EKS Resilience · FinOps
AWS EKS · Chaos Engineering · FinOps
How often must an availability zone fail before multi-AZ pays for itself? I built two identical EKS platforms differing by a 15-line Terraform diff, then broke them with AWS Fault Injection Simulator to measure — not assume — the true cost and benefit of high availability.
Single-AZ never recovered (2.71% availability); multi-AZ held at 98.81% across three chaos runs. The finding: resilience cost lives in compute spread, not the database — Multi-AZ RDS is $32/mo, spreading the workload is $103/mo. Hardened with Pod Identity, IMDSv2, KMS envelope encryption and default-deny NetworkPolicy.