ChaosCat

Chaos engineering for Pull Requests. ChaosCat merges open pull requests at random, subject to configurable parameters.

Clone and run with stack:

git clone
cd chaoscat

# Set environment variables
export CHAOSCAT_OWNER=torvaney
export CHAOSCAT_REPO=example-pr-repo
export CHAOSCAT_AUTH=abcdefghijklmnop  # Put your actual GitHub auth token here
export CHAOSCAT_RATE=0.5       # Avg hours (after minimum) before closing a pull request
export CHAOSCAT_MINHOURS=0.15  # Minimum hours before a pull request is closed

stack run

How does it work?

ChaosCat will merge Pull Requests according to an exponential distribution. This means that Pull Requests will be merged independently of one another, and at a constant average rate.

This average rate is configured with the CHAOSCAT_RATE environment variable, which sets the average (mean) time before a Pull Request is closed.

You can also set a minimum amount of time before a Pull Request gets merged, with the CHAOSCAT_MINHOURS environment variable.

We can see what this distribution looks like in practice when applied to the torvaney/example-pr-repo repo with a minimum time of 0.15 hours (9 mins) and a rate of 0.5 hours (30 mins):

As you can see, no Pull Requests are merged initially. But once the minimum time threshold is passed, they are merged at random, following the exponential curve as expected (pink line).

But… why?

ChaosCat is a response to the question no one has been asking: what if we applied the principles of Chaos Engineering to our development practices themselves?

principlesofchaos.org states that

Chaos Engineering is the discipline of experimenting on a system in order to build confidence in the system’s capability to withstand turbulent conditions in production.

In the same way, if we want to ensure that you can ship features to production frequently, and gracefully handle bugs, what better way than to ship features… automatically?

Chaoscat (named after GitHub’s Octocat and Netflix’s Chaos Monkey) merges your pull requests at random, forcing you to develop robust processes around code review and feature releases