In the previous setup, we were always running with a single instance of the microservice. The URL is hardcoded both in client as well as in the service-to-service calls. In the real world, this is not a recommended approach, since there could be more than one service instance. If there are multiple instances, then ideally, we should use a load balancer or a local DNS server to abstract the actual instance locations, and configure an alias name or the load balancer address in the clients. The load balancer then receives the alias name, and resolves it with one of the available instances. With this approach, we can configure as many instances behind a load balancer. It also helps us to handle server failures transparent to the client.

This is achievable with Spring Cloud Netflix Ribbon. Ribbon is a client-side load balancer which can do round-robin load balancing across a set of servers. There could be other load balancing algorithms possible with the Ribbon library. Spring Cloud offers a declarative way to configure and use the Ribbon client.

As shown in the preceding diagram, the Ribbon client looks for the Config server to get the list of available microservice instances, and, by default, applies a round-robin load balancing algorithm.

In order to use the Ribbon client, we will have to add the following dependency to the pom.xml file:

<dependency>
  <groupId>org.springframework.cloud</groupId>
  <artifactId>spring-cloud-starter-ribbon</artifactId>
</dependency>

In case of development from ground up, this can be selected from the Spring Starter libraries, or from http://start.spring.io/. Ribbon is available under Cloud Routing:

Update the Booking microservice configuration file, booking-service.properties, to include a new property to keep the list of the Fare microservices:

fares-proxy.ribbon.listOfServers=localhost:8080,localhost:8081

Going back and editing the FareServiceProxy class created in the previous section to use the Ribbon client, we note that the value of the @RequestMapping annotations is changed from /get to /fares/get so that we can move the host name and port to the configuration easily:

@FeignClient(name="fares-proxy")
@RibbonClient(name="fares")
public interface FareServiceProxy {
  @RequestMapping(value = "fares/get", method=RequestMethod.GET)

We can now run two instances of the Fares microservices. Start one of them on 8080, and the other one on 8081:

java -jar -Dserver.port=8080 fares-1.0.jar
java -jar -Dserver.port=8081 fares-1.0.jar

Run the Booking microservice. When the Booking microservice is bootstrapped, the CommandLineRunner automatically inserts one booking record. This will go to the first server.

When running the website project, it calls the Booking service. This request will go to the second server.

On the Booking service, we see the following trace, which says there are two servers enlisted:

DynamicServerListLoadBalancer:{NFLoadBalancer:name=fares-proxy,current 

list of Servers=[localhost:8080, localhost:8081],Load balancer stats=Zone stats: {unknown=[Zone:unknown;  Instance count:2;  Active connections count: 0;  Circuit breaker tripped count: 0;  Active connections per server: 0.0;]
},