the tests/integration/targets/setup_abstract_service/tasks/main.yml file to install and run the service:

- name: Install abstract service
  package:
    name: abstract_service

- name: Run the service
  systemd:
    name: abstract_service
    state: started

This is a very simplified example.

  1. Add the target for the module you are testing.

Let’s say the module is called abstract_service_info. Create the following directory structure in the target:

mkdir -p tests/integration/targets/abstract_service_info/tasks
mkdir -p tests/integration/targets/abstract_service_info/meta

Add all of the needed subdirectories. For example, if you are going to use defaults and files, add the defaults and files directories, and so on. The approach is the same as when you are creating a role.

  1. To make the setup_abstract_service target run before the module’s target, add the following lines to the tests/integration/targets/abstract_service_info/meta/main.yml file.

dependencies:
  - setup_abstract_service
  1. Start with writing a single stand-alone task to check that your module can interact with the service.

We assume that the abstract_service_info module fetches some information from the abstract_service and that it has two connection parameters.

Among other fields, it returns a field called version containing a service version.

Add the following to tests/integration/targets/abstract_service_info/tasks/main.yml:

- name: Fetch info from abstract service
  abstract_service_info:
    host: 127.0.0.1  # We assume the service accepts local connection by default
    port: 1234       # We assume that the service is listening to this port by default
  register: result   # This variable will contain the returned JSON including the server version

- name: Test the output
  assert:
    that:
      - result.version == '1.0.0'  # Check version field contains what we expect
  1. Run the tests with the -vvv argument.

If there are any issues with connectivity (for example, the service is not accepting connections) or with the code, the play will fail.

Examine the output to see at which step the failure occurred. Investigate the reason, fix it, and run again. Repeat the cycle until the test passes.

  1. If the test succeeds, write more tests. Refer to the Recommendations on coverage section for details.

community.postgresql example

Here is a real example of writing integration tests from scratch for the community.postgresql.postgresql_info module.

For the sake of simplicity, we will create very basic tests which we will run using the Ubuntu 20.04 test container.

We use Linux as a work environment and have git and docker installed and running.

We also installed ansible-core.

  1. Create the following directories in your home directory:

mkdir -p ~/ansible_collections/community
  1. Fork the collection repository through the GitHub web interface.

  2. Clone the forked repository from your profile to the created path:

git clone https://github.com/YOURACC/community.postgresql.git ~/ansible_collections/community/postgresql

If you prefer to use the SSH protocol:

git clone git@github.com:YOURACC/community.postgresql.git ~/ansible_collections/community/postgresql
  1. Go to the cloned repository:

cd ~/ansible_collections/community/postgresql
  1. Be sure you are in the default branch:

git status
  1. Checkout a test branch:

git checkout -b postgresql_info_tests
  1. Since we already have tests for the postgresql_info module, we will run the following command:

rm -rf tests/integration/targets/*

With all of the targets now removed, the current state is as if we do not have any integration tests for the community.postgresql collection at all. We can now start writing integration tests from scratch.

  1. We will start with creating a setup target that will install all required packages and will launch PostgreSQL. Create the following directories:

mkdir -p tests/integration/targets/setup_postgresql_db/tasks
  1. Create the tests/integration/targets/setup_postgresql_db/tasks/main.yml file and add the following tasks to it:

- name: Install required packages
  package:
    name:
      - apt-utils
      - postgresql
      - postgresql-common
      - python3-psycopg2

- name: Initialize PostgreSQL
  shell: . /usr/share/postgresql-common/maintscripts-functions && set_system_locale && /usr/bin/pg_createcluster -u postgres 12 main
  args:
    creates: /etc/postgresql/12/

- name: Start PostgreSQL service
  ansible.builtin.service:
    name: postgresql
    state: started

That is enough for our very basic example.

  1. Then, create the following directories for the postgresql_info target:

mkdir -p tests/integration/targets/postgresql_info/tasks tests/integration/targets/postgresql_info/meta
  1. To make the setup_postgresql_db target run before the postgresql_info target as a dependency, create the tests/integration/targets/postgresql_info/meta/main.yml file and add the following code to it:

dependencies:
  - setup_postgresql_db
  1. Now we are ready to add our first test task for the postgresql_info module. Create the tests/integration/targets/postgresql_info/tasks/main.yml file and add the following code to it:

- name: Test postgresql_info module
  become: true
  become_user: postgres
  community.postgresql.postgresql_info:
    login_user: postgres
    login_db: postgres
  register: result

- name: Check the module returns what we expect
  assert:
    that:
      - result is not changed
      - result.version.major == 12
      - result.version.minor == 8

In the first task, we run the postgresql_info module to fetch information from the database we installed and launched with the setup_postgresql_db target. We are saving the values returned by the module into the result variable.

In the second task, we check the result variable, which is what the first task returned, with the assert module. We expect that, among other things, the result has the version and reports that the system state has not been changed.

  1. Run the tests in the Ubuntu 20.04 docker container:

ansible-test integration postgresql_info --docker ubuntu2004 -vvv

The tests should pass. If we look at the output, we should see something like the following:

TASK [postgresql_info : Check the module returns what we expect] ***************
ok: [testhost] => {
  "changed": false,
  "msg": "All assertions passed"
}

If your tests fail when you are working on your project, examine the output to see at which step the failure occurred. Investigate the reason, fix it, and run again. Repeat the cycle until the test passes. If the test succeeds, write more tests. Refer to the Recommendations on coverage section for details.