What it does
LitmusChaos is an open-source chaos engineering platform for testing the resilience of cloud-native infrastructure and applications. Its control plane organizes experiments into workflows, while Kubernetes custom resources describe targets, faults and steady-state probes. Teams can select experiment templates from ChaosHub, add checks for expected application behavior, then review experiment results and exported Prometheus metrics. It is most relevant when a team already has a Kubernetes environment and can safely scope fault injection to a test workload.
Test the system's response, not only the fault command
A Litmus experiment combines a fault with a steady-state hypothesis. In the project model, a ChaosExperiment describes a fault and its parameters, a ChaosEngine links that fault to a workload or infrastructure target, and a workflow can order multiple experiments. Probes check whether the application is still meeting its chosen conditions; ChaosResult records validation and rollback outcomes. This gives a team more useful evidence than simply confirming that a pod was deleted or a node was stressed.
The official installation guide describes self-hosting through Helm or Kubernetes manifests. It expects a Kubernetes cluster and a persistent volume for ChaosCenter state; the guide recommends 20 GB, with a smaller volume possible for test use. After installing into a namespace, choose an experiment from ChaosHub or create one, constrain its target, define probes for the service's expected steady state and run the workflow against a non-production workload first. Results can be reviewed in ChaosCenter, and the project exports Prometheus metrics for external dashboards.
This is infrastructure testing, not a general-purpose unit-test runner. The team supplies the cluster, target application, experiment scope and health conditions. Because faults are intentionally injected, environment access and rollback behavior deserve careful review before a run. The website also mentions a hosted option, but its pricing and service terms were not established here.
Strengths
- Kubernetes custom resources let teams describe experiments, targets, probes and results declaratively.
- ChaosHub provides experiment templates, while custom experiments can extend the catalog.
- Steady-state probes and rollback information help assess the effect of a fault on the target workload.
- Experiment metrics can be exported to Prometheus and viewed in external observability tools.
Watch out for
- Self-hosting depends on a Kubernetes cluster and persistent storage, so it is not a lightweight local test runner.
- The operator must provide and scope a target workload and define its own success conditions before injecting faults.
Best for
- SRE and platform teams validating application resilience on Kubernetes.
- Developers adding controlled fault experiments to integration or delivery workflows.
- Teams that need experiment history and metrics alongside workload health checks.
Sources
- LitmusChaos - Open Source Chaos Engineering Platform (Oct 26, 2026)
- ChaosCenter installation | Litmus Docs (Oct 26, 2026)
- GitHub - litmuschaos/litmus: Litmus helps SREs and developers practice chaos engineering in a Cloud-native way (Oct 26, 2026)
- litmus/LICENSE at master · litmuschaos/litmus · GitHub (Oct 26, 2026)