ATEX

ATEX is a framework for configurable test execution.

It is a set of Python-based abstract APIs and several implementations utilizing them, providing building blocks for you to make simple Python-based scripts that control the execution and result processing of your tests.

Its main building blocks are:

• Provisioners that give you systems to run tests on
• Executors that prepare and run the tests on them
• Aggregators that collect results from multiple tests
• Orchestrators that string everything up together, using Provisioners to get systems for Executors to run tests on, calling an Aggregator to ingest all test results

ATEX is not a linear pipeline like Provision -> Execute -> Report, the building blocks can be used independently and at any time.

Even during orchestration, Provisioners run in parallel to Executors, so that re-runs of failed tests can get fresh systems, and tests can start running as soon as one system is provisioned. Aggregators can upload to 3rd party services as soon as any one test finishes.

You are in control

The key part is that this is a framework to be used by YOU. Your script controls what gets used and how.

You can download / fetch tests from multiple repositories, modify their metadata on-the-fly, do anything you want via normal Python code, and also call ATEX building blocks to help you.

There is no "vendor lock-in" to one tool or ecosystem. You don't need ATEX to implement feature XYZ when you can write a trivial piece of Python code to do it (ie. pre-processing test metadata, post-processing results).

There are no boundaries for you to stay within - you don't need to implement a Provisioner using the Provisioner API. Just obtain the system somehow, wrap its SSH details in an SSHConnection*, and give that to an FMFExecutor.

You don't need to write a "plugin for ATEX", you just write Python code.

How it works

Each building block defines one or more abstract base classes, forming a sort-of stable reference API for everyone to use:

class Brewer:
    def intake(self, ingredients):
        """Input `ingredients` for brewing."""

    def brew(self):
        """Brew the beverage and return it."""

class CoffeeBrewer(Brewer):
    def __init__(self, kind, strength=None):
        self.kind = kind
        self.strength = strength if strength is not None else 5

    def _the_actual_brewing(self):
        ...

    def intake(self, ingredients, *, grind=True):
        ...

    def brew(self, *, speed=100):
        return self._the_actual_brewing()

Any piece of code can then state that it takes an initialized Brewer instance as an argument, and have a guarantee that it will have .intake(ingredients) and .brew() available, no matter the implementation.

b = CoffeeBrewer("espresso", 1000)
serve_to_employees(brewer=b)

See API Rules for more details.

Testing this project

There are some limited sanity tests provided via pytest, although:

• Some require additional variables (ie. Testing Farm) and will ERROR without them.
• Some take a long time (ie. Testing Farm) due to system provisioning taking a long time, so install pytest-xdist and run with a large -n.

Currently, the recommended approach is to split the execution:

# synchronously, because podman CLI has concurrency issues
pytest tests/provisioner/test_podman*.py

# in parallel, because provisioning takes a long time
export TESTING_FARM_API_TOKEN=...
export TESTING_FARM_COMPOSE=...
pytest -n 20 tests/provisioner/test_testingfarm.py

# needs a HVM-capable host to run, unprivileged qemu:///session is fine
#export LIBVIRT_DEFAULT_URI=...  # override default autodetection
export TEMPVIRT_LOCATION=...
pytest tests/provisioner/test_tempvirt.py

# these are hundreds of tests with all Fedora + CentOS Streams and podman/ssh
# versions, you might want to run only a sensible subset
pytest -k fedora tests/executor/{fmf,beakerlib}  # or -k "fedora and podman"

# fast enough for synchronous execution
pytest \
    tests/connection \
    tests/executor --ignore=tests/executor/{fmf,beakerlib} \
    tests/aggregator \
    tests/orchestrator \
    tests/provisioner/test_local.py

You can also export BASE_IMAGE as a podman image name or ID to be used across the tests as a baseline for container creation - for example, the executor tests above don't need -k as the matrix collapses to just the one image.

What it stands for

ATEX = Ad-hoc Test EXecution, named after the most prominent Orchestrator, originally the only one available.

The name comes from a (fairly unique to FMF/TMT ecosystem) approach that allows provisioning a pool of systems and scheduling tests on them as one would on an ad-hoc pool of thread/process workers - once a worker becomes free, it receives a test to run.

This is in contrast to a more common approach of splitting a large list of N tests onto M workers like N/M, which yields significant time penalties due to tests having very varied runtimes.