ben-bradley / generator-micro

A Yeoman generator for creating a ES6-capable, Hapi-based, REST microservices API.
11 stars 1 forks source link

generator-micro Build Status

This is still in development, use it with caution and please submit any bugs =)

Goals

The goal of this generator is to make it possible to spin up scalable, RESTful API microservices that let you focus on writing the business logic and not scaffolding out your infrastructure.

The components are built with the goal of staying as modular and composable as possible so that there are clear scopes of responsibility and that code is functional and self-contained.

generator-micro leverages Hapi's plugin framework to keep your microservices isolated and modular. Within a particular microservice, the route, handler, controller model is followed to enable DRY and ease of maintenance. Using Hapi plugins and package.json to manage dependencies and versions makes building and refactoring services fairly painless.

Generators

The micro generator has three main templates:

  1. yo micro - The server framework
  2. yo micro:service - The microservice within a server
  3. yo micro:module - An element to be CRUDed within the microservice

Server Framework

Running yo micro will build your basic server framework. It includes a gulp workflow that will build the ES6/2015 into ES5, run tests, and start the server with nodemon.

The server is responsible for:

Which microservices are deployed is based on the dependencies in your package.json. The server will read through the node_modules/ folder to find any folders that match the glob 'node_modules/microservice-*/index.js' and attempt to load them as Hapi plugins.

yo micro

This builds the server framework into which you will npm install or npm link your microservices. The app that this scaffolds is intended to be a codebase that is separate from your microservice codebase.

The following example assumes that you've already built out the microservice-awesome microservice and have published it to NPM, though you could just as easily put it in any git repo. While developing the microservice, I'd recommend using npm link to simplify dev/test.

$ cd ~/gits
$ mkdir awesomeServer && cd $_
$ yo micro
# ... answer the Yeoman's questions
$ npm link microservice-awesome # <- for development
# npm install --save microservice-awesome # <- for deploys
$ gulp develop # <- compile the ES6, watch /src & /test, & start the server

Gulps

Microservice

A microservice is really just a Hapi plugin. By leveraging this framework, it's possible to easily manage version control, deployment, and scalability while keeping the code base functional and modular.

Microservices are intended to be self-contained packages that can be tested and developed independently of each other even if they're hosted on the same Server.

Versioning & Routing

Microservice versioning is managed by reading the value in the major position of the version value in its package.json. For example, if your package.json looks like this:

{
  "name": "microservice-awesome",
  "version": "1.0.0",
  "prefix": "awesome",
  "description": "The Awesome microservice!",
    ...
}

Then all of the routes that the awesome microservice creates would be prefixed with awesome/1/

Calling the foo module in the awesome microservice looks like this:

GET http://localhost:3000/awesome/1/foo

yo micro:service

You should create a separate repo in which to build your microservice. Doing this enables the server framework to leverage API versioning, and dependency management that NPM provides, and modularity required to rapidly create and modify a microservice.

$ cd ~/gits
$ mkdir microservice-awesome && cd $_
$ yo micro:service
# ... answer the Yeoman's questions
$ npm link # <- for development
$ gulp develop # <- compile ES6 and watch /src & /test

Modules

A module is considered to be a combination of router, handler, controller and test. In the future a model may be added, but that's dependent on how patterns emerge.

A module is what manages the life-cycle of a request, each component has a specific part to play (scope of responsibility):

router

handler

controller

test

yo micro:module

You should think of a module as an object to be CRUDed. When you run yo micro:module from the microservice root directory, you'll get a new set of router, handler, controller, and test files to build out. They'll come pre-built with basic CRUD operations for reference, but you should absolutely replace them with your specific environment's needs.

$ cd ~/gits/microservice-awesome
$ yo micro:module
# ... answer the Yeoman's questions
# start hacking!