One of the challenges with a microservices architecture is understanding it. How do you figure out what connects to what, where things are deployed, what capabilities the total system of services has?
This is a talk about System-Z, the microservice-based microservices modelling system used at Spotify to understand their set of services. We use a combination of runtime collection of data and mostly-static service metadata. The system currently tracks about 1000 components deployed to about 8000 machines, and we are looking at extending it to also tracking our data collection and analysis pipelines, as well as possible client-side features.