CERN IT operates Ironic as their bare metal management platform since 2017 and has encountered several operational challenges during this time. This presentation will review the top five of them and explain how we mastered (or sometimes only get along with) them: transparent adoption of pre-Ironic production nodes into the system, user-facing resource overview / location specific scheduling of physical instances, update of missing inventory data with active introspection, reduction of the hardware manager development cycle, and scaling with conductor groups (with fast-track bonus). We will also share our “GRUBsetta” stone, our collection of boot errors and what they really mean.