A two-layer failover design for a resilient web app. CloudFront handles application errors at the edge; Route 53 redirects around a regional outage to a static S3 maintenance site.
Choose a scenario to trace traffic through the system. This is an illustrative architecture walkthrough, not a live connection to AWS.
Visitors reach the primary app through CloudFront and the load balancer.
Fast application-level recovery comes first. DNS failover provides a separate route when the primary region itself is unavailable.
CloudFront monitors the primary ALB origin. On configured origin failures, its origin group serves the S3 maintenance page without waiting for DNS propagation.
Route 53 health-checks the public endpoint and switches the failover record to the standby S3 website endpoint if the primary is unhealthy.
The reference implementation uses Terraform to provision primary-region compute and networking alongside a low-cost static standby.
Distribution, ALB origin, and an S3 origin group.
Two public subnets across availability zones, with an ALB target group.
Amazon Linux launch template; scales from two to six instances.
Maintenance page hosted in the secondary region.
Health check and primary/secondary failover records.
Health-status alarm with an email notification subscription.
The source material reports these as completed tests for its reference deployment. Results are documented here, not verified against a live AWS account.
| Test | Reported result |
|---|---|
| Terraform validation and plan | ✓ Passed · Plan showed 38 resources |
| CloudFront endpoint and Route 53 health check | ✓ Passed · Website reachable; health check healthy |
| Layer 1 · forced ALB 5xx responses | ✓ Passed · CloudFront served the S3 page |
| Layer 2 · primary instances terminated | ✓ Passed · DNS switched to S3 in about 90 seconds |
| Failback and Auto Scaling | ✓ Passed · Primary recovered; capacity scaled 2 → 4 → 2 |
DNS detection and failover time vary with health-check settings, DNS TTL, and resolver caching.
Keeping the standby page static avoids running a second application fleet. The two failover layers are independently useful and should be exercised separately during testing.
CloudFront distribution changes can take time to deploy. Keep health checks public and verify the S3 website access policy and failover behavior for your own security requirements.