So far, we know how to deploy applications in pods. The sole concept of pod is very powerful, but it lacks robustness. It is actually impossible to define scaling policies or even make sure that the pods remain alive if something happens (such as a node going down). This might be okay in some situations, but here is an interesting question. If we are biting the bullet on the overhead of maintaining a Kubernetes cluster, why don't we take the benefits of it?
In order to do that, we need to work with Replica Sets. A Replica Set is like a traffic cop in a road full of pods: they make sure that the traffic flows and everything works without crashing and moving the pods around so that we make the best use of the road (our cluster, ...
