How to Develop a Robust Operator for Day-2 (Lesson Learned on KubeVirt/HCO)

This presentation shares lessons learned from developing the Hyperconverged Cluster Operator (HCO) for the KubeVirt project, focusing on the complexities of day-2 operations for virtual machines. It covers essential aspects like control plane vs. workload upgrades, long-running upgrades, reliability concerns such as canary and fail-forward upgrades, and strategies for managing new and deprecated APIs. Attendees will learn how to implement declarative approaches, handle feature-gates, discriminate defaults, and maintain a manageable upgrade graph to build trustworthy operators for automatic and continuous upgrades.