Node Selectors

A nodeSelector is the simplest way to direct a Pod to a specific group of nodes. You add labels to nodes, then add a nodeSelector to the Pod spec that matches those labels. Only nodes with all the required labels can receive the Pod.

You already saw nodeSelector used in the DaemonSet scheduling lesson to restrict a DaemonSet to SSD nodes. The mechanism is identical for regular Pods.

Labeling nodes

kubectl get nodes --show-labels

Nodes already have default labels: kubernetes.io/hostname, kubernetes.io/os, and kubernetes.io/arch among others. You can add custom labels for your own scheduling needs.

kubectl label node sim-worker disk=ssd
kubectl label node sim-worker2 disk=hdd

Verify:

kubectl get nodes -L disk

The -L disk flag adds the disk label as a column in the output. Both nodes appear with their respective values.

Using nodeSelector in a Pod

nano ssd-pod.yaml

apiVersion: v1
kind: Pod
metadata:
  name: ssd-pod
spec:
  nodeSelector:
    disk: ssd
  containers:
    - name: app
      image: busybox:1.36
      command: ['sh', '-c', 'sleep 3600']

kubectl apply -f ssd-pod.yaml
kubectl get pod ssd-pod -o wide

The NODE column shows sim-worker. The Pod was scheduled on the SSD node.

kubectl describe pod ssd-pod

Under Node-Selectors, you see disk=ssd. If no node has this label, the Events section would show FailedScheduling: 0 nodes matched node selector.

Selecting by built-in labels

Built-in labels are useful for OS or architecture requirements:

nodeSelector:
  kubernetes.io/os: linux
  kubernetes.io/arch: amd64

This schedules the Pod only on Linux nodes with AMD64 architecture. Useful for clusters with mixed OS or architecture (Raspberry Pi nodes with arm64, for example).

The limitation of nodeSelector

nodeSelector only supports equality matching: key=value. You cannot express:

• “schedule on nodes with disk=ssd OR disk=nvme”
• “prefer nodes labeled region=eu-west but allow other regions if needed”
• “schedule only if the node has at least 8 CPUs AND is in zone A”

For any scheduling requirement more complex than exact label matching, use nodeAffinity, covered in the next lesson.

kubectl delete pod ssd-pod
kubectl label node sim-worker disk-
kubectl label node sim-worker2 disk-

nodeSelector is the simplest node targeting tool. Use it when you need exact label matching. For soft preferences, OR conditions, or range expressions, nodeAffinity provides the flexibility you need.