Amazon EKS — Fargate — Practice Exercises

2 minute read

Chapter 6 — Practice Exercises

Exercise 1: Fargate vs EC2 Decision

Question: You’re running a Kubernetes CronJob that processes data every hour for 15 minutes, using 1 vCPU and 2 GB memory. Compare the monthly cost on Fargate vs a dedicated t3.medium EC2 node.

Show Solution

Fargate cost:

Runtime: 15 min/hr × 24 hrs × 30 days = 10,800 min = 180 hours
Cost: (1 × $0.04048 + 2 × $0.004445) × 180 = $0.04937 × 180 = $8.89/month

EC2 (t3.medium, 24/7):

Cost: $0.0416/hr × 730 hrs = $30.37/month

Fargate is ~71% cheaper because you only pay for the 180 hours the job actually runs, not the 730 hours the EC2 node sits idle.

Exercise 2: Fargate Profile Design

Question: Design Fargate profiles for: production web app (should NOT go to Fargate), batch jobs (should go to Fargate), and dev environment (should go to Fargate).

Show Solution 
fargateProfiles:
  - name: batch
    selectors:
      - namespace: batch
        labels:
          compute: fargate

  - name: dev
    selectors:
      - namespace: dev

# Production web app — NO Fargate profile
# It runs in the 'production' namespace
# No matching profile = scheduled on EC2 nodes

Key insight: Fargate is opt-in. Only pods matching a profile run on Fargate. The production web app stays on EC2 because no profile matches the production namespace.

Exercise 3: DaemonSet Alternative

Question: You need to collect logs from all pods using Fluent Bit. On EC2 nodes, you’d use a DaemonSet. How do you achieve this on Fargate?

Show Solution

Use Fargate’s built-in log router or sidecar containers:

Option 1: Fargate built-in log router (FireLens):

apiVersion: v1
kind: Pod
metadata:
  annotations:
    eks.amazonaws.com/logging: enabled
spec:
  containers:
    - name: app
      image: my-app:latest

Option 2: Sidecar container:

spec:
  containers:
    - name: app
      image: my-app:latest
      volumeMounts:
        - name: shared-logs
          mountPath: /var/log/app
    - name: fluent-bit
      image: fluent/fluent-bit:latest
      volumeMounts:
        - name: shared-logs
          mountPath: /var/log/app
          readOnly: true
  volumes:
    - name: shared-logs
      emptyDir: {}

Exercise 4: Fargate Limitations

Question: Your team wants to run a PostgreSQL database on EKS with persistent EBS storage and a monitoring DaemonSet. Can this run on Fargate? If not, what’s the alternative?

Show Solution

No, this cannot run on Fargate. Two blockers:

  1. No EBS volumes — Fargate doesn’t support EBS. PostgreSQL needs persistent block storage.
  2. No DaemonSets — Monitoring agents can’t run as DaemonSets on Fargate.

Alternative: Mixed mode

  • Run PostgreSQL on an EC2 managed node group with EBS storage
  • Run monitoring as a DaemonSet on EC2 nodes
  • Run stateless application pods on Fargate

Exercise 5: Cost Optimization

Question: Your Fargate bill is $500/month for 20 pods running 24/7 at 1 vCPU, 2 GB each. How can you reduce costs?

Show Solution

Current Fargate cost:

20 pods × $0.04937/hr × 730 hrs = $720.82/month
(My estimate: $36/pod × 20 = ~$720)

Option 1: Switch steady workloads to EC2

3× m5.large (fit ~10 pods each) = 3 × $70 = $210/month
EKS cluster fee (already paying): $73
Total: ~$283/month (61% savings)

Option 2: Use Spot EC2 nodes

3× m5.large Spot = 3 × $25 = $75/month
Total: ~$148/month (79% savings)

Option 3: Fargate Savings Plans (1-year commit)

Up to 50% discount on Fargate
New cost: ~$360/month

Recommendation: Move 24/7 steady workloads to EC2 Spot. Keep Fargate only for bursty/intermittent workloads.

What’s Next?

On to networking: Chapter 7 — Networking.

You may also enjoy