-
Hello Karl,
You’re absolutely right that a database replica must be running in the backup region for replication to occur. In AWS disaster-recovery terminology, Option 3 describes a warm-standby architecture. In a warm-standby design, the database is not “scaled down” to zero or stopped, it is running on a smaller instance type and continuously receiving asynchronous replication from the primary. AWS explicitly defines warm-standby as a fully functional, scaled-down environment in which key components (including the database) remain online. In this context, “scaled-down” applies only to the web and application tiers, which are deployed in small Auto Scaling groups behind ALBs. The database replica remains running because it must be up for continuous cross-region replication.
This is why Option 3 meets the RTO/RPO requirements. The RPO of 1 hour is satisfied because asynchronous cross-region replication is active. The RTO of 5 minutes is achievable because failover does not require resizing or stopping/starting the database instance—failover simply promotes or points the application to the already-running standby. The only scaling event during failover is the web and app layer, which Auto Scaling Groups can expand very quickly, while Route 53 flips traffic to the backup region. This aligns with AWS guidance that warm-standby is the most cost-efficient DR strategy for single-digit-minute RTO.
Option 2, in contrast, fails for the exact reason you mentioned: it requires vertical resizing of the standby database during a disaster. AWS vertical resizing of EC2 instances is not fast enough to guarantee a 5-minute RTO, especially for large DB instances. Because its database starts at a smaller size and must be manually resized during failover, Option 2 cannot meet the RTO requirement in practice, even though the phrasing may imply some replication is in place.
So while the question could be more straightforward about the database instance not being “scaled down” in the sense of being shut off, Option 3 is the only option that aligns with AWS’s documented warm-standby DR pattern, meets the 5-minute RTO and 1-hour RPO, avoids slow resizing operations, and keeps cost lower than a fully duplicated multi-region environment.
Regards,
Nikee @ Tutorials Dojo
