The above approach used to run Pods is not the best one. We used the imperative way to tell Kubernetes what to do. Even though there are cases when that might be useful, most of the time we want to leverage the declarative approach.

​

We want to have a way to define what we need in a file and pass that information to Kubernetes. That way, we can have a documented and repeatable process, that can (and should) be version controlled as well.

​

Moreover, the kubectl run was reasonably simple. In real life, we need to declare much more than the name of the deployment and the image. Commands like kubectl can quickly become too long and, in many cases, very complicated. Instead, we’ll write specifications in YAML format.

​

​