Developing a chaos architecture mindset

We’ve seen cloud usage patterns begin with a faster data center and greenfield applications, move to cloud-native migrations, and end up with complete data center replacement strategies. These patterns are driving even more business-critical backend workloads to the cloud, and new patterns are emerging for highly automated, available, and durable cloud-based architectures.

However, a recurring problem with highly available architectures is that they don’t get enough exercise to ensure they will work correctly under turbulent conditions—and the weakest link is often the people operating the systems. Most enterprises have a backup data center, but in many cases disaster recovery failover and incident response isn’t practiced regularly.

Chaos engineering leverages carefully designed failure injection tests and the distributed automation inherent in cloud deployments to prove that there is enough margin to absorb failures in production. Adrian Cockcroft outlines the overall architectural principles of chaos engineering and shares methods engineers can use to exercise failure modes in safety and business-critical systems.