brendanbeck.com

Routing

All api requests go through /api/

All front end and static requests go through / (the root)

Meaning if you go to brendanbeck.com/gallery it will take you to the gallery page with html, css, javascript etc.. but if you go to brendanbeck.com/api/gallery` it will not grab html.. the route assumes you are trying to access an api functionality.

HTML Pages

All html pages are in the /views/pages/ directory. These are ejs files which call partials in ejs to include html in other pages.

Static Files (CSS, Javascript)

When an html page links to a stylesheet or some script, express serves the static files in the /static/ directory within the same direcory it is called inside of the /views/pages/ direcory.

For example:

/views/pages/index.ejs has a stylesheet href="css/index.css".

Since express serves our static files for us it actually grabs the css from /static/css/page.css.

This makes it easy to use relative addresses in hrefs of html files, so long as you continue this directory structure of placing static files in /static directory, the html in the views/pages/ directory, and the reusable html in the /views/partials/ directory.

Dev

Git submodules

As the brendanbeck.com enterprise grows and I transition to a microservice architecture, some files need to be shared to maintain consistency. To acheive this, I use a couple git submodules.

NPM

server.js is the main entry point

"scripts": {
    "dev": "NODE_ENV=dev nodemon server.js",
    "start": "NODE_ENV=prod node server.js"
  },

.env files

NODE_ENV is set in package.json, can also just be exported.

The NODE_ENV corresponds to the suffix of a corresponding .env file.

⚠️ Localhost for docker containers: 127.0.01 refers to the network interface of the host. When mapping a docker container locally, use 0.0.0.0. https://stackoverflow.com/a/59182290

Sample:

.env.dev:
HOST=0.0.0.0
PORT=8080

.env.prod:
HOST=127.0.0.1
PORT=80

AWS Infra

Layed out pretty well in the terraform.

1. Containers are published to ECR
2. Target group is 3 container instances stood up in ecs Fargate across 3 subnets
3. ECS Security group only allows traffic from ALB Security group
4. Application load balancer forwarding port 80 to it's target group
5. DNS is configured by Cloudflare, pointing to the DNS name of the ALB.

There are currently 2 workspaces configured, prod and test. The idea is that any infra changes occur in the test workspace for testing, and then once PRd into master the prod workspace can be caughtup.

I typically spin down the test workspace after I am done working on it, to reduce cost (with the exception of ECR)

• /registry has a dev and prod, creates the ecr
• /s3 single instance for static files to make the docker container smaller

TODO

• Github action on PR to run terraform apply on prod workspace
• Better tagging practice for containerized node app
• JWT server to put put other subdomains behind a login Devon Crawford Video