Orchestration Allows Microservices to Be Unreliable (And That's a Good Thing)
One of the first features I wanted to build for Kubernetes was service workflows: Service A starts, then B, then C. If B fails, A should know, and C shouldn't panic. Services need to know