Agrarian-Scale Kubernetes: Part 1

Visit the Kubernetes homepage, and one of the first things you’ll see is the promise of “Planet Scale.” The S-word is plastered across the pitch for Google’s managed version of Kubernetes as well (Google Kubernetes Engine, or “GKE”). Demand-based autoscaling! Support for thousands of host nodes! Cross-region federated clusters! The ability to scale easily to heavy workloads, across geographies and even multiple clouds makes Kubernetes and GKE great choices for high-traffic websites.

Because so many of the apps we create on the Interactive News Team are used only by reporters and editors in the building, we tend to take greater advantage of the less discussed dimension of scaling: fitting many applications into a single cluster. Consider our concerns on a kind of agrarian scale: a single plot of land (our newsroom) where our biggest concerns are crop rotation.

This guide is an outline of our devops philosophy on Interactive News at The New York Times, by way of tutorial. It assumes you can copy code into a command line, and that you have a rough understanding of what containers are. (Docker has a brief overview here.) You don’t need a background in cloud infrastructure.

Read more at Times Open