AWS Introduces Multi-Day AZ Evacuation Drills

Amazon Web Services (AWS) has detailed a new approach for simulating prolonged disruptions within cloud environments. The method utilizes AWS Application Recovery Controller (ARC) Zonal Shift to conduct multi-day Availability Zone (AZ) evacuation drills.

Traditional disaster recovery tests often focus on brief failover mechanisms, verifying that traffic can be shifted and applications respond. However, these short exercises may not expose issues that arise only during sustained outages. The new AWS approach mandates applications to handle full production load on N-1 capacity for 48–72 hours, thereby revealing problems missed by shorter tests.

These extended drills are designed to uncover failures related to auto-scaling policies not tuned for prolonged N-1 operation, stale DNS entries, time-based operational tasks, and the behavior of long-lived database connections. Crucially, they also test the operational capability of teams to manage systems for multiple days with reduced capacity.

The process verifies sufficient application capacity, database stability, and client reconnection behavior. It demonstrates the teams' ability to operate effectively during extended impairments. By directing traffic away from a single AZ, ARC Zonal Shift provides a realistic simulation of a long-term outage scenario, proving the application's resilience and the operational readiness of the supporting teams.