Skip to content

Latest commit

 

History

History
253 lines (190 loc) · 9.96 KB

testing.md

File metadata and controls

253 lines (190 loc) · 9.96 KB

Testing

All unit and lint tests are run using tox, with different versions of Python with specific constraints. We also use tox-setuptools-version to specify the correct setuptools version based on what is present in each release, and tox-pyenv to recognize the different local pyenv interpreters.

First, run the script to install and configure pyenv, and the tox dependencies:

./tools/setup_pyenv.sh

After that you need to set up your shell environment, according to the pyenv documentation. The guide has quick snippets to configure bash, zsh and fish shells.

Refresh your terminal to make sure pyenv is working. Then you can run the unit and lint tests:

tox

The client also includes built-in dep8 tests. These are run as follows:

autopkgtest -U --shell-fail . -- lxd ubuntu:xenial

Integration Tests

Ubuntu Pro Client uses behave for its integration testing.

The integration test definitions are stored in the features/ directory and consist of two parts: .feature files that define the tests we want to run, and .py files which implement the underlying logic for those tests.

By default, integration tests will do the following on a given cloud platform:

  • Launch an instance running latest daily image of the target Ubuntu release
  • Add the Ubuntu Pro client daily build PPA: ppa:ua-client/daily
  • Install the appropriate ubuntu-advantage-tools and ubuntu-advantage-pro deb
  • Run the integration tests on that instance.

The testing can be overridden to install ubuntu-advantage-tools from other sources instead of the daily PPA by providing UACLIENT_BEHAVE_INSTALL_FROM to the behave test runner. The default is UACLIENT_BEHAVE_INSTALL_FROM=daily, and the other available options are:

  • staging: install from the staging PPA
  • stable: install from the stable PPA
  • archive: install the latest version available in the archive, not adding any PPA
  • proposed: install the package from the -proposed pocket - specially useful for SRU testing (see the release guide)
  • custom: install from a custom provided PPA. If set, then two other variables need to be set: UACLIENT_BEHAVE_CUSTOM_PPA=<PPA URL> and UACLIENT_BEHAVE_CUSTOM_PPA_KEYID=<signing key for the PPA>.
  • local: install from a local copy of the ubuntu-advantage-client source code

local is particularly useful, as it runs the suite against the local code, thus including and validating the latest changes made. It is advised to run any related integration tests against local code changes before pushing them to be reviewed.

Note Note that we cache the source when running with UACLIENT_BEHAVE_INSTALL_FROM=local based on a hash, calculated from the repository state. If you change the python code locally and run the behave tests against your new version, there will be new debs in the cache source with the new repo state hash.

To run the tests, you can use tox:

tox -e behave-lxd-20.04

or, if you just want to run a specific file, or a test within a file:

tox -e behave-lxd-20.04 features/unattached_commands.feature
tox -e behave-lxd-20.04 features/unattached_commands.feature:28

As can be seen, this will run behave tests only for release 20.04 (Focal Fossa). We are currently supporting 5 distinct releases:

  • 21.10 (Kinetic Kudu)
  • 22.04 (Jammy Jellyfish)
  • 20.04 (Focal Fossa)
  • 18.04 (Bionic Beaver)
  • 16.04 (Xenial Xerus)

Therefore, to change which release to run the behave tests against, just change the release version on the behave command.

Furthermore, when developing/debugging a new scenario:

  1. Add a @wip tag decorator on the scenario
  2. To only run @wip scenarios run: tox -e behave-lxd-20.04 -- -w
  3. If you want to use a debugger:
    1. Add ipdb to integration-requirements.txt
    2. Add ipdb.set_trace() in the code block you wish to debug

(If you're getting started with behave, we recommend at least reading through the behave tutorial to get an idea of how it works, and how tests are written.)

Iterating Locally

To make running the tests repeatedly less time-intensive, our behave testing setup has support for reusing images between runs via two configuration options (provided in environment variables), UACLIENT_BEHAVE_IMAGE_CLEAN and UACLIENT_BEHAVE_REUSE_IMAGE.

To avoid the test framework cleaning up the image it creates, you can run it like this:

UACLIENT_BEHAVE_IMAGE_CLEAN=0 tox -e behave

which will emit a line like this above the test summary:

Image cleanup disabled, not deleting: behave-image-1572443113978755

You can then reuse that image by plugging its name into your next test run, like so:

UACLIENT_BEHAVE_REUSE_IMAGE=behave-image-1572443113978755 tox -e behave

If you've done this correctly, you should see something like reuse_image = behave-image-1572443113978755 in the "Config options" output, and test execution should start immediately (without the usual image build step).

