Amazon EKS — Storage — Practice Exercises

3 minute read

Chapter 8 — Practice Exercises

Exercise 1: EBS vs EFS Decision

Question: You’re deploying a WordPress application on EKS with 3 replicas that all need to access the same uploaded media files. Should you use EBS or EFS? Explain your reasoning.

Show Solution

Use EFS.

Reasoning:

  • WordPress uploads go to /wp-content/uploads/
  • 3 replicas need to read AND write to the same directory
  • EBS is ReadWriteOnce — only one pod can mount it
  • EFS is ReadWriteMany — all 3 pods can read/write
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: wp-uploads
spec:
  accessModes:
    - ReadWriteMany      # All pods access same data
  storageClassName: efs-sc
  resources:
    requests:
      storage: 50Gi

Cost consideration: EFS is 3x more expensive than EBS, but it’s the only option for shared access across multiple pods.

Exercise 2: StorageClass Design

Question: Create a StorageClass for a production PostgreSQL database that needs: encrypted volumes, high IOPS (50,000), data preserved after PVC deletion, and proper AZ alignment.

Show Solution 
apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: postgres-storage
provisioner: ebs.csi.aws.com
parameters:
  type: io2
  iops: "50000"
  encrypted: "true"
  fsType: ext4
reclaimPolicy: Retain               # Data preserved!
volumeBindingMode: WaitForFirstConsumer  # AZ alignment!
allowVolumeExpansion: true          # Can resize later

Key decisions:

  • io2 for consistent high IOPS (gp3 maxes at 16,000)
  • Retain so accidental PVC deletion doesn’t destroy data
  • WaitForFirstConsumer prevents AZ mismatch
  • allowVolumeExpansion for future growth

Exercise 3: StatefulSet

Question: Write a StatefulSet for a 3-replica Redis cluster where each replica needs its own 10Gi persistent volume. Include the headless Service.

Show Solution 
apiVersion: v1
kind: Service
metadata:
  name: redis
spec:
  clusterIP: None
  selector:
    app: redis
  ports:
    - port: 6379
---
apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: redis
spec:
  serviceName: redis
  replicas: 3
  selector:
    matchLabels:
      app: redis
  template:
    metadata:
      labels:
        app: redis
    spec:
      containers:
        - name: redis
          image: redis:7
          ports:
            - containerPort: 6379
          volumeMounts:
            - name: data
              mountPath: /data
  volumeClaimTemplates:
    - metadata:
        name: data
      spec:
        accessModes: ["ReadWriteOnce"]
        storageClassName: ebs-gp3
        resources:
          requests:
            storage: 10Gi

Creates: redis-0, redis-1, redis-2 with PVCs data-redis-0, data-redis-1, data-redis-2.

DNS: redis-0.redis.default.svc.cluster.local

Exercise 4: Troubleshoot PVC

Question: A PVC is stuck in Pending status. List 5 common causes and how to diagnose each.

Show Solution 
# First, check the PVC events
$ kubectl describe pvc <pvc-name>

5 common causes:

  1. No StorageClass exists: 
    $ kubectl get storageclass
# Fix: Create the referenced StorageClass
  2. EBS CSI driver not installed: 
    $ kubectl get pods -n kube-system -l app=ebs-csi-controller
# Fix: Install the EBS CSI driver addon
  3. IAM role missing permissions: 
    $ kubectl logs -n kube-system <ebs-csi-pod> -c ebs-plugin
# Look for: "AccessDenied" errors
  4. AZ mismatch (with Immediate binding): 
    $ kubectl describe pvc <name> | grep "waiting for first consumer"
# Fix: Use WaitForFirstConsumer binding mode
  5. Subnet has no nodes (WaitForFirstConsumer): 
    $ kubectl get nodes
# If no nodes in the AZ, PVC waits forever
# Fix: Ensure nodes exist in at least one AZ

Exercise 5: Storage Cost Optimization

Question: Your EKS cluster uses 50 × 100Gi gp3 EBS volumes at $0.08/GB/month = $400/month. 30 of these are for log storage pods that only need write performance. How would you reduce costs?

Show Solution

Option 1: Use st1 (throughput-optimized HDD) for log volumes:

30 × 100Gi × $0.045/GB = $135/month (vs $240)
Savings: $105/month (44%)

Option 2: Use EFS with Infrequent Access:

Hot data on EFS Standard: $0.30/GB
Old logs auto-tiered to IA: $0.016/GB
If 80% of logs are old: ~$0.07/GB effective

Option 3: Ship logs to S3 instead of storing on volumes:

S3 Standard: $0.023/GB/month
S3 IA: $0.0125/GB/month
Use Fluent Bit → S3 pipeline, no PVCs needed
Savings: ~$380/month (95%)

Recommended: Option 3 — ship logs to S3 via Fluent Bit, eliminate log PVCs entirely.

What’s Next?

Security: Chapter 9 — IAM & Security.

You may also enjoy