Chaos Mesh
By CNCF
Chaos Mesh is an open-source, cloud-native chaos engineering platform, hosted by the Cloud Native Computing Foundation, that injects controlled failures into Kubernetes environments to test how applications and infrastructure respond to…
Definition
Chaos Mesh is an open-source, cloud-native chaos engineering platform, hosted by the Cloud Native Computing Foundation, that injects controlled failures into Kubernetes environments to test how applications and infrastructure respond to faults such as pod crashes, network latency, disk I/O errors, and clock skew. It is designed to run as a Kubernetes-native operator, defining chaos experiments as custom resources managed the same way as other cluster workloads.
Overview
Kubernetes systems fail in ways that are hard to reproduce with normal testing: a network partition between services, a node running out of disk I/O bandwidth, or a container being OOM-killed at the wrong moment. Chaos Mesh exists to make these failure modes reproducible and observable on demand, rather than waiting for them to happen unpredictably in production. It grew out of work at PingCAP on testing distributed databases and was later donated to the CNCF, reflecting its origins in rigorously testing stateful, distributed systems. Mechanically, Chaos Mesh installs as a set of Kubernetes controllers and a sidecar-free chaos daemon running on each node. Experiments are defined as Kubernetes custom resources — for example, a PodChaos resource that kills a percentage of pods matching a label selector, or a NetworkChaos resource that adds latency or packet loss between specific services. Because experiments are just Kubernetes objects, they can be version-controlled, applied via GitOps, and scheduled or scoped using the same selectors and namespaces used for regular workloads. A companion web dashboard provides a visual experiment designer and real-time status view. Within the chaos engineering space, Chaos Mesh is most directly comparable to LitmusChaos, another CNCF chaos project, and to Gremlin, a commercial platform with a broader non-Kubernetes fault library. Chaos Mesh's differentiation is its deep, native integration with Kubernetes primitives and its fine-grained fault types tuned for distributed systems testing, including specialized JVM and I/O chaos for stateful workloads, whereas Gremlin trades some of that Kubernetes-native depth for an easier managed experience and support for non-containerized infrastructure. In practice, platform and SRE teams use Chaos Mesh to validate that a service degrades gracefully under network partition, that a database cluster survives a leader pod being killed, or that autoscaling and retry logic behave correctly under injected latency. It is commonly run in staging or dedicated chaos-testing namespaces first, then carefully extended to production as confidence and blast-radius controls mature, often as part of formal game days. The main limitations are that Chaos Mesh requires genuine Kubernetes expertise to configure safely — a misconfigured experiment can affect more of a cluster than intended — and it is scoped specifically to Kubernetes, so organizations with significant non-containerized infrastructure will need a complementary tool or a broader platform like Gremlin. Teams without mature observability to detect and roll back a bad experiment quickly should build that foundation before running chaos experiments at all.
Key Features
- Kubernetes-native chaos experiments defined as custom resources
- Fault types spanning pod, network, I/O, kernel, and JVM chaos
- Web dashboard for visual experiment design and monitoring
- Scheduling and scoping via standard Kubernetes label selectors
- GitOps-compatible experiment definitions for version control
- Workflow feature for chaining multiple chaos steps into scenarios
- CNCF-hosted with an active open-source contributor community
Use Cases
Alternatives
Frequently Asked Questions
From the Blog
Network Topology Explained: Star, Bus, Ring, and Mesh
Network topology is the physical or logical arrangement of devices and connections in a network, and it determines a network's speed, cost, and fault tolerance. This guide compares star, bus, ring, mesh, and hybrid topologies with real trade-offs.
Read More Cloud & CybersecurityWhat Is a Mesh Network? How Mesh Topology Works
A mesh network connects every device to several others, so data can take multiple possible paths instead of relying on a single central point. This guide explains mesh topology, its advantages, and where mesh networks are used today.
Read More ProgrammingPython Virtual Environments: venv, conda, and poetry Explained
Installing packages globally is fine until it isn't — then you have version conflicts, broken projects, and chaos. This guide explains virtual environments from first principles and shows you how to use venv, pip, poetry, and conda to keep your projects isolated and reproducible.
Read More