Skip to content

Latest commit

 

History

History
246 lines (194 loc) · 5.15 KB

README.md

File metadata and controls

246 lines (194 loc) · 5.15 KB
Raw

Kubernetes Lab: Scheduling with Taints, Tolerations, Node Affinity, and Pod Anti-Affinity

This lab demonstrates how Kubernetes uses taints, tolerations, node affinity, and pod anti-affinity to control pod scheduling.

Prerequisites

  1. A running Kubernetes cluster (e.g., Minikube with two nodes: master and worker).
  2. kubectl installed and configured.

Cluster Setup

Step 1: Start Minikube with Two Nodes

  1. Start Minikube:

    minikube start --driver=docker --cni=cilium --kubernetes-version=stable --extra-config=kubelet.authentication-token-webhook=true --extra-config=kubelet.authorization-mode=AlwaysAllow --extra-config=kubelet.cgroup-driver=systemd --extra-config=kubelet.read-only-port=10255 --insecure-registry="registry.k8s.io"

  2. Add a worker node:

    minikube node add

  3. Verify nodes:

    kubectl get nodes

You should see:

NAME           STATUS   ROLES                  AGE   VERSION
minikube       Ready    control-plane,master   Xs    v1.XX.X
minikube-m02   Ready    <none>                 Xs    v1.XX.X

Taint the Master Node

To prevent non-system pods from being scheduled on the master node:

  1. Add a taint to the master node:

    kubectl taint nodes minikube node-role.kubernetes.io/master:NoSchedule

  2. Verify the taint:

    kubectl describe node minikube | grep Taint

Lab Objectives

  1. Understand and configure taints and tolerations.
  2. Learn how to use node affinity to schedule pods on specific nodes.
  3. Use pod anti-affinity to distribute pods across nodes.

1. Taints and Tolerations

Step 1: Taint the Worker Node

Taint the worker node to prevent pods from being scheduled unless they tolerate the taint:

kubectl taint nodes minikube-m02 key1=value1:NoSchedule

Verify the taint:

kubectl describe node minikube-m02 | grep Taint

Step 2: Deploy a Pod Without Tolerations

View Pod YAML 
apiVersion: v1
kind: Pod
metadata:
  name: no-toleration-pod
spec:
  containers:
  - name: nginx
    image: nginx

Apply the pod:

kubectl apply -f no-toleration-pod.yaml

The pod will be Pending because it cannot be scheduled on the tainted worker node or the master node (master is also tainted).

Step 3: Deploy a Pod with Tolerations

View Pod YAML 
apiVersion: v1
kind: Pod
metadata:
  name: toleration-pod
spec:
  tolerations:
  - key: "key1"
    operator: "Equal"
    value: "value1"
    effect: "NoSchedule"
  containers:
  - name: nginx
    image: nginx

Apply the pod:

kubectl apply -f toleration-pod.yaml

The pod should now run on the tainted worker node.

2. Node Affinity

Step 1: Label the Worker Node

Label the worker node:

kubectl label nodes minikube-m02 disktype=ssd

Verify the label:

kubectl get nodes --show-labels

Step 2: Deploy a Pod with Node Affinity

View Pod YAML 
apiVersion: v1
kind: Pod
metadata:
  name: node-affinity-pod
spec:
  affinity:
    nodeAffinity:
      requiredDuringSchedulingIgnoredDuringExecution:
        nodeSelectorTerms:
        - matchExpressions:
          - key: disktype
            operator: In
            values:
            - ssd
  containers:
  - name: nginx
    image: nginx

Apply the pod:

kubectl apply -f node-affinity-pod.yaml

The pod should be scheduled on the worker node.

3. Pod Anti-Affinity

Step 1: Deploy a Pod with Anti-Affinity

View Deployment YAML 
apiVersion: apps/v1
kind: Deployment
metadata:
  name: anti-affinity-app
spec:
  replicas: 2
  selector:
    matchLabels:
      app: anti-affinity
  template:
    metadata:
      labels:
        app: anti-affinity
    spec:
      affinity:
        podAntiAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
          - labelSelector:
              matchLabels:
                app: anti-affinity
            topologyKey: "kubernetes.io/hostname"
      containers:
      - name: nginx
        image: nginx

Apply the deployment:

kubectl apply -f anti-affinity-deployment.yaml

Step 2: Verify Pod Distribution

Check where the pods are scheduled:

kubectl get pods -o wide

The pods should be distributed across the nodes.

Cleanup

Remove all resources created during the lab:

kubectl delete pod no-toleration-pod toleration-pod node-affinity-pod
kubectl delete deployment anti-affinity-app
kubectl taint nodes minikube-m02 key1=value1:NoSchedule-
kubectl label nodes minikube-m02 disktype-
kubectl taint nodes minikube node-role.kubernetes.io/master:NoSchedule-

Lab Summary

  1. Taints and Tolerations: Prevent pods from being scheduled on nodes unless they tolerate the taints.
  2. Node Affinity: Schedule pods on specific nodes based on labels.
  3. Pod Anti-Affinity: Distribute pods across nodes to avoid co-locating them.

This lab demonstrates how Kubernetes uses scheduling controls to manage workload placement.