Switching Colors in Production: Blue/Green Deployments with ECS + Terraform

Blue/Green deployments have always felt like a magic trick — two identical environments running in parallel, one live, one in waiting, and with a single switch you move production traffic over without anyone noticing. This month, I implemented exactly that for our FeedbackHub application on AWS ECS Fargate using Terraform.

This is Phase 3 of my AWS-Learning roadmap — and it’s a milestone worth documenting.

---

🎯 Why Blue/Green?

• Zero downtime: Swap environments without taking production offline.

• Safe rollback: If the new version (Green) misbehaves, instantly flip back to Blue.

• Confidence in release: Test Green in production without risking the main environment.

Before Blue/Green became common, around the 2015–2020 era, we used to create a full replica of production, deploy new code there, run quick smoke tests, and then switch the DNS. It worked, but it was clunky, risky, and stressful. Blue/Green eliminates that hassle.

That’s why I often say: “Before DevOps was a buzzword, I was already scaling Drupal on EC2s. Now I’m just rewriting it with the help of modern tools and techniques!”

---

🏗 Architecture Overview (Pyramid View)

The setup:

• ALB at the top routing all production traffic

• Two Target Groups (Blue & Green) beneath it

• Two ECS Services (Blue active, Green staging) beneath Target Groups

---

🔍 Visual: Pyramid Flow

Legend:

• User Requests → Incoming traffic from users

• ALB Listener → Routes traffic based on weights

• Blue Target Group → Current production service (receives majority traffic)

• Green Target Group → New version under test (receives minority traffic)

• Rollback Path → Instant switch back to Blue if needed

• Promote to Prod → Switch Green to full production when stable

---

⚙️ Implementation Highlights

• Terraform modules for clean separation:

  • alb → ALB, TGs, listener rules

  • ecs/feedbackhub_service → Blue service

  • ecs/feedbackhub_service_green → Green service

• Weighted rule:

forward {
  target_group {
    arn    = aws_lb_target_group.app.arn # Blue
    weight = 90
  }
  target_group {
    arn    = aws_lb_target_group.feedbackhub_green_tg.arn # Green
    weight = 10
  }
}

• Rollout strategy: Shift gradually, monitor CloudWatch logs, rollback if needed

---

🧪 Testing the Cutover

1. Deploy new version to Green

2. Gradually increase traffic weight to Green

3. Validate health checks & logs in CloudWatch

4. If stable → shift 100% traffic to Green

5. If issues → rollback instantly to Blue

Fun fact: Our /green test URL had some Next.js quirks (redirects), so we validated Green by routing traffic through the ALB weighted rules instead — a very real production lesson in handling frontend routing.

---

📚 Lessons Learned

• Blue/Green is less about URLs and more about traffic control at ALB level

• Weighted rules are a powerful safety net

• Clean Terraform modules make this setup repeatable and maintainable

• Always have a rollback command ready

---

🚀 What’s Next

This completes Phase 3 of my AWS-Learning roadmap. Next up:

• Phase 3.5 → AWS Bedrock integration to analyze ECS CloudWatch logs (AI-powered observability!)

• Phase 4 → CI/CD pipeline for Blue/Green deployments using GitHub Actions

Stay tuned — we’re bringing AI into the mix next.

---

💬 Have you implemented Blue/Green with ECS or another platform? What was your biggest lesson?

🔗 Resources & Links

💻 Repo: GitHub Repository

🌐 Live App: FeedbackHub Deployment (May be snoozing to save AWS bill!)

🔗 Connect: LinkedIn Profile (Come for the DevOps talk, stay for the ECS nap jokes!)