What are the best practices for Disaster Recovery (DR) in a Cloud-Native environment?

All answers to this question.

To meet an RPO of 15 minutes, you definitely need "Pilot Light" at a minimum, which involves keeping your core data live in a second region. For your database, use "Aurora Global Database" or "RDS Cross-Region Read Replicas"—these handle the heavy lifting of asynchronous replication with very low lag. For failover, Route 53 with "Health Checks" is the standard, but AWS Global Accelerator is faster because it doesn't rely on DNS TTL, which can be cached by ISPs. To test your plan, use "Chaos Engineering" principles. Run a "Game Day" once a quarter where you simulate a region failure in your staging environment to see if your automation actually kicks in.

   Answered 2024-01-19 by Susan Martinez

How are you managing the synchronization of your application secrets and IAM roles across regions? That's usually what breaks during an actual failover.

   Answered 2024-01-21 by Kevin Adams

• Kevin, we use Terraform to ensure our infrastructure is identical in both regions. For secrets, we use AWS Secrets Manager with "Multi-Region Replication" enabled. This ensures that if we fail over to us-west-2, the local Secrets Manager already has the latest DB credentials synced from us-east-1. We also use "IAM Global Endpoints" to ensure our permissions are consistent. The key is to never manually change anything in the secondary region; it must be a mirror of the primary.

     Commented 2024-01-22 by Mark Robinson

Remember that "Data Transfer Out" costs for cross-region replication can be the biggest part of your bill. Always compress your data before sending it if you can.

   Answered 2024-01-23 by Dorothy Miller

• Great point, Dorothy. We actually filtered our replication streams so we only send "Critical" tables to the DR region, which saved us about 40% on inter-region transfer fees.

     Commented 2024-01-24 by Jessica Thompson