[Bug]: Deprecated `cdi.default` CRD default causes persistent Argo CD drift

[dentiny]

Describe the bug
Hi team, I found there's inconsistency between cluster policy definition and helm-rendered cluster policy, which leads to ArgoCD keep reporting out-of-sync.

The issue is

The deprecated spec.cdi.default field is no longer rendered by the ClusterPolicy Helm template.
The rendered resource therefore contains:

spec:
  cdi:
    enabled: true

However, the ClusterPolicy CRD still defines spec.cdi.default with a schema default of false.
The corresponding Go API definition also marks the field as deprecated and unused while retaining +kubebuilder:default=false.
After the rendered resource is applied, the Kubernetes API server adds the default:

spec:
  cdi:
    enabled: true
    default: false

This creates a persistent difference:

Desired: spec.cdi.default is absent
Live:    spec.cdi.default is false

Argo CD reports the ClusterPolicy as OutOfSync. Re-syncing cannot resolve the difference because the API server adds the field again after every apply.

To Reproduce
N/A

Expected behavior
These two files are kept in sync.

Environment (please provide the following information):

• GPU Operator Version: v26.3.3
• OS: N/A, unrelated
• Kernel Version: N/A, unrelated
• Container Runtime Version: N/A, unrelated
• Kubernetes Distro and Version: N/A, unrelated

Information to attach (optional if deemed irrelevant)

• kubernetes pods status: kubectl get pods -n OPERATOR_NAMESPACE
• kubernetes daemonset status: kubectl get ds -n OPERATOR_NAMESPACE
• If a pod/ds is in an error state or pending state kubectl describe pod -n OPERATOR_NAMESPACE POD_NAME
• If a pod/ds is in an error state or pending state kubectl logs -n OPERATOR_NAMESPACE POD_NAME --all-containers
• Output from running nvidia-smi from the driver container: kubectl exec DRIVER_POD_NAME -n OPERATOR_NAMESPACE -c nvidia-driver-ctr -- nvidia-smi
• containerd logs journalctl -u containerd > containerd.log

Collecting full debug bundle (optional):

N/A

[tariq1890]

@dentiny This seems to be more of an ArgoCD issue rather than a GPU Operator issue. I am not sure I agree that the gpu-operator CRD should be modified (removing a default property is a non-additive change) to accommodate an ArgoCD behaviour.

I believe this issue highlights the same problem you're facing and it is solved by enabling the server-side diff strategy in your ArgoCD setup. Docs linked here.

[tariq1890]

@dentiny Can you confirm the ArgoCD version you were using when you faced this issue?