Node affinity and anti-affinity in Kubernetes - Mini Project: Build & Apply

Node affinity and anti-affinity

📖 Scenario: You are managing a Kubernetes cluster for a small company. You want to control where your application pods run by using node affinity and anti-affinity rules. This helps keep your app stable and balanced across the cluster.

🎯 Goal: Learn how to write a Kubernetes pod specification that uses node affinity and anti-affinity to schedule pods on specific nodes or avoid certain nodes.

📋 What You'll Learn

Create a pod manifest with a basic container

Add node affinity rules to prefer nodes with a specific label

Add node anti-affinity rules to avoid nodes with another label

Display the final pod manifest YAML

💡 Why This Matters

🌍 Real World

Node affinity and anti-affinity help control where pods run in a Kubernetes cluster. This is useful to improve performance, reliability, and resource use by placing pods on suitable nodes or avoiding problematic ones.

💼 Career

Understanding node affinity and anti-affinity is important for Kubernetes administrators and DevOps engineers to manage cluster workloads effectively and ensure application stability.

1

Create a basic pod manifest

Create a Kubernetes pod manifest named pod.yaml with a pod named myapp-pod that runs the container image nginx:latest. Include the apiVersion, kind, metadata, and spec sections with the container definition.

Kubernetes

# Create the basic pod manifest with name 'myapp-pod' and container image 'nginx:latest'
# Your code here

Need a hint?

Start with the basic pod YAML structure. Use apiVersion: v1 and kind: Pod. Add metadata with the pod name and a spec with one container running nginx.

2

Add node affinity to prefer nodes with label

Add a nodeAffinity section under spec to prefer scheduling the pod on nodes with the label disktype: ssd. Use preferredDuringSchedulingIgnoredDuringExecution with a weight of 1 and a match expression for the label key disktype and value ssd.

Kubernetes

apiVersion: v1
kind: Pod
metadata:
  name: myapp-pod
spec:
  containers:
  - name: nginx-container
    image: nginx:latest
  # Add nodeAffinity to prefer nodes with label disktype=ssd
  # Your code here

Need a hint?

Under spec, add affinity with nodeAffinity. Use preferredDuringSchedulingIgnoredDuringExecution with a weight and a match expression for the label disktype: ssd.

3

Add node anti-affinity to avoid nodes with label

Add a nodeAffinity requiredDuringSchedulingIgnoredDuringExecution rule to avoid scheduling the pod on nodes with the label zone: east. Use a matchExpressions with key zone, operator NotIn, and value east.

Kubernetes

apiVersion: v1
kind: Pod
metadata:
  name: myapp-pod
spec:
  containers:
  - name: nginx-container
    image: nginx:latest
  affinity:
    nodeAffinity:
      preferredDuringSchedulingIgnoredDuringExecution:
      - weight: 1
        preference:
          matchExpressions:
          - key: disktype
            operator: In
            values:
            - ssd
      # Add requiredDuringSchedulingIgnoredDuringExecution to avoid nodes with label zone=east
      # Your code here

Need a hint?

Under nodeAffinity, add requiredDuringSchedulingIgnoredDuringExecution with a nodeSelectorTerms list. Use a matchExpressions to exclude nodes labeled zone: east.

4

Display the final pod manifest

Print the complete pod manifest YAML to the console exactly as written in the previous steps.

Kubernetes

apiVersion: v1
kind: Pod
metadata:
  name: myapp-pod
spec:
  containers:
  - name: nginx-container
    image: nginx:latest
  affinity:
    nodeAffinity:
      preferredDuringSchedulingIgnoredDuringExecution:
      - weight: 1
        preference:
          matchExpressions:
          - key: disktype
            operator: In
            values:
            - ssd
      requiredDuringSchedulingIgnoredDuringExecution:
        nodeSelectorTerms:
        - matchExpressions:
          - key: zone
            operator: NotIn
            values:
            - east

# Print the pod manifest YAML below
# Your code here

Need a hint?

Use a print statement to output the full pod manifest YAML exactly as defined.