Amazon Application Recovery Controller (ARC) provides controls for recovering applications from Availability Zone and Regional impairments. Its capabilities address different failure scopes: zonal shift and zonal autoshift operate within a Region, while routing controls and Region switch support multi-Region recovery. Readiness checks help identify configuration and capacity differences that could interfere with failover. These capabilities can be adopted separately according to the application's recovery requirements.
Zonal shift temporarily moves traffic for a supported resource away from one Availability Zone. Traffic goes to the remaining eligible Availability Zones rather than to an operator-selected destination. Engineers can initiate a shift when application telemetry indicates an isolated zonal problem. Shifts have an expiration time, so incident procedures should include monitoring, extension when necessary, and cancellation after recovery. Resource eligibility and prerequisites vary by integration; verify the requirements for the specific load balancer, Auto Scaling group, or EKS configuration.
Zonal autoshift allows AWS to initiate a traffic shift for enabled resources when AWS detects a potential infrastructure impairment. Practice runs exercise the application's ability to operate without an Availability Zone. Scheduled and on-demand practice runs help expose insufficient capacity or application dependencies before an incident. Configure appropriate monitoring and practice-run controls, and ensure that the remaining zones can sustain expected production traffic.
Routing controls are on/off switches hosted on an ARC cluster. They connect to Amazon Route 53 health checks associated with DNS records, allowing operators or automation to change which application endpoint receives traffic. Safety rules can reject state changes that violate configured safeguards, such as disabling every recovery destination. DNS caching and existing connections affect how quickly clients move; a routing-control update does not immediately redirect every request.
Region switch orchestrates the steps required to operate an application in another Region. Plans can coordinate supported service actions and custom Lambda actions, with execution visibility and logging. Depending on the workload, a plan might scale destination compute, perform a database recovery action, wait for target health, and change traffic routing. Define action ordering, permissions, success criteria, and recovery-time objectives explicitly. Test both failover and the procedure for returning to the original Region.
Readiness checks assess supported resource configurations, capacity, and related recovery prerequisites. They are preparation signals, not substitutes for application health monitoring or end-to-end recovery tests. ARC does not independently replicate application data or guarantee a recovery point objective. Database replication, storage consistency, dependency availability, and sufficient destination capacity remain architectural responsibilities.
Prepare incident access before an outage. For routing controls, retain the cluster's Regional data-plane endpoints and implement endpoint retry behavior rather than depending on configuration discovery during recovery. Readiness and recovery-control configuration operations use the US West (Oregon) Region, while routing-control state changes use cluster endpoints. Check feature-specific Regional availability, IAM permissions, service quotas, and downstream API limits before deploying recovery automation.