Testing Keystone

Running Tests

Before running tests, you should have tox installed and available in your environment (in addition to the other external dependencies in Setting up Keystone):

$ pip install tox


You may need to perform both the above operation and the next inside a python virtualenv, or prefix the above command with sudo, depending on your preference.

To execute the full suite of tests maintained within keystone, simply run:

$ tox

This iterates over multiple configuration variations, and uses external projects to do light integration testing to verify the Identity API against other projects.


The first time you run tox, it will take additional time to build virtualenvs. You can later use the -r option with tox to rebuild your virtualenv in a similar manner.

To run tests for one or more specific test environments (for example, the most common configuration of Python 3.6 and PEP-8), list the environments with the -e option, separated by spaces:

$ tox -e py36,pep8


Keystone dropped the support of python 2.7 in the Ussuri release of Openstack.

Use tox --listenvs to list all testing environments specified in keystone’s tox.ini file.

Interactive debugging

Using pdb breakpoints with tox and testr normally doesn’t work since the tests just fail with a BdbQuit exception rather than stopping at the breakpoint.

To capture breakpoints while running tests, use the debug environment. The following example uses the environment while invoking a specific test run.

$ tox -e debug keystone.tests.unit.test_module.TestClass.test_case

For reference, the debug environment implements the instructions here: https://wiki.openstack.org/wiki/Testr#Debugging_.28pdb.29_Tests

Building the Documentation

The docs and api-ref environments will automatically generate documentation and the API reference respectively. The results are written to doc/ and api-ref/.

For example, use the following command to render all documentation and manual pages:

$ tox -e docs

Tests Structure

Not all of the tests in the keystone/tests/unit directory are strictly unit tests. Keystone intentionally includes tests that run the service locally and drives the entire configuration to achieve basic functional testing.

For the functional tests, an in-memory key-value store or in-memory SQLite database is used to keep the tests fast.

Within the tests directory, the general structure of the backend tests is a basic set of tests represented under a test class, and then subclasses of those tests under other classes with different configurations to drive different backends through the APIs. To add tests covering all drivers, update the base test class in test_backend.py.


The structure of backend testing is in transition, migrating from having all classes in a single file (test_backend.py) to one where there is a directory structure to reduce the size of the test files. See:

  • keystone.tests.unit.backend.role

  • keystone.tests.unit.backend.domain_config

To add new drivers, subclass the base class at test_backend.py (look at test_backend_sql.py for examples) and update the configuration of the test class in setUp().

For example, test_backend.py has a sequence of tests under the class keystone.tests.unit.test_backend.IdentityTests that will work with the default drivers. The test_backend_sql.py module subclasses those tests, changing the configuration by overriding with configuration files stored in the tests/unit/config_files directory aimed at enabling the SQL backend for the Identity module.

Testing Schema Migrations

Tests for database migrations can be found in keystone/tests/unit/test_sql_upgrade.py and keystone/tests/unit/test_sql_banned_operations.py.

LDAP Tests

LDAP has a fake backend that performs rudimentary operations. If you are building more significant LDAP functionality, you should test against a live LDAP server. Devstack has an option to set up a directory server for Keystone to use. Add ldap to the ENABLED_SERVICES environment variable, and set environment variables KEYSTONE_IDENTITY_BACKEND=ldap and KEYSTONE_CLEAR_LDAP=yes in your localrc file.

The unit tests can be run against a live server with keystone/tests/unit/test_ldap_livetest.py and keystone/tests/unit/test_ldap_pool_livetest.py. The default password is test but if you have installed devstack with a different LDAP password, modify the file keystone/tests/unit/config_files/backend_liveldap.conf and keystone/tests/unit/config_files/backend_pool_liveldap.conf to reflect your password.


To run the live tests you need to set the environment variable ENABLE_LDAP_LIVE_TEST to a non-negative value.

“Work in progress” Tests

Work in progress (WIP) tests are very useful in a variety of situations including:

  • While doing test-driven-development they can be used to add tests to a review while they are not yet working and will not cause test failures. They can be removed when the functionality is fixed in a later patch set.

  • A common practice is to recreate bugs by exposing the broken behavior in a functional or unit test. To encapsulate the correct behavior in the test, the test will usually assert the correct outcome, which will break without a fix. Marking the test as WIP gives us the ability to capture the broken behavior in code if a fix isn’t ready yet.

The keystone.tests.unit.utils.wip() decorator can be used to mark a test as WIP. A WIP test will always be run. If the test fails then a SkipTest exception is raised because we expect the test to fail. We do not pass the test in this case so that it doesn’t count toward the number of successfully run tests. If the test passes an AssertionError exception is raised so that the developer knows they made the test pass. This is a reminder to remove the decorator.

The keystone.tests.unit.utils.wip() decorator requires that the author provides a message. This message is important because it will tell other developers why this test is marked as a work in progress. Reviewers will require that these messages are descriptive and accurate.


The keystone.tests.unit.utils.wip() decorator is not a replacement for skipping tests.

@wip('waiting on bug #000000')
def test():


Another strategy is to not use the wip decorator and instead show how the code currently incorrectly works. Which strategy is chosen is up to the developer.

