Helm installation does not reflect helm's --namespace argument, but hardcodes kube-system (or rather hides it in the values.yaml)

[kastl-ars]

Describe the bug
Normally one can install a helm release into a discrete namespace by using helm install --namespace foo ....

This does not work with the retina chart, as it puts the helm release into that namespace, but the resources into the kube-system namespace.

e.g. https://github.com/microsoft/retina/blob/main/deploy/legacy/manifests/controller/helm/retina/values.yaml#L47

If there is a technical reason for doing so, this should be documented. If not, this should be fixed to get the usual behaviour.

To Reproduce
Steps to reproduce the behavior:

1. Install using the following command:

helm upgrade --install retina oci://ghcr.io/microsoft/retina/charts/retina \
    --namespace retina \
    --create-namespace \
    --version $VERSION \
    --namespace kube-system \
    --set image.tag=$VERSION \
    --set operator.tag=$VERSION \
    --set logLevel=info \
    --set enabledPlugin_linux="\[dropreason\,packetforward\,linuxutil\,dns\]"

3. Issue helm ls -n retina and find the release there, but no pods or daemonsets (as kubectl get pod,ds -n retina shows)
4. Check for daemonsets in the kube-system namespace using kubectl get pod,ds -n kube-system
5. Voilá!

Expected behavior
The chart should respect helm's --namespace argument

Platform (please complete the following information):

• OS: openSUSE Leap 15.6
• Kubernetes Version: v1.30.4+k3s1
• Host: self-hosted VM
• Retina Version: 0.0.16

[kastl-ars]

In addition, the operator daemonset always gets deploy to kube-system, even if its configMap is created in e.g. the retina namespace.

https://github.com/microsoft/retina/blob/main/deploy/legacy/manifests/controller/helm/retina/templates/operator.yaml#L6

[BeegiiK]

@kastl-ars

Thank you for creating an issue. During my investigation, there's two things to mention:

• The operator was always going to get deployed to kube-system. Now the daemonset can get deployed in any namepsace provided by the user.
• The commands you provided needs to be updated. The --namespace option will get overrided if the namespace chart value exists (which in this case does) and that is why you'll never see it deployed elsewhere. You'll need to use the --set namespace=X option to override the value instead. I've tested this and it works!

[kastl-ars]

Thanks for checking @BeegiiK.

If there is no technical reason to use the namespace value, I would recommend to remove it and use what most other charts are using: helm's --namespace argument...