avoid creating nodescaleadjuster artefacts when disabled

[slaupster]

the Deployment already has this guard. It should be extended to include its IAM and service.

[romanbaron]

Thanks for suggesting this change, it makes a lot of sense.
The reason the validation test fails is because nodescaleadjuster ClusterRole is generated here:
https://github.com/NVIDIA/KAI-Scheduler/blob/main/Makefile#L65
And the generated version doesn't have the if guard that you added.

[slaupster]

Thanks for suggesting this change, it makes a lot of sense. The reason the validation test fails is because nodescaleadjuster ClusterRole is generated here: https://github.com/NVIDIA/KAI-Scheduler/blob/main/Makefile#L65 And the generated version doesn't have the if guard that you added.

I have to say I'm not following what the manifests target is intending to do - just generate what is missing ? If I just remove it like last commit we accept going forward none of it is generated because it has go template for the helm ?

[romanbaron]

All ClusterRoles yaml files are generated from kubebuilder annotations, it assists with keeping track of where the resources in the ClusterRole are actually being used. Here is an example:

KAI-Scheduler/pkg/podgrouper/pod_controller.go

Line 57 in 2387592

// +kubebuilder:rbac:groups="",resources=configmaps,verbs=get;list;watch

So these files cannot be manually modified. In addition 'make validate' step in CI attempts to catch a possible case where someone added a kubebuilder annotation but did not re-generate the yaml files.
Usually yaml files can be patched with Kustomize (example: Kustomize) but I am not sure if there is a way to wrap the entire file with a condition using this method.