kube-monkey

What kube-monkey is

kube-monkey is a self-hosted, open-source tool that randomly deletes Kubernetes pods to help develop and validate services designed to withstand failures. Applications opt in using Kubernetes labels, including an enabled label and settings such as an identifier and MTBF. Dry run is enabled by default, so the tool logs planned terminations rather than carrying them out. Users can configure run days, schedule creation time, termination hours, timezone, and the pod shutdown grace period. Deployments, StatefulSets, and DaemonSets are supported directly; other resource kinds can be configured as custom resources with matching RBAC. Namespace whitelists can narrow eligible workloads, while blacklists can exclude namespaces; kube-system is included in the default blacklist. The README provides Helm installation instructions for deploying the tool into kube-system. Project documentation describes notifications to Slack or a user-provided API, as well as a Prometheus metrics endpoint. Namespace pattern filtering requires cluster-wide permission to list workloads; a namespace-scoped Role installation logs an error and schedules nothing. The project is distributed under the Apache License 2.0.

Who it suits

kube-monkey suits Kubernetes teams that want to exercise failure-resilience by scheduling pod termination experiments on workloads that opt in. It is intended for self-hosted deployments and offers namespace controls, dry run, and configurable scheduling.

Limits

Dry run logs planned terminations instead of executing them by default. Namespace pattern filtering needs cluster-wide workload list permission; a namespace-scoped Role installation logs an error and schedules nothing.

Getting started

1. Review the project README and contribution guidance on GitHub.
2. Deploy kube-monkey into the kube-system namespace using the README's Helm instructions.
3. Configure Kubernetes workloads to opt in through labels and settings such as an identifier and MTBF.
4. Set run days, schedule creation time, termination hours, timezone, and pod shutdown grace period.
5. Keep dry run enabled to log planned terminations, or configure the tool for actual terminations.
6. Configure namespace controls and, if desired, notifications to Slack or a user-provided API.

Verdict

Choose kube-monkey if you need a free, self-hosted Kubernetes tool for scheduled pod-termination experiments with opt-in labels and namespace controls. Look elsewhere if you need cloud or network fault injection; also account for the cluster-wide permission requirement when using namespace pattern filtering.