High Performance MySQL, 4th Edition

Appendix B. MySQL on Kubernetes

If you have been working in tech at all in the past five years, you very likely have heard of Kubernetes, work with teams that run Kubernetes, or have watched a lot of conference talks or read a lot of blog posts that explain Kubernetes. If your organization runs its own Kubernetes clusters, you will at some point get asked whether running MySQL on them too is a good idea. And on the surface it seems like a reasonable path to take. Managing many Kubernetes clusters is a complex task that typically needs dedicated human resources, and it is reasonable for your organization to want to leverage that expertise for more than only stateless workloads. But there are good reasons to explore running MySQL on Kubernetes and not so good reasons to do so. Let’s demystify some of the FUD (fear, uncertainty, doubt) around running MySQL on Kubernetes here.

Provisioning Resources with Kubernetes

Before Kubernetes reached peak tech popularity, a lot of companies either built entirely bespoke tech stacks for provisioning and managing VMs and bare metal servers or glued together open source projects that did smaller parts of the life cycle of a resource. Then came Kubernetes as a more complete ecosystem for managing both compute and storage resources, and the prospect of using it as the provisioning stack to rule them all has become more and more appealing. Yet stateful loads such as MySQL remained behind and left out of that added value because the common wisdom ...