If a service in production goes down, what immediate steps would you take to diagnose and resolve the issue?