Skip to content

autotraderuk/chaos-kraken

Repository files navigation

Chaos Kraken

Build Release

An application that can be used to simulate JVM based failure scenarios when running on a delivery platform.

Terrorise the Shipp'ng lanes with the Chaos Kraken, an incomplete FAAS (Failures as a Service).

The first version of Chaos Kraken was created at Auto Trader UK back in 2017. It was originally built to test various application failure conditions on their private cloud infrastructure.

As Auto Trader started their migration to public cloud and Kubernetes, the Chaos Kraken evolved to cater for more failure modes.

Chaos Kraken is actively used within Auto Trader to verify various behaviours of their GKE based delivery platform.

Usage

A GitHub package containing the executable JAR for Chaos Kraken is created every time a new release is created using a GitHub Actions workflow.

Once you've downloaded the JAR file simply run it, for example:

java -jar chaos-kraken-0.1.2.jar

Alt text

Start-up failures

Set the FAIL_ON_START environment variable to one of the failure types.

FAIL_ON_START=killapp ./gradlew clean bootRun 

Runtime failures

Send a POST request to your desired failure e.g.

curl -X POST http://localhost:8080/simulate/memoryleak

To return a desired status code, send a GET request to /echostatus/{CODE} e.g.

curl http://localhost:8080/echostatus/403

Failure types

  • toggle-service-health

    Toggle application health check between healthy and unhealthy.

  • unhealthy-service

    Toggle application health check to return unhealthy status. It will be unhealthy forever.

  • memoryleak

    Start allocating memory until error, catch the OutOfMemoryError.

  • memoryleak-oom

    Start allocating memory until error, uncaught OutOfMemoryError.

  • wastecpu

    Start hashing random strings forever.

  • threadbomb

    Spin up an infinite amount of threads that never end.

  • filehandlebomb

    Start opening random new files and never closing them.

  • filecreator

    Start creating random 1KB files in java.io.tmpdir.

  • stdoutbomb

    Start repeatedly writing timestamps to stdout.

    Accepts periodMillis as a query parameter to set the rate (defaults to 1ms)

  • diskbomb

    Start writing random 1GB files to random locations on disk.

  • killapp

    Emit a log message and then immediately terminate the JVM.

  • selfconnectionsbomb

    Start Listening on a random local port and open 5000 connections to that port

  • directmemoryleak

    Start allocating direct memory until error.

    Accepts limitMB as a query parameter to set the amount of memory to leak (default is no limit, ie until OOM)

Build

Build

Standalone

As the project is based on Gradle and SpringBoot, you can get started quickly by running:

./gradlew bootRun 

The app will be available at http://localhost:8080

Alternatively to build an executable JAR run:

./gradlew build

Docker

A Dockerfile is provided to demonstrate how you may containerise the app for deployment to a platform such as Kubernetes. A docker-compose.yaml file is also provided for convenience.

To build the image you can run:

docker-compose build

To spin up the app locally run the following:

docker-compose up

Release

To release a new version you simply push a new tag. The Publish Release action will then build, package and publish the artifacts.

For example:

$ git tag v0.2.9
$ git push --tags

NOTE: You'll need push access for the main repo to push the tag.