1. Can I create a Master / Slave kind of an architecture with Kubernetes? Say I have Microservice A running as a Docker container. Can I have Kubernetes manage this as a Master and have 2 or 3 other instances as slaves?
To answer the first question, Kubernetes does not use the concept of containers directly and instead the atomic
unit in Kubernetes is a Pod, which could consist of one or more containers for different Docker images. A cluster in Kubernetes consists of one or more Pods and a Kubernetes service could be used as an interface for the Pod/s. Clients do not access the Pods directly but do so via the service. A service is exposed on one or more endpoints corresponding to the number of Pods managed by the service; one Pod-one endpoint, two Pods-two endpoints. If Microservice A is running, it could be managed on a Master/Slave architecture, more specifically a Master/Workers architecture, which consists of two or more nodes. Multi-node architecture is discussed in Chapter 14.
2. If the Microservice A is exposing ports, how would this behave with rudundancy? Should I have multiple ports open?
You could have multiple ports for the different interfaces provided by an application. For example, for Oracle Database you could have port 1521 for the TCP/IP database listener and port 8080 as the HTTP port. Redundancy is provided by the multiple Pods in the microservice. By scaling a cluster to 10 Pods for example, 10 different clients could access the service. The clients access one of the endpoints provided by the service and the service forwards the request to one of the Pods managed by the service, selecting a Pod in a round-robin manner. Chapter 2 discusses how scaling up a cluster of Pods adds endpoints.
