What it does
Chaos Mesh is an open-source chaos-engineering platform for Kubernetes. It uses Kubernetes custom resources and a web dashboard to define fault experiments, select which workloads they affect, and observe their status. Teams can model one-time, scheduled or cyclic failures, such as pod or network faults, and use the results to investigate how an application recovers.
Define a controlled fault, then watch the recovery
Chaos Mesh runs in a Kubernetes environment and models failure tests as experiments. Install it into a cluster using the documented Helm path, then create an experiment that specifies the fault type, target selectors and duration. A one-time experiment injects the configured fault immediately; the current guide says a duration can restore the fault automatically, while pausing or deleting the experiment restores it immediately. Scheduled or cyclic experiments can repeat according to configured rules.
The dashboard helps users create experiments and inspect their state, while Kubernetes custom resources offer a declarative route for teams that keep infrastructure configuration in version control. For example, a staging team might introduce a bounded network delay for selected pods, then observe whether a dependent service retries and remains available. That is an illustrative test scenario, not a claim that a cluster was exercised here.
Scope matters: Chaos Mesh selects targets using rules such as labels and namespace controls, and its documentation describes RBAC protections for dashboard access. Start in a test cluster, constrain selectors to the intended workloads and keep the experiment duration explicit. The project is available under Apache License 2.0 and is intended to be deployed and operated by the team that owns the Kubernetes cluster.
Strengths
- The project provides several fault-simulation types and can orchestrate one-time, scheduled or cyclic experiments.
- Target selectors and namespace controls help teams constrain which workloads receive an injected fault.
- Experiments can restore faults after a configured duration or immediately when paused or deleted.
- A web dashboard and Kubernetes custom resources support both visual and declarative workflows.
Watch out for
- Chaos Mesh requires an existing Kubernetes cluster and permissions to deploy and configure its components.
- Faults act on the selected targets, so selectors and experiment duration must be bounded to avoid disrupting unrelated workloads.
Best for
- Platform teams validating resilience and recovery behavior in Kubernetes staging environments.
- Engineers testing failover, retries and dependency behavior through controlled workload faults.
- Teams that want chaos experiments represented as Kubernetes resources alongside their cluster configuration.
Sources
- Chaos Mesh: A Powerful Chaos Engineering Platform for Kubernetes (Oct 25, 2026)
- Chaos Mesh Overview | Chaos Mesh (Oct 25, 2026)
- Run a Chaos Experiment | Chaos Mesh (Oct 25, 2026)
- Chaos Mesh (Oct 25, 2026)
- Apache License Version 2.0 (Oct 25, 2026)