Create utility function for getting Kubernetes client

[pcieslinski]

Thanks for contributing to Prefect!

Please describe your work and make sure your PR:

• adds new tests (if appropriate)
• add a changelog entry in the changes/ directory (if appropriate)
• updates docstrings for any new functions or function arguments, including docs/outline.toml for API reference docs (if appropriate)

Note that your PR will not be reviewed unless all three boxes are checked.

What does this PR change?

While working on resolving this issue: #2281, I discovered a lot of duplicate code and tests. I managed to extract into a separate utility function the logic responsible for creating and authorizing the client for Kubernetes, which is used in the task library.

get_kubernetes_client was created in a similar way to get_boto_client. Thanks to this, I wanted to be consistent with the rest of the codebase.

Example of use:

from prefect.utilities.kubernetes import get_kubernetes_client

api_client = get_kubernetes_client(resource="pod", kubernetes_api_key_secret="KUBERNETES_API_KEY")

When the logic of creating and authorizing a client for Kubernetes was extracted into a separate function, some of the tests from the tasks were removed. Now all testing is done at the level of function get_kubernetes_client.

Why is this PR important?

Generally, this PR strives to stick to good software development practices and DRY principle. Thanks to this refactor we will have gathered in one place the logic to create and authorize the client for Kubernetes. We can use this part of the code throughout the codebase.

By removing duplicate tests, the project code will be easier to maintain.

[pcieslinski]

I missed that this api_client, which is used in Kubernetes tasks, is mocked in other tests for these tasks. I will fix these tests and everything should be fine. The logic of operation is the same, only the form of the code changed. ;)

[joshmeek]

Thanks for the big PR @pcieslinski! Yeah the client is mocked heavily for these tests and with this new utility could definitely be cleaned up / possibly separated into a test fixture.

[codecov]

Codecov Report

Merging #2845 into master will decrease coverage by 0.12%.
The diff coverage is 100.00%.

[pcieslinski]

@joshmeek I know that this PR has unfortunately grown quite a lot and can be difficult to check, but it is mainly because the same code was used in many places. I have improved the tests for all Kubernetes tasks and put the mocked Kubernetes client in the fixture. :)

[jcrist]

Thanks @pcieslinski, this looks good to me! Love to see that many deleted lines of code :).