Experimenting on distributed systems in a controlled way, using chaos engineering can help identifying failures before they become outages.