What Happens When an AWS Region Goes Down?
Designing a simple multi-region failover architecture using Amazon Route 53 and custom API domains.
Introduction
When building applications on AWS, many developers deploy their backend in a single region. It works perfectly most of the time, but what happens if that region becomes unavailable?
AWS outages are rare, but they do happen. If your application depends on only one region, your entire system may go offline.
In this article, we will understand:
What happens when an AWS region goes down
Which AWS services are global vs regional
How to design a simple failover architecture
Understanding AWS Regions
AWS infrastructure is divided into regions, and each region contains multiple Availability Zones (AZs).
Example regions:
ap-south-1– Mumbaius-east-1– Virginiaeu-west-1– Ireland
If your backend runs entirely in one region and that region fails, your entire application can become unavailable.
Global vs Regional AWS Services
Some AWS services are global, while others operate in a specific region.
Global Services
Amazon Route 53 – Global DNS and traffic routing
Amazon CloudFront – Global content delivery network
These services are not tied to a single region and can route users to different regions.
Regional Services
AWS Lambda
Amazon API Gateway
Amazon RDS
These services run inside specific regions.
Mixed Behavior
- Amazon S3
S3 buckets exist in a region, but the service has a global namespace and high availability across AZs.
The Problem with Hardcoded Region Endpoints
Many frontend applications directly call API Gateway URLs like:
This tightly couples your application to one region.
If that region goes down, the frontend cannot reach the backend.
Architecture:
Problem:
If the Mumbai region (ap-south-1) goes down:
The frontend cannot reach the API
Users see errors
Your entire application becomes unavailable
Your whole application stops.
Better Approach: Use a Custom Domain
Instead of calling a regional API URL, use a domain like:
The frontend always calls this domain, while DNS decides where the request goes.
Architecture:
Solution:
If Region A fails:
Route53 → automatically routes traffic → Region B
Users barely notice.
Request Flow Example
If the primary region fails, Route53 automatically routes traffic to the secondary region.
Benefits of This Architecture
No region dependency in the frontend
Automatic failover
Improved availability
Global performance improvements
| Architecture | Risk | Complexity |
|---|---|---|
| Single Region | Region outage breaks app | Simple |
| Multi-AZ | Data center failure protected | Medium |
| Multi-Region | Region failure protected | Complex |
Practical Tip for Developers
A simple way to implement this architecture:
Deploy your backend in two regions
Use Amazon Route 53 health checks
Configure failover routing
Use Amazon CloudFront or a custom domain like:
Now, if the primary region fails, traffic automatically shifts to the backup region.
How You Can Implement This in AWS (Step-by-Step)
Here is a simplified way to implement this architecture using Amazon Web Services.
1. Deploy Backend in Two Regions
Deploy the same backend stack in two regions:
Primary region:
ap-south-1(Mumbai)Secondary region:
ap-southeast-1(Singapore)
Each region contains:
2. Create a Custom API Domain
Instead of calling the regional endpoint:
Create a custom domain like:
You can configure this using Amazon API Gateway.
3. Configure DNS Failover
Use Amazon Route 53 to manage DNS routing.
Create two records:
Primary Record
api.example.com → Region A API
Routing: Failover (Primary)
Health Check: Enabled
Secondary Record
api.example.com → Region B API
Routing: Failover (Secondary)
If the primary region fails, Route53 automatically routes traffic to the backup region.
4. Optional: Add Global CDN
For even better performance and caching, you can place Amazon CloudFront in front of the API.
Benefits:
Faster global latency
Additional resilience
Edge caching
Final Thoughts
While full multi-region setups can be complex, understanding global services like Route53 and CloudFront helps you design more resilient systems.
Even simple architectural decisions like using a custom domain instead of regional endpoints can significantly improve reliability.
Have you ever implemented a multi-region architecture on AWS, or are you still running everything in a single region? I’d love to hear how you handle reliability in your systems.
