Kubernetes Batch + HPC Day North America 2022

Managed Kubernetes offerings provide users with a simple way to automatically deploy, scale, and manage Kubernetes - generally everything you need to quickly deploy a production ready Kubernetes cluster. However, as a cluster operator, you are responsible for more than simply deploying containerized business logic. When you adopt a product with a managed life cycle you need to know what exactly to plan for: more specifically, where does your responsibility end and where does the provider’s responsibility start? When new Kubernetes versions are released, is it your responsibility as an operator to update your control plane? How about your worker nodes? The nature of Kubernetes as a tool with a control plane and a data plane further complicates things. For example, users generally are not responsible for managing and operating control plane components including kube-apiserver, kube-controller-manager, kube-scheduler, or etcd, but worker nodes generally exist in user tenancies and because nodes execute private code and store sensitive data providers’ access is limited. This talk will explain the support boundaries and shared responsibility of a managed Kubernetes service through the eyes of a cloud provider. It will advise users where to look for information about the parts of their system in need of care and feeding and those that can be comfortably trusted to a knowledgeable provider.