End to End Test Workflow

This documentation covers the details of Akri's end to end testing workflow, which runs Akri on several Kubernetes distributions and versions each time a commit is pushed to a PR or a PR is merged.

It will explain how it works, how the CI runs it, how to run it locally and how to write new tests.

The end to end test framework is based on Python and pytest, the dependencies are managed using Poetry. It aims to test different scenarios directly in a Kubernetes instance.

How tests are done

The test suite assumes a working, clean Kubernetes cluster (can be single node) and also needs a working kube config and helm client.

For every test suite, Akri will be installed (and uninstalled at the end of it) using helm with the needed discovery handlers.

All tests suite and cases can be run independently in any order.

The CI Test K3s, Kubernetes (Kubeadm) and MicroK8s Workflow

File: akri/.github/workflows/run-test-cases.yml

A GitHub workflow that:

  • runs Python pytest-based end-to-end tests;

  • through 4 different Kubernetes versions: 1.24, 1.25, 1.26, 1.27;

  • on 3 different Kubernetes distros: K3s, Kubernetes (Kubeadm), MicroK8s.

Jobs|Steps

The workflow comprises two jobs (build-containers and test-cases).

build-containers

build-containers builds container images for Akri 'controller', 'agent' and discovery handlers based upon the commit that triggers the workflow. Once build, these images are shared across the test-cases job, using GitHub Action upload-artifact.

When not running in a PR context, this is skipped

test-cases

test-cases uses a GitHub strategy to run its steps across the different Kubernetes distros and versions summarized at the top of this document.

New Kubernetes distro versions may be added to the job by adding entries to jobs.test-cases.strategy.matrix.kube. Each array entry must include:

PropertyDescription

runtime

The Kubernetes distribution name (k3s, k8s or microk8s)

version

A distro-specific unique identifier for the Kubernetes version

Notes:

  • runtime is used by subsequent steps as a way to determine the distro, e.g. startsWith(matrix.kube.runtime, 'k3s')

  • version is used by each distro to determine which binary, snap etc. to install. Refer to each distro's documentation to determine the value required

Distro installation and Akri container images insertion

Each distro has an installation step and a step to import the Akri images created by the build-containers job.

The installation steps are identified by:

if: startsWith(matrix.kube.runtime, ${DISTRO})

The installation instructions map closely with the installation instructions provided for the distro.

The container image import steps are identified by:

if: (startsWith(github.event_name, 'pull_request')) && (startsWith(matrix.kube.runtime, ${DISTRO}))

Tests

Of all the steps, only one is needed to run the Python end-to-end script.

The scripts arguments contains all needed information for it to run (see Run the tests locally for details on arguments).

stdout|stderr from the script can be logged to the workflow.

Upload logs

Once the end-to-end script is complete, the workflow uses the GitHub Action upload-artifact again to upload /tmp/log and so that these remain available (for download) once the workflow completes.

Run the tests locally

In order to run the test suite on your computer, you need:

You also need a clean Kubernetes cluster, one can be easily created using k3d.

All further commands are expected to be run from the /test/e2e/ directory.

To install the dependencies run poetry install.

To run all the tests run poetry run pytest -v --distribution ${DISTRO}, with DISTRO either k3s, k8s, or microk8s.

To run specific test suite, add the suite file as argument to the pytest command (e.g add test_webhook.py).

To run a specific test in a test suite add the fully quialified test name as argument to the pytest command (e.g add test_webhook::test_valid_configuration_accepted).

You can specify multiple tests or test suites in you pytest command.

By default, the tests will run on latest akri-dev chart.

There are other options in addition to --distribution that affect the test run:

OptionDescription

--distribution

Specify the target distribution for the tests, can be one of k3s, k8s or microk8s

--release

Use akri chart instead of akri-dev

--test-version

Version of the chart to use

--use-local

Use local chart (i.e /deployment/helm and local images)

--local-tag

When using local images, the tag used by images (by default will look for pr tag)

Technical details and writing new tests

All end to end tests suites are located in /test/e2e/, as we use pytest to run those, a test suite file must be named test_${SUITE}.py. Within these files every test_* functions will run independently, fixtures are here to help in setting up and tearing down the test evironment.

Every fixture will get set up when first used in scope (everything before the yield is executed), and teared down after last use in scope (everything after the yield is executed). Useful scopes for our usecases are session for the entire time of the pytest run, module for a specific suite and function for a specific test. An autouse fixture will get automatically added without explicitely asking for it.

Akri will be installed thanks to the autouse fixture install_akri, the scope of this fixture is module, meaning it will get installed and uninstalled for every test suite. This fixture is configured by the module variable discovery_handlers that must be set to a list of discovery handlers to enable.

A test passes if the function return; a test fails if the function raises an exception.

Be careful when writing new tests, the test order is not guarenteed, so make sure your test reverts any modification.

Last updated