You (probably) need liveness and readiness probes

Note: This article was originally published by the author on the Red Hat Developer Blog. You can read the original on the Red Hat Developer Blog here.

One of the most common questions I get as a consultant is, “What is the difference between a liveness and a readiness probe?” The next most frequent question is, “Which one does my application need?”

Anyone who has tried Duck Duck Go-ing these questions knows that they are difficult to answer using an internet search. In this article, I hope to help you answer these questions for yourself. I will share my opinion about the best way to use liveness and readiness probes in applications deployed to Red Hat OpenShift. I’m not offering a hard prescription but rather a general framework that you can use to make your own architectural decisions. Each application is different, and these differences might require adapting the “rules” you learn here.

To help make the abstract more concrete, I offer four generic example applications. For each one, we’ll explore whether and how to configure liveness and readiness probes. Before we dive into the examples, let’s look more closely at the two different probe types.

Note: Kubernetes has recently adopted a new “startup” probe available in OpenShift 4.5 clusters. The startup probe does not replace liveness and readiness probes. You’ll quickly understand the startup probe once you understand liveness and readiness probes. I won’t cover startup probes here.