Automated Multi-Cloud Multi-Flavor Kubernetes Cluster Upgrades Using Operators

Learn how Databricks manages and automates Kubernetes cluster upgrades across multiple cloud providers in this technical conference talk. Explore the architecture of a system built on Kubernetes operators that handles zero-downtime blue-green rolling updates for over a thousand clusters running on EKS, AKS, GKE, and self-managed Kubernetes environments. Discover implementation details for operating system updates, Kubernetes version management, and configuration changes while maintaining service availability through features like PDBs, maintenance windows, deferred node draining, and custom workload handling plugins. Gain insights into building collaborative operators using the controller-runtime framework, designing declarative interfaces, and achieving consistent behavior across AWS, Azure, and Google Cloud platforms. Master the techniques for implementing reliable infrastructure changes, vulnerability patching, and maintaining robust observability with minimal operational overhead.