Per-node configuration

Note

This is a beta feature. Please provide feedback and file a GitHub issue if you experience any problems.

The Cilium agent process (a.k.a. DaemonSet) supports setting configuration on a per-node basis. This allows overriding cilium-config ConfigMap for a node or set of nodes. It is managed by CiliumNodeConfig objects.

This feature is useful for:

CiliumNodeConfig objects

A CiliumNodeConfig object allows for overriding ConfigMap / Agent arguments. It consists of a set of fields and a label selector. The label selector defines to which nodes the configuration applies. As is the standard with Kubernetes, an empty LabelSelector (e.g. {}) selects all nodes.

Note

Creating or modifying a CiliumNodeConfig will not cause changes to take effect until pods are deleted and re-created (or their node is restarted).

Example: selective XDP enablement

To enable LoadBalancer & NodePort XDP Acceleration only on nodes with necessary hardware, one would label the relevant nodes and override their configuration.

apiVersion: cilium.io/v2alpha1
kind: CiliumNodeConfig
metadata:
  namespace: kube-system
  name: enable-xdp
spec:
  nodeSelector:
    matchLabels:
      io.cilium.xdp-offload: "true"
  defaults:
    bpf-lb-acceleration: native

Example: KubeProxyReplacement Rollout

To roll out kube-proxy replacement in a gradual manner, you may also wish to use the CiliumNodeConfig feature. This will label all migrated nodes with io.cilium.migration/kube-proxy-replacement: true

Warning

You must have installed Cilium with the Helm values k8sServiceHost and k8sServicePort. Otherwise Cilium will not be able to reach the Kubernetes APIServer after kube-proxy is uninstalled.

You can apply these two values to a running cluster via helm upgrade.

  1. Patch kube-proxy to only run on unmigrated nodes.

    kubectl -n kube-system patch daemonset kube-proxy --patch '{"spec": {"template": {"spec": {"affinity": {"nodeAffinity": {"requiredDuringSchedulingIgnoredDuringExecution": {"nodeSelectorTerms": [{"matchExpressions": [{"key": "io.cilium.migration/kube-proxy-replacement", "operator": "NotIn", "values": ["true"]}]}]}}}}}}}'
    
  2. Configure Cilium to use kube-proxy replacement on migrated nodes

    cat <<EOF | kubectl apply --server-side -f -
    apiVersion: cilium.io/v2alpha1
    kind: CiliumNodeConfig
    metadata:
      namespace: kube-system
      name: kube-proxy-replacement
    spec:
      nodeSelector:
        matchLabels:
          io.cilium.migration/kube-proxy-replacement: true
      defaults:
        kube-proxy-replacement: true
        kube-proxy-replacement-healthz-bind-address: "0.0.0.0:10256"
    
    EOF
    
  3. Select a node to migrate. Optionally, cordon and drain that node:

    export NODE=kind-worker
    kubectl label node $NODE --overwrite 'io.cilium.migration/kube-proxy-replacement=true'
    kubectl cordon $NODE
    
  4. Delete Cilium DaemonSet to reload configuration:

    kubectl -n kube-system delete pod -l k8s-app=cilium --field-selector spec.nodeName=$NODE
    
  5. Ensure Cilium has the correct configuration:

    kubectl -n kube-system exec $(kubectl -n kube-system get pod -l k8s-app=cilium --field-selector spec.nodeName=$NODE -o name) -c cilium-agent -- \
        cilium config get kube-proxy-replacement
    true
    
  6. Uncordon node

    kubectl uncordon $NODE
    
  7. Cleanup: set default to kube-proxy-replacement:

    cilium config set --restart=false kube-proxy-replacement true
    cilium config set --restart=false kube-proxy-replacement-healthz-bind-address "0.0.0.0:10256"
    kubectl -n kube-system delete ciliumnodeconfig kube-proxy-replacement
    
  8. Cleanup: delete kube-proxy daemonset, unlabel nodes

    kubectl -n kube-system delete daemonset kube-proxy
    kubectl label node --all --overwrite 'io.cilium.migration/kube-proxy-replacement-'