Production Kubernetes

Chapter 2. Deployment Models

The first step to using Kubernetes in production is obvious: make Kubernetes exist. This includes installing systems to provision Kubernetes clusters and to manage future upgrades. Being that Kubernetes is a distributed software system, deploying Kubernetes largely boils down to a software installation exercise. The important difference compared with most other software installs is that Kubernetes is intrinsically tied to the infrastructure. As such, the software installation and the infrastructure it’s being installed on need to be simultaneously solved for.

In this chapter we will first address preliminary questions around deploying Kubernetes clusters and how much you should leverage managed services and existing products or projects. For those that heavily leverage existing services, products, and projects, most of this chapter may not be of interest because about 90% of the content in this chapter covers how to approach custom automation. This chapter can still be of interest if you are evaluating tools for deploying Kubernetes so that you can reason about the different approaches available. For those in the uncommon position of having to build custom automation for deploying Kubernetes, we will address overarching architectural concerns, including special considerations for etcd as well as how to manage the various clusters under management. We will also look at useful patterns for managing the various software installations as well as the infrastructure ...