Salt's Test Suite: An Introduction

Note

This tutorial makes a couple of assumptions. The first assumption is that you have a basic knowledge of Salt. To get up to speed, check out the Salt Walkthrough.

The second assumption is that your Salt development environment is already configured and that you have a basic understanding of contributing to the Salt codebase. If you're unfamiliar with either of these topics, please refer to the Installing Salt for Development and the Contributing pages, respectively.

Salt comes with a powerful integration and unit test suite. The test suite allows for the fully automated run of integration and/or unit tests from a single interface.

Salt's test suite is located under the tests directory in the root of Salt's code base and is divided into two main types of tests: unit tests and integration tests. The unit and integration sub-test-suites are located in the tests directory, which is where the majority of Salt's test cases are housed.

Getting Set Up For Tests

First of all you will need to ensure you install nox.

pip install nox

Test Directory Structure

As noted in the introduction to this tutorial, Salt's test suite is located in the tests directory in the root of Salt's code base. From there, the tests are divided into two groups integration and unit. Within each of these directories, the directory structure roughly mirrors the directory structure of Salt's own codebase. For example, the files inside tests/integration/modules contains tests for the files located within salt/modules.

Note

tests/integration and tests/unit are the only directories discussed in this tutorial. With the exception of the tests/runtests.py file, which is used below in the Running the Test Suite section, the other directories and files located in tests are outside the scope of this tutorial.

Integration vs. Unit

Given that Salt's test suite contains two powerful, though very different, testing approaches, when should you write integration tests and when should you write unit tests?

Integration tests use Salt masters, minions, and a syndic to test salt functionality directly and focus on testing the interaction of these components. Salt's integration test runner includes functionality to run Salt execution modules, runners, states, shell commands, salt-ssh commands, salt-api commands, and more. This provides a tremendous ability to use Salt to test itself and makes writing such tests a breeze. Integration tests are the preferred method of testing Salt functionality when possible.

Unit tests do not spin up any Salt daemons, but instead find their value in testing singular implementations of individual functions. Instead of testing against specific interactions, unit tests should be used to test a function's logic. Unit tests should be used to test a function's exit point(s) such as any return or raises statements.

Unit tests are also useful in cases where writing an integration test might not be possible. While the integration test suite is extremely powerful, unfortunately at this time, it does not cover all functional areas of Salt's ecosystem. For example, at the time of this writing, there is not a way to write integration tests for Proxy Minions. Since the test runner will need to be adjusted to account for Proxy Minion processes, unit tests can still provide some testing support in the interim by testing the logic contained inside Proxy Minion functions.

Running the Test Suite

Once all of the requirements are installed, the nox command is used to instantiate Salt's test suite:

nox -e 'pytest-zeromq-3(coverage=False)'

The command above, if executed without any options, will run the entire suite of integration and unit tests. Some tests require certain flags to run, such as destructive tests. If these flags are not included, then the test suite will only perform the tests that don't require special attention.

At the end of the test run, you will see a summary output of the tests that passed, failed, or were skipped.

You can pass any pytest options after the nox command like so:

nox -e 'pytest-zeromq-3(coverage=False)' -- tests/unit/modules/test_ps.py

The above command will run the test_ps.py test with the zeromq transport, python3, and pytest. The option after -- can include any pytest options.

Running Integration Tests

Salt's set of integration tests use Salt to test itself. The integration portion of the test suite includes some built-in Salt daemons that will spin up in preparation of the test run. This list of Salt daemon processes includes:

  • 2 Salt Masters

  • 2 Salt Minions

  • 1 Salt Syndic

These various daemons are used to execute Salt commands and functionality within the test suite, allowing you to write tests to assert against expected or unexpected behaviors.

A simple example of a test utilizing a typical master/minion execution module command is the test for the test_ping function in the tests/integration/modules/test_test.py file:

def test_ping(self):
    """
    test.ping
    """
    self.assertTrue(self.run_function("test.ping"))

The test above is a very simple example where the test.ping function is executed by Salt's test suite runner and is asserting that the minion returned with a True response.

Test Selection Options

If you want to run only a subset of tests, this is easily done with pytest. You only need to point the test runner to the directory. For example if you want to run all integration module tests:

nox -e 'pytest-zeromq-3(coverage=False)' -- tests/integration/modules/

Running Unit Tests

If you want to run only the unit tests, you can just pass the unit test directory as an option to the test runner.

