Running integration tests leaves orphaned processes

[c-pius]

time-boxed to 1d

Description

Running the integration tests seems to keep K8s-related processes running after they are finished (processes like kube-apiserver, etcd, ). When consecutively executing them for a larger amount of times, it eats up the machine resources and a restart is necessary.

Steps to reproduce

run make test for a couple of times

Environment Type

Managed

Environment Info

Kubernetes Version: 1.30.3
Docker Desktop Version: 4.29.0
macOS: 15.1.1
Command Line Tools for Xcode: 16.1

Attachments

[amritanshusikdar]

Potential Cause:
In our DualClusterFactory, we always start the skrEnv inside the Init() function, but we never actually call skrEnv.Stop().
In order to understand this better, have a look at our integration tests which call the NewDualClusterFactory() function to spin up a test SKR. In the respective suite_test.go AfterSuite you'll find a kcpEnv.Stop(), however, as mentioned above, there's no call to stop the SKR test environment.

This could be easily solved with some minor refactoring and code style changes. Some efforts have been made as part of #2324

[c-pius]

Probably we miss "closing statements like env.Stop()". Tomek has a suspicion that it is closed correctly on successful tests, but not on failing tests.