Open Source Summit North America 2017

Description
With Kubernetes 1.6, etcd v3 becomes the preferred storage backbone of every Kubernetes cluster. Do you know what’s changed? How to recover from failures? This talk is a look at what’s new to etcd v3. It will act as a refresher on what failure scenarios admins need to be mindful of, in order to keep their cluster safe.

Abstract
The most notable change to etcd v3 is the introduction of gRPC. We’ll talk about why the change was made and how it affects the old REST API. Did you know that HTTP and gRPC are namespaced differently? You will after this talk, and you’ll also discover what implications it has for running your cluster.

After covering changes, we’ll take a look at common failure scenarios for etcd. We’ll discuss common misconceptions about leader election. We’ll explore the risks associated with 3 node cluster vs a 9 node, and a regional cluster vs a global. Finally, we’ll end with a live demo of how to backup your cluster, and restore from said backup.

This talk is applicable to any developer that relies on an etcd backed platform. That includes Kubernetes, and many other cloud native projects. Only a glancing familiarity with etcd and distributed consensus are required.