khaos-monkey

What khaos-monkey is

khaos-monkey randomly terminates Kubernetes pods according to configured rules, for teams that want to schedule experiments on a cluster. It can terminate a percentage of pods, a fixed count, or pods until a chosen number remain. Attack intervals and kill counts can be randomized. Targeting can select namespaces or individual pods labeled `khaos-enabled: true`; pods labeled `khaos-enabled: false` are excluded. Namespaces can also be blacklisted, with documented defaults of `kube-system`, `kube-public`, and `kube-node-lease`. Pods are grouped by ReplicaSet unless custom `khaos-group` labels define other groups. The project is built in Rust using kube-rs and is presented as a lightweight option for smaller experimental clusters. The README describes deploying it as a Kubernetes Deployment with the `dagandersen/khaos-monkey:latest` image, or running it locally with Cargo and a kubeconfig. Its installation example grants a Kubernetes ClusterRole list and delete permissions for pods and namespaces. The project is free and open source; its page includes a README and issue tracker but states no support plan.

Who it suits

It suits operators of smaller experimental Kubernetes clusters who want configurable pod-termination experiments. Teams should consider the listed permissions before setting it up.

Limits

The example installation requires a ClusterRole with list and delete permissions for pods and namespaces. The project is positioned as a simple option for smaller experimental clusters, with more advanced tooling suggested for mature infrastructure.

Getting started

1. Open the project README at https://github.com/dag-andersen/khaos-monkey.
2. Choose a Kubernetes Deployment using the dagandersen/khaos-monkey:latest image, or run locally with Cargo and a kubeconfig.
3. Review the example ClusterRole permissions for listing and deleting pods and namespaces.
4. Configure the kill mode, targeting labels or namespaces, and any exclusions.
5. Set attack intervals and kill counts, including randomization if desired.
6. Define custom pod groups with khaos-group labels if needed.

Verdict

khaos-monkey offers several ways to target and schedule pod termination experiments. Its stated focus is smaller experimental clusters, and its installation permissions merit review before deployment.