This course was focused on delivering the different applications/softwares as a service. We were asked to implement a QnA system with some specific functionalities. The project that we had to deliver had to be developed with two different architectures, so we decided to use MVC Architecture and Microservices Architecture.
The only constraint of the project was to use NodeJS.
To develop the project with the MVC Architecture we used:
- ExpressJS Framework
- PostgreSQL for the database
- Sequelize as an ORM
- EJS for the Views
To develop the project with the Microservices Architecture we started to make a plan on how to distribute the different functionalities of the project to different services. The distribution of functionalities to services is:
- Accounts Service: Signin, Signup
- Analytics Service: User stats (Questions Asked, Questions Answered, Contributions per Day)
- Answers Service: Get Question by ID, Answer Question by ID
- Ask Question Service: Post Question by ID
- Browse Questions Service: Get All Questions, Get All Questions asked by User with ID = id
- Event Bus Service: Notify subscribed services that a specific event was emitted
- Graphs Service: Get Top Keywords data, Get Questions Per Day data
- Front End: The presentation layer of the project
We decided that our MSA Project had to be protected, so as only the Front End that was developed by us could communicate with the services. To make this happen we made a custom Header, in which we have a SECRET-WORD that exists in the .env variables. This word is been encrypted with a specific algorithm, and is set as value of the custom Header. So when a service receives a request, it first searches for the specific custom header. It decodes its value and tries to match it with the SECRET-WORD that has also stored in the .env variables. If it matches the request continues through the service, else it gets a 401 status code. Also the service that sent the request must always have the custom header set, or else will receive a 401 status code.
The event bus was developed from scratch by our team. The Event Bus has the /events endpoint (POST). When an Event occurs (i.e USER CREATED) the service that was used to create the user also emits the event on that specific endpoint of the Event Bus. Then the Event Bus notifies all subscribers. The subscribers need to have a specific endpoint which is called /events (POST). So when the Event Bus receives a request on its /events endpoint, it makes sure to POST on every service that has the /events endpoint developed. So every service/subscriber has the event.
We make sure that every emitted event has a TYPE. The types that exist in our project are:
- USER CREATE
- QUESTION CREATE
- ANSWER CREATE
So when every service receives the event from the Event Bus, parses this TYPE and knows if has to do something with it or ignore it. With this TYPE attribute in each event object it is like we have three different channels that each subscriber listens to.
We make sure to store any occured Event in the Event Bus database. The reasons we made this decision are:
- When a service is down, it might lose some Events. So we make sure that every time a service/subscriber is (re)starting to ask for the lost events.
- If we want to develop a new service, we would like to sync it with all the past Events.
We make sure to store how many events each Service received and parsed, so as when the service/subscriber (re)starts to make a GET request on the /events/:id endpoint of the Event Bus. The /events/:id endpoint simply returns all the Events of id greated than the requested param id Also this task is performed asynchronously because the lost Events that the service eventually receives will be posted on its database (if needed) by using the async functions of the Sequelize ORM.
The deployment was done using Heroku. MVC Project was deployed as one Application, while Microservices Project was deployed as 8 different Applications