Details

What steps did you take and what happened:
After upgrading HelmRelease to 3.5.0 for amd-gpu-node-labeler with a pre-defined RBAC configuration (ClusterRole, ClusterRoleBinding, and ServiceAccount), the serviceAccount configuration did not seem to be honored which caused the pods in the Daemonset to use the Default ServiceAccount instead of the custom defined one which has the correct permissions to read the node.

    serviceAccount:
      create: false
      name: amd-gpu-node-labeller

What did you expect to happen:

The Pods should use the defined service account

Anything else you would like to add:

Reverting the app-template to 3.4.0 fixes the issue.

Additional Information: