Maintaining configurations for different kinds of nodes and cloud resources in a [micro]service architecture can be an operational nightmare, especially if not managed with the application codebase. CI and CD job environments diverge from production configuration yielding their results unpredictable at best or produce false positives in the worst case. Code pushes to staging and production can have unintended consequences that can't be reasoned about before deploy and often can’t be inspected thoroughly on a dry run. Leading to unhappy users when problems do arise.
This session will demonstrate the use of the Nix and NixOS ecosystem to define and build packages in a referentially transparent way which can be leveraged as a solid foundation to configure systems and test multiple [virtual] machines with coordinated scenarios. We also look at how reliable packaging allows us to build a consistent CI/CD pipeline where upgrading your version of the JVM doesn't break your CI build servers for days.