The unit tests do not spin up any Salt testing daemons as the integration tests do and execute very quickly compared to the integration tests.

nox -e 'pytest-zeromq-3(coverage=False)' -- tests/unit/

Running Specific Tests

There are times when a specific test file, test class, or even a single, individual test need to be executed, such as when writing new tests. In these situations, you should use the pytest syntax to select the specific tests.

For running a single test file, such as the pillar module test file in the integration test directory, you must provide the file path.

nox -e 'pytest-zeromq-3(coverage=False)' -- tests/integration/modules/test_pillar.py

Some test files contain only one test class while other test files contain multiple test classes. To run a specific test class within the file, append the name of the test class to the end of the file path:

nox -e 'pytest-zeromq-3(coverage=False)' -- tests/integration/modules/test_pillar.py::PillarModuleTest

To run a single test within a file, append both the name of the test class the individual test belongs to, as well as the name of the test itself:

nox -e 'pytest-zeromq-3(coverage=False)' -- tests/integration/modules/test_pillar.py::PillarModuleTest::test_data

The following command is an example of how to execute a single test found in the tests/unit/modules/test_cp.py file:

nox -e 'pytest-zeromq-3(coverage=False)' -- tests/unit/modules/test_cp.py::CpTestCase::test_get_file_not_found

Writing Tests for Salt

Once you're comfortable running tests, you can now start writing them! Be sure to review the Integration vs. Unit section of this tutorial to determine what type of test makes the most sense for the code you're testing.

Note

There are many decorators, naming conventions, and code specifications required for Salt test files. We will not be covering all of the these specifics in this tutorial. Please refer to the testing documentation links listed below in the Additional Testing Documentation section to learn more about these requirements.

In the following sections, the test examples assume the "new" test is added to a test file that is already present and regularly running in the test suite and is written with the correct requirements.

Writing Integration Tests

Since integration tests validate against a running environment, as explained in the Running Integration Tests section of this tutorial, integration tests are very easy to write and are generally the preferred method of writing Salt tests.

The following integration test is an example taken from the test.py file in the tests/integration/modules directory. This test uses the run_function method to test the functionality of a traditional execution module command.

The run_function method uses the integration test daemons to execute a module.function command as you would with Salt. The minion runs the function and returns. The test also uses Python's Assert Functions to test that the minion's return is expected.

def test_ping(self):
    """
    test.ping
    """
    self.assertTrue(self.run_function("test.ping"))

Args can be passed in to the run_function method as well:

def test_echo(self):
    """
    test.echo
    """
    self.assertEqual(self.run_function("test.echo", ["text"]), "text")

The next example is taken from the tests/integration/modules/test_aliases.py file and demonstrates how to pass kwargs to the run_function call. Also note that this test uses another salt function to ensure the correct data is present (via the aliases.set_target call) before attempting to assert what the aliases.get_target call should return.

def test_set_target(self):
    """
    aliases.set_target and aliases.get_target
    """
    set_ret = self.run_function("aliases.set_target", alias="fred", target="bob")
    self.assertTrue(set_ret)
    tgt_ret = self.run_function("aliases.get_target", alias="fred")
    self.assertEqual(tgt_ret, "bob")

Using multiple Salt commands in this manner provides two useful benefits. The first is that it provides some additional coverage for the aliases.set_target function. The second benefit is the call to aliases.get_target is not dependent on the presence of any aliases set outside of this test. Tests should not be dependent on the previous execution, success, or failure of other tests. They should be isolated from other tests as much as possible.

While it might be tempting to build out a test file where tests depend on one another before running, this should be avoided. SaltStack recommends that each test should test a single functionality and not rely on other tests. Therefore, when possible, individual tests should also be broken up into singular pieces. These are not hard-and-fast rules, but serve more as recommendations to keep the test suite simple. This helps with debugging code and related tests when failures occur and problems are exposed. There may be instances where large tests use many asserts to set up a use case that protects against potential regressions.

Note

The examples above all use the run_function option to test execution module functions in a traditional master/minion environment. To see examples of how to test other common Salt components such as runners, salt-api, and more, please refer to the Integration Test Class Examples documentation.

Destructive vs Non-destructive Tests

Since Salt is used to change the settings and behavior of systems, often, the best approach to run tests is to make actual changes to an underlying system. This is where the concept of destructive integration tests comes into play. Tests can be written to alter the system they are running on. This capability is what fills in the gap needed to properly test aspects of system management like package installation.

