ChaosBlade

Visit Website

Introduction

ChaosBlade is an easy-to-use chaos engineering toolkit that originated at Alibaba. The README says it is based on nearly ten years of Alibaba’s failure testing and drill practice, and it is now an open-source project with a CNCF sandbox history.

Its strength is breadth from a single command-line interface. The README groups experiment scenarios into basic resources (CPU, memory, network, disk and process), Java applications (databases, caches, messaging, the JVM itself and microservices, with the ability to target any class method), containers (killing containers or stressing CPU, memory, network, disk and processes inside them), and Kubernetes resources such as nodes, pods and containers. Cloud scenarios cover some cloud resources too.

Each scenario is implemented by an executor project: chaosblade-exec-os for host faults, chaosblade-exec-jvm for Java bytecode-level injection, chaosblade-exec-docker and chaosblade-exec-cri for containers through the Docker API or the Container Runtime Interface, and chaosblade-operator for Kubernetes, where experiments are defined as ChaosBlade custom resources. The CLI follows a consistent grammar, blade create <target> <action> [flags], and every experiment returns an ID you can use to query status or destroy it.

The JVM executor is a notable capability. It can attach to a running Java process and inject delays, exceptions or return-value changes into specific methods, Dubbo calls, JDBC operations, Redis or HTTP clients, which lets teams test application-level resilience without code changes.

Recent releases add Blade AI, described in the README as an intelligent agent layer that invokes ChaosBlade for fault injection and adds orchestration such as intent understanding, security auditing, effect verification, safe recovery and structured reporting. ChaosBlade Box, a separate project, provides a web console for managing experiments.

In the quality loop ChaosBlade belongs at the production and pre-production resilience stage. It suits teams that want one toolkit for host, container, Kubernetes and Java-level faults, and it is especially well documented for the Java ecosystem.

The source code is at chaosblade-io/chaosblade on GitHub under the Apache-2.0 licence, with about 6,500 stars; the most recent commit was on 29 October 2026.

Recording each experiment’s hypothesis and outcome builds a library of known failure modes.

Features

1. Host faults: CPU, memory, disk, network and process experiments.
2. JVM injection: Delays, exceptions and return values in chosen Java methods.
3. Middleware scenarios: Dubbo, JDBC, Redis, HTTP and other Java clients.
4. Container faults: Docker and CRI-based container experiments.
5. Kubernetes operator: ChaosBlade custom resources for nodes, pods and containers.
6. Consistent CLI: blade create, blade status, blade destroy.
7. Experiment IDs: Track and roll back each experiment.
8. Blade AI: Agent layer for orchestration and reporting.
9. Web console: ChaosBlade Box for managing experiments.
10. Extensible executors: Separate projects per target platform.

Pros

1. Breadth: One toolkit covers hosts, containers, Kubernetes and Java apps.
2. Java depth: Method-level fault injection without code changes.
3. Free to use: Open source under the Apache-2.0 licence, so the core tool costs nothing to use.
4. Actively maintained: The repository had commits as recent as 29 October 2026.
5. Community: About 6,500 GitHub stars.

Cons

1. Documentation mix: Some documentation and community discussion is primarily in Chinese.
2. Many components: Executors and operator are separate projects to install and upgrade.
3. Safety is on you: Blast radius and abort conditions must be planned carefully.
4. Smaller Western community: Fewer English tutorials than Chaos Mesh or Litmus.

Usage and Links (Example)

Install and run a CPU experiment on a host:

blade create cpu fullload --cpu-percent 60 --timeout 120
blade status --type create
blade destroy <experiment-id>

Inject a delay into a Java method:

blade prepare jvm --process my-service
blade create jvm delay --time 1500 --classname com.example.OrderService --methodname place

On Kubernetes, with the operator installed:

apiVersion: chaosblade.io/v1alpha1
kind: ChaosBlade
metadata: { name: pod-network-delay }
spec:
  experiments:
    - scope: pod
      target: network
      action: delay
      matchers:
        - { name: names, value: ["checkout-7d9c"] }
        - { name: namespace, value: ["shop"] }
        - { name: interface, value: ["eth0"] }
        - { name: time, value: ["3000"] }

Practical tips:

1. Always set --timeout so a forgotten experiment cleans itself up.
2. Start in staging with clear steady-state metrics, then move to production carefully.
3. Use JVM experiments to test retries, timeouts and fallbacks in Java services.

Resources:

• ChaosBlade website
• GitHub repository

Pricing

ChaosBlade is free and open source under the Apache-2.0 licence.

Recommended for?

1. Java-heavy organisations that want method-level fault injection.
2. Platform teams wanting one CLI for host, container and Kubernetes faults.
3. Teams running regular game days with varied failure scenarios.
4. Consider Chaos Mesh or Litmus if you want a Kubernetes-first project with a larger English-language community.

Conclusion

ChaosBlade is a broad, practical chaos toolkit with unusual depth for Java applications. Its consistent CLI and Kubernetes operator make it easy to script experiments across very different targets.

Plan safety first: set timeouts, define steady-state metrics, and keep experiments small until the team trusts both the tool and the system.

On CheckOps, ChaosBlade is filed under the Production monitoring & observability and Integration testing stages. Related tools listed here include Chaos Mesh, LitmusChaos, Chaos Toolkit, Gremlin, Steadybit and Toxiproxy. Which one fits best depends mostly on your stack, your team’s size and skills, and whether you prefer to run tools yourself or pay for a managed service. This page reflects the official website, documentation and repository as they stood on 11 November 2026.