Chaos Engineering: an introduction


Chaos Engineering. Building Confidence in System Behavior through Experiments

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

Netflix has been a practitioner of Chaos Engineering for a long time and at the origin of what can be called now a discipline – some organizations have a dedicated chaos engineering team. Testing is about performing well defined tests in well defined conditions and expecting a well known answer. Chaos Engineering is more about conducting experiments by producing a failure and learning from it (how the system behave). According to the result it could be time to work on improving things.

According to each use case chaos engineering can spread from:

  • lower layers: at infrastructure level by rebooting an instance,
  • to higher layers: at service level by flooding an API with a kind of DoS,
  • or even at team level by producing a failure requiring manual intervention, in this case it’s called game days.

The key point behind all of that is if you have not experienced such cases, believing and saying to your customers that your services are resilient is just a fairy tale – even if well designed and documented. And the last point is that it’s much more confortable to experiment such failures when you know that they will happen and that you are ready to handle it, than in an unexpected way when you are forced to fix it in urge while a lots of customers are stuck, your boss – and also his boss – is on the phone and it’s Friday 5 PM. Mature organizations like Netflix create failures 24/7 in an automated way with its Simian Army, but it’s better to start smaller.

There is so much to say on this topic that I will write next posts. In the mean time my advice is to start with the free ebook – my advice is always to start by reading a book – Chaos Engineering: Building Confidence in System Behavior through Experiments (the illustration of this article has been borrowed from the book cover), you can also check – and contribute to – the list of books dedicated to this topic I’ve created on Goodreads: Chaos Engineering.