End-to-end recovery from AZ impairments in Amazon EKS using EKS Zonal shift and Istio

This article explains how to implement end-to-end recovery from Availability Zone impairments in Amazon EKS using zonal shift and Istio service mesh.

• Addresses gray failures: AZ degradation that appears healthy but impacts customer experience
• Combines NLB zonal shift, EKS zonal shift, and Istio to manage three traffic patterns
• NLB zonal shift removes impaired AZ from load balancer DNS routing
• EKS zonal shift removes pods from impaired AZ via EndpointSlice updates
• Istio ServiceEntry and DestinationRule enable locality-aware routing to external dependencies
• Aurora writer failover required if primary resides in impaired AZ
• Includes sample application deployment walkthrough and cleanup procedures

This comprehensive approach ensures applications maintain performance during AZ disruptions by redirecting all inbound, internal, and outbound traffic away from degraded zones.