Skip to content

alphagov/notifications-functional-tests

Repository files navigation

notifications-functional-tests

The tests are:

  • functional/: tests of the Notify user interface and API (notifications.service.gov.uk)
  • document_download/: tests of the Documents user interface and API (documents.service.gov.uk)
  • provider_delivery/: tests for delivery of notifications (Staging and Production only)

These tests are run against preview, staging and production using Concourse. We run a full set of tests on preview but only a smaller set of tests, also known as smoke tests, on staging and production.

The Concourse jobs are defined in our infrastructure repo.

Uses

These tests are not intended to be used for load testing.

Installation

brew install --cask chromedriver # needs to be >= v2.32
brew install zbar  # Used for QR code reading

# https://github.com/npinchot/zbar/issues/3
mkdir -p ~/lib
ln -s $(brew --prefix zbar)/lib/libzbar.dylib ~/lib/libzbar.dylib

make bootstrap # install dependencies, etc.

Note: when you run the tests for the first time on a Mac, you may need to authorise chromedriver in your security settings ("System Preferences > Security & Privacy > General").

Running tests

Note, there is an order dependency in the main tests. The registration test must run before any of the other tests as a new user account created for each test run. That user account is used for all later browser based tests. Each test run will first register a user account using the configured FUNCTIONAL_TEST_EMAIL. The email account will have random characters added so that we do not have uniqueness issues with the email address of registered user.

Running the tests against your local development environment

Note: this is currently only supported for functional/ and document_download/ tests. See the next section if you want to run the provider_delivery/ tests locally.

If you are running Notify locally using docker-compose via notifications-local, then you need to set the following environment variables:

export FUNCTIONAL_TESTS_LOCAL_API_HOST=http://notify-api.localhost:6011
export FUNCTIONAL_TESTS_LOCAL_ADMIN_HOST=http://notify.localhost:6012

Populate the local database with fixture data using the make target, this will call database fixtures located in notifications-api:

make generate-local-dev-db-fixtures

Now run the following in other tabs / windows:

Fetch the required environment variables from notifications-credentials:

pass credentials/functional-tests/local-functional > environment_local.sh

Then source this script and run the tests:

source environment_local.sh

# run all the tests
make test

# run a specific test
pytest tests/functional/preview_and_dev/test_seeded_user.py

Running the tests against preview, staging or production

Users with the required services and templates have already been set up for each of these environments. The details for these are found in our credentials repo, under credentials/functional-tests. There are different sets of credentials depending on the tests you want to run e.g. staging-provider, staging-functional . Decrypt the credentials you need and paste them locally in a separate file e.g. environment_staging.sh. Then:

source environment_{env_name}.sh

# run specific tests, for example:
pytest tests/document_download/preview_and_dev

Running tests in parallel

We can reduce the duration of the test suite by running some tests in parallel.

We use the pytest x-dist plugin to support running tests in parallel automatically. The number of test runners is determined automatically using the pytest -n auto option. This will be set to the number of CPUs available on the test machine.

Each test runner launches a separate selenium/chromedriver instance so browser state is isolated between runners.

Defining parallel groups

We execute tests in parallel groups using the pytest --dist loadgroup option. This allows us to group tests by the authenticated user type or other logical domain - this is useful for functional tests that rely on a particular state of a real user environment during execution.

Parallel tests executed using the same user type can cause race conditions and interfere with other tests. Tests belonging to different groups are executed in parallel. Each test within the same group is executed sequentially by the same test runner.

We use the following annotations on test methods to define the groups:

@pytest.mark.xdist_group(name="registration-flow")
@pytest.mark.xdist_group(name="api-client")
@pytest.mark.xdist_group(name="seeded-email")
@pytest.mark.xdist_group(name="api-letters")

More groups generally equals better parallelisation (limited by test runner count). However, in the case of functional tests, increased parallelisation increases the risk of side effects and race conditions in the shared environment unless grouped carefully.

Database fixtures

For the functional tests to pass in any environment, they require certain database fixtures to exist.

For your local environment, these fixtures are now generated using the custom flask command functional-test-fixtures (defined in notifications-api/app/functional_tests_fixtures/init.py). The command will create all the database rows required for the funcitonal tests in an idempotent way and output an environment file the functional tests can use to execute against the environment. There are instructions on how to update the local database fixtures.

For our preview environment, these fixtures are not yet stored in code, but will be similar to (but not the same as) the local fixtures.

For our staging environment, these fixtures are found in db_fixtures/staging.sql.j2. You'll need to generate the actual fixtures by populating the jinja template with some secrets from the credentials repo. You can do this using make generate-staging-db-fixtures. You should only need to do this in the very rare case that our staging database has lost the fixtures that exist in there already.

For our production environment, these fixtures are not yet stored in code, but will be similar to (but not the same as) the staging fixtures.

Pre-commit

We use pre-commit to ensure that committed code meets basic standards for formatting, and will make basic fixes for you to save time and aggravation.

Install pre-commit system-wide with, eg brew install pre-commit. Then, install the hooks in this repository with pre-commit install --install-hooks.

Further documentation