HumanTalks : Microservices

This repository contains the code I used at the May's HumanTalk conference (Lyon).

It contains two services :

• The Front service which redirect and load-balance any request received on /echo
• The Echo service which echo any body sent to it (adding its IP adress so one is able to identify which instance answered the request).

To run the system, just type :

git clone https://github.com/MaximeGaudin/HT-Microservices.git && cd HT-Microservices && mvn clean install && docker-compose up

It will spawn both of the service and a special instance of the echo service that is a bit slower to demonstrate the latency-aware load-balancing. It can be pretty long depending on how many maven repositories you already pulled and how many docker base image are cached on your machine.

Why is it so slow to start ?

Because of the architecture, each instance has to register to the service registry (Eureka). By design, it could take at most 3 heartbeats (a heartbeat is sent every 3 secondes) to register an instance. Added to the service startup time, one may wait a full minute before the API is available. For more information, please see here.

Can I scale the echo service up ?

Yes : docker-compose scale echo=3 for 3 instances.

Weird, when I scale down, every containers shut down

Bug report here.

The speaker service

Because @jdray is speaking right after me, I added a dummy endpoint which send the list of every 2015 HT speaks. To request it, run the system and perform the following request : GET http://localhost:8081/echo/speakers

For instance :

docker-compose up -d && until $(curl --output /dev/null --silent --head --fail http://localhost:8081/echo/speakers); do; printf '.'; sleep 1; done && curl -XGET 'http://localhost:8081/echo/speakers'

Requirements

• Docker >= 1.5 : wget -qO- https://get.docker.com/ | sh
• Docker-compose : sudo pip install docker-compose
• Maven >= 3.0
• Java 8