AWS Load Balancers: How Deregistration Delay Ensures Seamless Shutdowns

One Friday evening, Arjun got a page from his monitoring system.

“One of your EC2 instances is showing high error rates. Auto Scaling will replace it shortly.”

He sighed in relief—Auto Scaling would handle it.

But then... some users reported failed requests and broken sessions.

“Why did users get cut off mid-upload?” he asked himself.

That’s when Arjun learned about a quiet hero in AWS Load Balancers:
Connection Draining, also known as Deregistration Delay.

---

🧠 What Is Connection Draining / Deregistration Delay?

When an EC2 instance is:

• Being removed from a load balancer

• Marked unhealthy

• Or getting terminated by Auto Scaling

The load balancer needs to stop sending it new requests — but also let existing users finish their sessions gracefully.

That’s exactly what Connection Draining (in Classic Load Balancer) and Deregistration Delay (in Application or Network Load Balancer) do.

📘 SAA Tip:
The concept is the same, but the name depends on the load balancer:

---

🚦 Why Is This Important?

Without connection draining:

• An instance that’s in use could be cut off immediately

• In-flight requests (like file uploads, payments) may fail abruptly

• Users can experience errors or partial actions

With connection draining enabled:

✅ Current users finish their requests
✅ No new traffic is sent to that instance
✅ Instance gracefully exits the pool

---

🧪 Real-Life Example: Arjun’s EC2 Shutdown

Arjun had 3 EC2 instances behind his Application Load Balancer.

Suddenly, one instance started failing health checks. ALB began removing it.

But because Deregistration Delay was enabled, here’s what happened:

1. Existing users still on that EC2 were allowed to finish their requests

2. ALB stopped sending new traffic to that instance

3. Once all sessions ended (or timeout expired), the instance was removed

This avoided interrupted uploads, lost forms, or broken logins.

---

⏲️ How Long Is the Draining Period?

You can configure the Connection Draining / Deregistration Delay time between:

• 1 and 3,600 seconds (1 hour)

• Default: 300 seconds (5 minutes)

To disable draining entirely, set it to 0.

---

⚖️ Choosing the Right Timeout: Arjun’s Tip Sheet

🛠️ Tip: Balance user experience vs. speed of replacement.

---

✅ What Happens During the Draining Period?

While an instance is draining:

• ✅ Existing connections stay active until complete

• ❌ No new requests go to the instance

• ✅ Once all connections close (or timeout hits), the instance is removed

“It’s like a restaurant closing:
No new customers allowed in, but the current diners can finish their meal.”

---

🧠 Summary for SAA Exam

---

🎯 Final Thoughts from Arjun

Arjun updated his infrastructure with:

• A 300-second deregistration delay

• Health checks + Auto Scaling policies

• Awareness that graceful exits matter just as much as fast launches

“Now my app scales up and down without hurting users mid-request. That’s true resilience.”

More AWS SAA Articles

• AWS Network Load Balancer

• Complete Overview of AWS Application Load Balancer

• How ELB Manages Traffic with Precision

• AWS Route 53 Explained

• Amazon EFS Made Easy: The File System That Scales With You

• EBS Multi-Attach: The Power of Shared Storage in AWS

• EBS Volume Types

• What is Elastic Network Interface?

• What is AWS EC2 Placement Groups?

• Why You Must Handle IAM API Keys with Extreme Care in AWS?

• AWS Security Groups

• EC2 Instance Types

• AWS IAM Best Practice

• AWS IAM Security Tools

• IAM Roles for Services

• AWS IAM User, Root User and Groups

• AWS IAM Policies and their structure

Follow me for more such content

• My YouTube Channel

• My Site

• My Blogs

• LinkedIn

• Instagram

• Twitter

• Stackoverflow

• Github

• Dev