Litmus Chaos
By CNCF
LitmusChaos is an open-source, Kubernetes-native chaos engineering framework, hosted by the Cloud Native Computing Foundation, used to run controlled fault-injection experiments such as pod deletion, network disruption, and resource…
Definition
LitmusChaos is an open-source, Kubernetes-native chaos engineering framework, hosted by the Cloud Native Computing Foundation, used to run controlled fault-injection experiments such as pod deletion, network disruption, and resource exhaustion against applications running on Kubernetes. It packages experiments as reusable, community-maintained "chaos charts" published to a public catalog, letting teams compose them into larger workflows to validate system resilience without writing custom fault-injection scripts themselves.
Overview
Validating that a Kubernetes application survives real failure conditions typically requires writing bespoke scripts to kill pods, throttle networks, or fill disks, then manually observing the result. LitmusChaos was created to standardize that process by turning individual failure scenarios into reusable, shareable building blocks rather than one-off scripts each team writes from scratch. It originated at MayaData and was contributed to the CNCF, where it now operates as a community-governed project alongside other Kubernetes-native tooling. Architecturally, LitmusChaos runs a chaos operator in the cluster that watches for ChaosEngine custom resources, each of which references a specific ChaosExperiment — a containerized, self-contained unit of fault-injection logic pulled from the public ChaosHub catalog of community-contributed experiments. This experiment-as-a-package model means a team does not need to write pod-kill or network-latency logic themselves; they select an existing chart, configure its target and parameters, and apply it as a Kubernetes resource. Results, called "chaos results," are recorded as their own custom resources, making outcomes queryable and auditable through standard Kubernetes tooling. LitmusChaos sits alongside Chaos Mesh as one of the two major CNCF-hosted, Kubernetes-native chaos engineering projects, and both are frequently compared to commercial platforms like Gremlin. Its differentiation is the ChaosHub ecosystem model — a marketplace-like catalog of prebuilt, community-contributed experiments — which favors teams that want to adopt tested fault scenarios quickly rather than author their own from primitives, whereas Chaos Mesh leans more toward direct, fine-grained custom resource authoring. In practice, platform teams use LitmusChaos to build resilience test suites that run as part of CI/CD pipelines, chaining together experiments like pod failure, node drain, and network partition into a single workflow that gates a release or nightly build. SRE teams also use it during game days to rehearse incident response against realistic, repeatable failure scenarios rather than relying on tribal knowledge of past outages. The framework requires meaningful Kubernetes and chaos engineering maturity to use safely, since composing multiple experiments into pipeline gates raises the risk of an experiment affecting shared infrastructure if scoping is not carefully constrained. Teams new to the project also face a learning curve around the ChaosEngine and ChaosExperiment resource model before they can confidently author or safely modify a workflow rather than only running examples pulled directly from ChaosHub. It is also Kubernetes-specific, so, like Chaos Mesh, it does not cover chaos testing for non-containerized systems, virtual machines, or traditional infrastructure without additional tooling, and organizations with mixed environments typically pair it with a separate solution for the non-Kubernetes portion of their estate.
Key Features
- ChaosHub catalog of reusable, community-contributed experiment charts
- ChaosEngine and ChaosExperiment custom resources for declarative configuration
- Recorded ChaosResult resources for auditable experiment outcomes
- Chaos workflows for chaining multiple experiments into pipelines
- CI/CD integration for gating releases on resilience test outcomes
- Litmus Portal for centralized experiment scheduling and observability
- CNCF governance with active open-source community contributions