API & Scenario Tests

Keystone provides API and scenario tests via a tempest plugin which is located in a separate repository. This tempest plugin is mainly intended for specific scenarios that require a special deployment, such as the tests for the Federated Identity feature or live testing against LDAP. For the deployment of these scenarios, keystone also provides a devstack plugin.

For example, to setup a working federated environment, add the following lines in your devstack local.conf` file:

enable_plugin keystone https://opendev.org/openstack/keystone
enable_service keystone-saml2-federation

Clone and install keystone-tempest-plugin.

git clone https://opendev.org/openstack/keystone-tempest-plugin
sudo pip install ./keystone-tempest-plugin

Finally, to run keystone’s API and scenario tests, deploy tempest with devstack (using the configuration above) and then run the following command from the tempest directory:

tox -e all -- keystone_tempest_plugin


Most of keystone’s API tests are implemented in tempest and it is usually the correct place to add new tests.

Writing new API & Scenario Tests

When writing tests for the keystone tempest plugin, we should follow the official tempest guidelines, details about the guidelines can be found at the tempest coding guide. There are also specific guides for the API and scenario tests: Tempest Field Guide to API tests and Tempest Field Guide to Scenario tests.

The keystone tempest plugin also provides a base class. For most cases, the tests should inherit from it: keystone_tempest_plugin.tests.base.BaseIdentityTest. This class already setups the identity API version and is the container of all API services clients. New API services clients keystone_tempest_plugin.services (which are used to communicate with the REST API from the services) should also be added to this class. For example, below we have a snippet from the tests at keystone_tempest_plugin.tests.api.identity.v3.test_identity_providers.py.

class IdentityProvidersTest(base.BaseIdentityTest):


def _create_idp(self, idp_id, idp_ref):
    idp = self.idps_client.create_identity_provider(
        idp_id, **idp_ref)['identity_provider']
        self.idps_client.delete_identity_provider, idp_id)
    return idp

def test_identity_provider_create(self):
    idp_id = data_utils.rand_uuid_hex()
    idp_ref = fixtures.idp_ref()
    idp = self._create_idp(idp_id, idp_ref)

    # The identity provider is disabled by default
    idp_ref['enabled'] = False

    # The remote_ids attribute should be set to an empty list by default
    idp_ref['remote_ids'] = []

    self._assert_identity_provider_attributes(idp, idp_id, idp_ref)

The test class extends keystone_tempest_plugin.tests.base.BaseIdentityTest. Also, the _create_idp method calls keystone’s API using the idps_client, which is an instance from. keystone_tempest_plugin.tests.services.identity.v3.identity_providers_client.IdentityProvidersClient.

Additionally, to illustrate the construction of a new test class, below we have a snippet from the scenario test that checks the complete federated authentication workflow ( keystone_tempest_plugin.tests.scenario.test_federated_authentication.py). In the test setup, all of the needed resources are created using the API service clients. Since it is a scenario test, it is common to need some customized settings that will come from the environment (in this case, from the devstack plugin) - these settings are collected in the _setup_settings method.

class TestSaml2EcpFederatedAuthentication(base.BaseIdentityTest):


def _setup_settings(self):
    self.idp_id = CONF.fed_scenario.idp_id
    self.idp_url = CONF.fed_scenario.idp_ecp_url
    self.keystone_v3_endpoint = CONF.identity.uri_v3
    self.password = CONF.fed_scenario.idp_password
    self.protocol_id = CONF.fed_scenario.protocol_id
    self.username = CONF.fed_scenario.idp_username


def setUp(self):
    super(TestSaml2EcpFederatedAuthentication, self).setUp()

    # Reset client's session to avoid getting garbage from another runs

    # Setup identity provider, mapping and protocol

Finally, the tests perform the complete workflow of the feature, asserting correctness in each step:

def _request_unscoped_token(self):
    resp = self.saml2_client.send_service_provider_request(
        self.keystone_v3_endpoint, self.idp_id, self.protocol_id)
    self.assertEqual(http_client.OK, resp.status_code)
    saml2_authn_request = etree.XML(resp.content)

    relay_state = self._str_from_xml(
        saml2_authn_request, self.ECP_RELAY_STATE)
    sp_consumer_url = self._str_from_xml(
        saml2_authn_request, self.ECP_SERVICE_PROVIDER_CONSUMER_URL)

    # Perform the authn request to the identity provider
    resp = self.saml2_client.send_identity_provider_authn_request(
        saml2_authn_request, self.idp_url, self.username, self.password)
    self.assertEqual(http_client.OK, resp.status_code)
    saml2_idp_authn_response = etree.XML(resp.content)

    idp_consumer_url = self._str_from_xml(
        saml2_idp_authn_response, self.ECP_IDP_CONSUMER_URL)

    # Assert that both saml2_authn_request and saml2_idp_authn_response
    # have the same consumer URL.
    self.assertEqual(sp_consumer_url, idp_consumer_url)


                      "Federated Identity feature not enabled")
def test_request_unscoped_token(self):

Notice that the test_request_unscoped_token test only executes if the federation feature flag is enabled.


For each patch submitted upstream, all of the tests from the keystone tempest plugin are executed in the gate-keystone-dsvm-functional-v3-only-* job.