(Note that this handling is specific to our behave tests as it's performed in features/environment.py, so don't expect to find documentation about it outside of this codebase.)

Optimizing total run time of integration tests with snapshots

When UACLIENT_BEHAVE_SNAPSHOT_STRATEGY=1 we create a snapshot of an instance with ubuntu-advantage-tools installed and restore from that snapshot for all tests. This adds an upfront cost that is amortized across several test scenarios.

Based on some rough testing in July 2021, these are the situations when you should set UACLIENT_BEHAVE_SNAPSHOT_STRATEGY=1

At time of writing, starting a lxd.vm instance from a local snapshot takes longer than starting a fresh lxd.vm instance and installing ua.

machine_type condition
lxd.container num_scenarios > 7
lxd.vm never
gcp num_scenarios > 5
azure num_scenarios > 14
aws num_scenarios > 11

Integration testing on EC2

The following tox environments allow for testing focal on EC2:

  # To test ubuntu-pro-images
  tox -e behave-awspro-20.04
  # To test Canonical cloud images (non-ubuntu-pro)
  tox -e behave-awsgeneric-20.04

To run the test for a different release, just update the release version string. For example, to run AWS pro xenial tests, you can run:

tox -e behave-awspro-16.04

In order to run EC2 tests, please set up the pycloudlib toml file with the required EC2 credentials.

To specifically run non-ubuntu pro tests using canonical cloud-images an additional token obtained from https://ubuntu.com/pro needs to be set:

  • UACLIENT_BEHAVE_CONTRACT_TOKEN=<your_token>

By default, the public AMIs for Ubuntu Pro testing used for each Ubuntu release are defined in features/aws-ids.yaml. These ami-ids are determined by running ./tools/refresh-aws-pro-ids.

Integration tests will read features/aws-ids.yaml to determine which default AMI id to use for each supported Ubuntu release.

To update features/aws-ids.yaml, run ./tools/refresh-aws-pro-ids and put up a pull request against this repo to updated that content from the ua-contracts marketplace definitions.

  • To manually run EC2 integration tests with a specific AMI Id provide the following environment variable to launch your specific AMI instead of building a daily ubuntu-advantage-tools image.
UACLIENT_BEHAVE_REUSE_IMAGE=your-custom-ami tox -e behave-awspro-20.04

Integration testing on Azure

The following tox environments allow for testing focal on Azure:

  # To test ubuntu-pro-images
  tox -e behave-azurepro-20.04
  # To test Canonical cloud images (non-ubuntu-pro)
  tox -e behave-azuregeneric-20.04

To run the test for a different release, just update the release version string. For example, to run Azure pro xenial tests, you can run:

tox -e behave-azurepro-16.04

In order to run Azure tests, please set up the pycloudlib toml file with the required Azure credentials.

To specifically run non-ubuntu pro tests using canonical cloud-images an additional token obtained from https://ubuntu.com/pro needs to be set:

  • UACLIENT_BEHAVE_CONTRACT_TOKEN=<your_token>
  • To manually run Azure integration tests with a specific Image Id provide the following environment variable to launch your specific Image Id instead of building a daily ubuntu-advantage-tools image.
UACLIENT_BEHAVE_REUSE_IMAGE=your-custom-image-id tox -e behave-azurepro-20.04

Integration testing on GCP

The following tox environments allow for testing focal on GCP:

  # To test ubuntu-pro-images
  tox -e behave-gcppro-20.04
  # To test Canonical cloud images (non-ubuntu-pro)
  tox -e behave-gcpgeneric-20.04

To run the test for a different release, just update the release version string. For example, to run GCP pro xenial tests, you can run:

tox -e behave-gcppro-16.04

In order to run GCP tests, please set up the pycloudlib toml file with the required GCP credentials.

To specifically run non-ubuntu pro tests using canonical cloud-images an additional token obtained from https://ubuntu.com/pro needs to be set:

  • UACLIENT_BEHAVE_CONTRACT_TOKEN=<your_token>
  • To manually run GCP integration tests with a specific Image Id provide the following environment variable to launch your specific Image Id instead of building a daily ubuntu-advantage-tools image.
UACLIENT_BEHAVE_REUSE_IMAGE=your-custom-image-id tox -e behave-gcppro-20.04