Metric "kube_node_role" missing (AKS)

[R-Studio]

What happened:

• We installed the kube-prometheus-stack Helm chart 61.8.0 on our AKS cluster (1.29.6) with kube-state-metrics enabled.
• It looks like it works but the metric kube_node_role is missing because the label node-role.kubernetes.io/* is missing.
• In AKS node-role.kubernetes.io/* is deprecated and the recommended substitute is kubernetes.azure.com/role=* (more in the docs)
• Unfortunately we are not able and it is not allowed because the following prefixes are AKS reserved prefixes and can't be used for any node (docs):
  • kubernetes.azure.com/
  • kubernetes.io/

What you expected to happen:

• The metric kube_node_role is available.

How to reproduce it (as minimally and precisely as possible):

• Install the kube-prometheus-stack Helm chart 61.8.0 on our AKS cluster (1.29.6) (no special values).
• Create a port-forward to the kube-state-metrics pod and open http://localhost:8080/metrics.
• Search the metric kube_node_role and you can see there is only the description and help of the metric but no metric data:

  ...
  # HELP kube_node_role The role of a cluster node.
  # TYPE kube_node_role gauge
  ...
  

Anything else we need to know?:

• It looks like we are not able to tell kube-state-metrics to use the label kubernetes.azure.com/role=* instead of node-role.kubernetes.io/*, because it is hardcoded.
• For us the metric kube_node_role is very important, because we have a lot of recording- and alertingrules that uses it.

Environment:

• kube-state-metrics version: v2.13.0
• Kubernetes version (use kubectl version): 1.29.6
• Cloud provider or hardware configuration: AKS

[R-Studio]

Workaround / Solution

It looks like, it is not allowed to set node-role.kubernetes.io/agent* but you can set any other role if you want.
For example we set the following labels on the two node pools:

• Control-Plane Nodes = node-role.kubernetes.io/control-plane: 'true'
• Worker Nodes = node-role.kubernetes.io/worker: 'true'

[R-Studio]

Workaround not working!

Unfortunately our workaround is not really working, because we cannot set the labels on the AKS node pools. Here the answer of Microsoft support:
The error message you're seeing indicates that you're trying to set a label on a Kubernetes node pool using a key that has a prefix reserved for Kubernetes' internal use. Labels with prefixes like kubernetes.io/ are reserved for use by Kubernetes itself and should not be used for custom labels.

[dgrisonnet]

/triage accepted
/assign

[abossard]

I opened a related issue in the AKS project: Azure/AKS#4628

[R-Studio]

@abossard any news?