ChaosKube is similar to Netflix’s chaos monkey for Kubernetes clusters. It schedules random termination of pods in order to test the fault tolerance of a highly available system.


As a part of our charts development we needed a way to test the fault tolerance of our deployments.


Using ChaosKube is a manual step we do after our weekly demos. The intended use case of ChaosKube is to kill pods randomly at random times during a working day to test the ability to recover. The way we use it is a bit different, we manually launch ChaosKube in debug mode and manually identify the weak points of our deployment.

Later, we intend to integrate it into our CI pipeline, so whenever new changes are rolled out we have a ChaosKube run for that release.


The installs and unleashes ChaosKube by scheduling a run 10m after installing ChaosKube by default. It also sets up the needed service account and role if RBAC is enabled.

After you clone the charts repository, to install and unleash ChaosKube, run:

scripts/ up


ChaosKube can be configured by editing the scripts/chaoskube-resources/values.yaml file. For more info read the official ChaosKube docs.

You can also configure the deployment with flags on the script. To find all available options, run:

scripts/ -h

Visit the README’s values section for a full list of options to pass via --set arguments.