Ignore control plane nodes

[aivanov-citc]

We have a cluster with 6 nodes, three of them are control plane nodes. We set NFD only on worker nodes.

When applying this rule,

apiVersion: nfd.k8s-sigs.io/v1alpha1
kind: NodeFeatureRule
metadata:
  name: paravirtualization-type
spec:
  rules:
    - name: "x86-paravirtualization-type-VMX"
      labels:
        "feature.node.kubernetes.io/x86-paravirtualization-type": "VMX"
      matchFeatures:
        - feature: cpu.cpuid
          matchExpressions:
            VMX: {op: Exists}
    - name: "x86-paravirtualization-type-SVM"
      labels:
        "feature.node.kubernetes.io/x86-paravirtualization-type": "SVM"
      matchFeatures:
        - feature: cpu.cpuid
          matchExpressions:
            SVM: {op: Exists}

we constantly see errors in the logs that the control planes do not have the necessary attributes.

Is there a way to filter nodes, for example, by labels or something else, so as not to read the attributes of unnecessary nodes?

[marquiz]

@aivanov-citc thanks for reporting this issue. I remember being annoyed by this myself, too, but been busy to do anything about it 🙄

I think this should be an easy fix, at least for the common usage scenario. Basically do not try to process the NodeFeatureRule(s) if a node does not have any NodeFeature(s). There might be some nice-to-have rework on the logging/error handling of the rules, too, but that basic thing should fix most of the problems.

EDIT: #1894