Run Chaos Toolkit locally with Docker¶

Docker is a fantastic approach to package up your Chaos Toolkit experiment dependencies and make it easy for anyone to run your experiments.

Official Images¶

Chaos Toolkit offers three base images you can use.

chaostoolkit/chaostoolkit:latest This is an Alpine based image and does not contain any extensions
chaostoolkit/chaostoolkit:basic This is an Ubuntu based image and does not contain any extensions
chaostoolkit/chaostoolkit:full This is an Ubuntu based image and does contains many popular extensions

Create your Own Container Image¶

While an image such as chaostoolkit/chaostoolkit:full is a good default to start with, you will likely want to create your own image with specific extensions and dependencies for your needs.

Let’s see how to get on about it.

Dockerfile

FROM chaostoolkit/chaostoolkit:basic  Let’s start with the Ubuntu base


COPY pyproject.toml pdm.lock /home/svc/  Copy the file listing the Python packages we want to install, see in the next section


USER root  Switch back to the admin user as we need to run priviledged commands


WORKDIR /home/svc

RUN apt-get update && \  Refresh the system

    apt-get install -y --no-install-recommends curl python3.11-venv build-essential gcc && \  We need the curl command to retrieve the Python package manager PDM used to install yopur packages. We also install the essentials to build extensions that require compilation. If your do not require it, you can usually safely remove build-essential and gcc to speed up the build

    curl -o install-pdm.py -sSL https://raw.githubusercontent.com/pdm-project/pdm/main/install-pdm.py && \  Fetch PDM

    python3.11 install-pdm.py -p /home/svc && \  Install the pdm tool in the existing home directory

    export PATH=/home/svc/bin:$PATH && \
    pdm use .venv && \  Switch to the Python environment created on the base image

    pdm install --no-editable --no-self && \  Install your Python dependencies

    rm install-pdm.py && \
    chown --recursive svc:svc /home/svc/.venv  && \  Make everything we just installed owned by the svc user created in the base image

    apt-get remove -y build-essential gcc && \  Tidy up

    apt-get clean && \
    rm -rf /var/lib/apt/lists/*

USER 1001  Switch bask to the svc user from now on

Before you can create your image, you now need to create a pyproject.toml file listing the dependencies you wish to install. Here a basic one:

pyproject.toml

[project]
name = "my-experiment"  Set any name you need, for the purpose of this file this name is not relevant

version = "0.1.0"  Not relevant to our needs here so leave as-is

dependencies = []  Contains the list of dependencies

requires-python = ">=3.11"

[build-system]
requires = ["pdm-backend"]
build-backend = "pdm.backend"

[tool.pdm]
distribution = false

To add dependencies, run the following command:

pdm add chaostoolkit-kubernetes

Tip

Install pdm from pdm.

The first time this command is run, it generates the pdm.lock file describing the specific version to install.

We’re done!

You can now create the image with:

docker build -t my/chaostoolkit .

Once this image has been published to a container registry, it is available to use by anyone with access or from other environments such as Kubernetes or serverless providers.

Run an Experiment¶

Let’s assume you have this Chaos Toolkit experiment:

experiment.json

{
    "version": "1.0.0",
    "title": "Hello world!",
    "description": "Say hello to the world.",
    "method": [
      {
        "type": "action",
        "name": "say-hello",
        "provider": {
          "type": "process",
          "path": "echo",
          "arguments": "hello"
        }
      }
    ]
}

You can run this experiment as follows:

docker run -v `pwd`/experiment.json:/home/svc/experiment.json my/chaostoolkit run experiment.json