Labeling Nodes for Selection
What you'll learn: How to add custom labels to your Kubernetes nodes and reference them in pod specifications to control where pods run.
The Core Idea
Now that you understand node selectors, you need to know how to prepare your nodes by giving them meaningful labels. Think of labels as tags you attach to nodes to describe their characteristics—like "has-gpu=true" or "environment=production"—so you can later use those tags to guide pod placement.
Adding Labels to Nodes
You use kubectl label to tag nodes with key-value pairs:
kubectl label nodes worker-node-1 disktype=ssd
kubectl label nodes worker-node-2 gpu=nvidia-t4
kubectl label nodes worker-node-3 environment=production
To verify your labels:
kubectl get nodes --show-labels
Or filter nodes by a specific label:
kubectl get nodes -l disktype=ssd
Using Node Labels in Pod Specs
Once your nodes are labeled, reference those labels in your pod's nodeSelector:
apiVersion: v1
kind: Pod
metadata:
name: gpu-workload
spec:
nodeSelector:
gpu: nvidia-t4
containers:
- name: training-container
image: ml-training:latest
This pod will only schedule on nodes with the label gpu=nvidia-t4.
Practical Analogy
Labeling nodes is like putting colored stickers on boxes in a warehouse. Later, when you need to store something temperature-sensitive, you can say "only use boxes with the blue 'refrigerated' sticker"—you don't need to know specific box IDs, just the characteristic you need.
Key Takeaway: Label your nodes with meaningful key-value pairs using kubectl label, then use those labels in nodeSelector to control pod placement based on node characteristics.