~/ nmadiraju.com Amazon EKS · Failure testing with AWS FIS
arrow_back Back to nmadiraju
Visual Walkthrough · Resilience

How Failure Testing with AWS FIS Works on Amazon EKS

Follow checkout's 3 replicas through three AWS FIS experiments before a peak sale, then check yourself with 8 questions.

1Visual Walkthrough

AWS FIS experiments against the checkout service on an EKS cluster across three Availability Zones Top: your team, the AWS FIS experiment template, the CloudWatch alarm checkout-5xx used as a stop condition, the ALB for checkout and the node capacity (a managed node group, later Karpenter). Bottom: cluster shop-staging in your VPC with node-1 in us-east-1a, node-2 and node-4 in us-east-1b and node-3 in us-east-1c, running checkout Pods and checkout-redis. The walkthrough shows a node termination taking all replicas, spreading them, a Pod CPU stress handled by readiness, and a zone power interruption fixed with Karpenter and data kept in two zones. Your team aws fis · kubectl AWS · us-east-1 CloudWatch alarm checkout-5xx ALB 5xx > 1% for 1 min ALB · checkout target type: ip Service checkout :8080 VPC · CLUSTER shop-staging us-east-1a us-east-1b us-east-1c node-1 m6i.large · us-east-1a gone node-2 m6i.large · us-east-1b gone m6i.large · us-east-1c gone node-4 · by Karpenter m6i.large · us-east-1c gone checkout-redis-0 EBS gp3 volume in 1a checkout Pod Running · Ready checkout Pod Running · Ready checkout Pod Running · Ready checkout Pod Running · Ready checkout Pod Running · Ready checkout Pod Running · Ready checkout Pod Running · Ready FIS Pod · fis-sa injects and watches the fault checkout-redis-1 replica · own volume in 1c checkout Pod cart, orders Pods use the rest of the CPU cart, orders Pods use the rest of the CPU cart, orders Pods use the rest of the CPU

    2Quiz to check understanding

    Check yourself