To write a destructive test, import and use the destructiveTest decorator for the test method:

import integration
from tests.support.helpers import destructiveTest


class PkgTest(integration.ModuleCase):
    @destructiveTest
    def test_pkg_install(self):
        ret = self.run_function("pkg.install", name="finch")
        self.assertSaltTrueReturn(ret)
        ret = self.run_function("pkg.purge", name="finch")
        self.assertSaltTrueReturn(ret)

Writing Unit Tests

As explained in the Integration vs. Unit section above, unit tests should be written to test the logic of a function. This includes focusing on testing return and raises statements. Substantial effort should be made to mock external resources that are used in the code being tested.

External resources that should be mocked include, but are not limited to, APIs, function calls, external data either globally available or passed in through function arguments, file data, etc. This practice helps to isolate unit tests to test Salt logic. One handy way to think about writing unit tests is to "block all of the exits". More information about how to properly mock external resources can be found in Salt's Unit Test documentation.

Salt's unit tests utilize Python's mock class as well as MagicMock. The @patch decorator is also heavily used when "blocking all the exits".

A simple example of a unit test currently in use in Salt is the test_get_file_not_found test in the tests/unit/modules/test_cp.py file. This test uses the @patch decorator and MagicMock to mock the return of the call to Salt's cp.hash_file execution module function. This ensures that we're testing the cp.get_file function directly, instead of inadvertently testing the call to cp.hash_file, which is used in cp.get_file.

def test_get_file_not_found(self):
    """
    Test if get_file can't find the file.
    """
    with patch("salt.modules.cp.hash_file", MagicMock(return_value=False)):
        path = "salt://saltines"
        dest = "/srv/salt/cheese"
        ret = ""
        assert cp.get_file(path, dest) == ret

Note that Salt's cp module is imported at the top of the file, along with all of the other necessary testing imports. The get_file function is then called directed in the testing function, instead of using the run_function method as the integration test examples do above.

The call to cp.get_file returns an empty string when a hash_file isn't found. Therefore, the example above is a good illustration of a unit test "blocking the exits" via the @patch decorator, as well as testing logic via asserting against the return statement in the if clause. In this example we used the python assert to verify the return from cp.get_file. Pytest allows you to use these asserts when writing your tests and, in fact, plain asserts is the preferred way to assert anything in your tests. As Salt dives deeper into Pytest, the use of unittest.TestClass will be replaced by plain test functions, or test functions grouped in a class, which does not subclass unittest.TestClass, which, of course, doesn't work with unittest assert functions.

There are more examples of writing unit tests of varying complexities available in the following docs:

Note

Considerable care should be made to ensure that you're testing something useful in your test functions. It is very easy to fall into a situation where you have mocked so much of the original function that the test results in only asserting against the data you have provided. This results in a poor and fragile unit test.

Add a python module dependency to the test run

The test dependencies for python modules are managed under the requirements/static directory. You will need to add your module to the appropriate file under requirements/static. When pre-commit is run it will create all of the needed requirement files under requirements/static/py3{5,6,7}. Nox will then use these files to install the requirements for the tests.

Add a system dependency to the test run

If you need to add a system dependency for the test run, this will need to be added in the salt jenkins repo. This repo uses salt states to install system dependencies. You need to update the state-tree/golden-images-provision.sls file with your dependency to ensure it is installed. Once your PR is merged the core team will need to promote the new images with your new dependency installed.

Checking for Log Messages

To test to see if a given log message has been emitted, the following pattern can be used

# Import logging handler
from tests.support.helpers import TstSuiteLoggingHandler

# .. inside test
with TstSuiteLoggingHandler() as handler:
    for message in handler.messages:
        if message.startswith("ERROR: This is the error message we seek"):
            break
        else:
            raise AssertionError("Did not find error message")

Automated Test Runs

SaltStack maintains a Jenkins server which can be viewed at https://jenkinsci.saltstack.com. The tests executed from this Jenkins server create fresh virtual machines for each test run, then execute the destructive tests on the new, clean virtual machine. This allows for the execution of tests across supported platforms.

Additional Testing Documentation

In addition to this tutorial, there are some other helpful resources and documentation that go into more depth on Salt's test runner, writing tests for Salt code, and general Python testing documentation. Please see the follow references for more information: