Masters host the control plane. That is a fancy way of referring to the brains of the cluster.

With this in mind, it is good practice to have more than one Master in order to maintain high availability (HA). This way, if one of them fails, the cluster can remain operational. It is common to have 3 or 5 Masters in a production cluster and to spread them across failure domains. It is not wise to stick them all in the same room, under the same leaky air conditioning unit, functioning on the same glitchy electricity supply.

The illustration below shows a highly available control plane with 3 Masters. Each one is in a separate failure domain with different network infrastructures and distinct power infrastructures, etc.

Control plane

Masters run the following services that form the control plane:

API Server

Scheduler

Store

Cloud controller  

Masters run the back-end services that keep the cluster running. They run services that form the control plane (brains of the clusters), namely:

API Server

Scheduler

Store

Cloud controller

Masters can also run other services, but those are not relevant to the subject matter of this course.