How can I ensure that all pods in an EKS cluster are stressed during an AWS FIS EKS CPU Stress experiment?

Question

We’re operating a microservice on an EKS cluster with a pod range of 2 to 10. We’ve been conducting an AWS FIS EKS CPU Stress experiment using the aws:eks:pod-cpu-stress action.

However, we’ve noticed that only the initial 2 pods are being stressed by the CPU. When the pod count scales up to 10, the newly created pods are not stressed because the target resolution completes before the CPU stress starts. 
How can we ensure that all 10 pods undergo CPU stress during the experiment?

Answer

As FIS resolves the targets before starting the task (as you correctly mention) And the targets for this specific action are individual pods, I don't think FIS can currently support stress testing including autoscaled containers. 
>AWS FIS identifies all targets at the start of the experiment, before starting any of the actions in the actions set. AWS FIS uses the target resources that it selects for the entire experiment.

My workaround for running a load test on the max number of pods would be to fix the replicas to 10 before starting the test, or run a first short test to "warm up" the deployment before launching your full test.

How can I ensure that all pods in an EKS cluster are stressed during an AWS FIS EKS CPU Stress experiment?

関連するコンテンツ