KubeCon + CloudNativeCon North America 2024

Kubernetes once favored Etcd as a database for all cluster data. Back then, relational databases lacked the availability and scalability characteristics required by Kubernetes. However, as Etcd encountered challenges with various Kubernetes workloads, relational databases continued to evolve. This session is a practical guide for deploying fault-tolerant and scalable Kubernetes clusters on distributed PostgreSQL. We’ll begin with Kine, which integrates into the Kubernetes architecture, enabling relational databases for cluster metadata management. Then, we’ll use Kine to deploy Kubernetes on a single-server PostgreSQL instance. After that, we’ll migrate to a multi-node PostgreSQL instance, allowing Kubernetes to tolerate zone and region outages and scale to thousands of nodes on demand.