clusterloader2 should be covered with unit tests

[oxddr]

Currently clusterloader2 has not tests at all. Testing new functionality requires spinning up a new cluster and running newly added code against it. It's costly and feedback loop is slow.

[oxddr]

/kind cleanup

[oxddr]

/area clusterloader

[fejta-bot]

Issues go stale after 90d of inactivity.
Mark the issue as fresh with /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.

If this issue is safe to close now please do so with /close.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/lifecycle stale

[wojtek-t]

/remove-lifecycle stale
/lifecycle frozen

[oxddr]

/help-wanted
/good-first-issue

This is a good opportunity to learn clusterloader2. Tests can be added package-by-package.

[k8s-ci-robot]

@oxddr:
This request has been marked as suitable for new contributors.

Please ensure the request meets the requirements listed here.

If this request no longer meets these requirements, the label can be removed
by commenting with the /remove-good-first-issue command.

In response to this:

/help-wanted
/good-first-issue

This is a good opportunity to learn clusterloader2. Tests can be added package-by-package.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

[rahulii]

I would like to give it a try.
/cc @oxddr , could you give some pointers to where should i start!

[marseel]

@rahulii I think it would make sense to start testing measurements based on prometheus.
In the past we had multiple issues when PromQL query was written incorrectly and we have quite a lot measurements, which are based on prometheus queries.
Here you can find some resources how PromQL queries can be tested (although I only glanced at it):
https://prometheus.io/docs/prometheus/latest/configuration/unit_testing_rules/#test-yml
https://zerokspot.com/weblog/2019/11/25/testing-prometheus-alerts/
https://github.com/prometheus/prometheus/blob/main/cmd/promtool/unittest.go#L155

For example, we could start with scheduling throughput measurement:
https://github.com/kubernetes/perf-tests/blob/master/clusterloader2/pkg/measurement/common/scheduling_throughput_prometheus.go

What we would like to do is check if for given query it returns correct result. This is a bit open-ended problem so I might not have all the answers.