Vinícius A dos Santos

Vinícius A dos Santos

About MeEmail Me

AWS API Gateway + Terraform + Serverless Framework - Part 1

aws api gateway
rest api

Hi, everyone! For this "Hands on!" we're going to build a REST API with AWS API Gateway, provisioned with Terraform and backed by AWS Lambda built with Serverless Framework.
The REST API will allow us to send SMS Messages using AWS SNS. Sounds like a lot of things, but it's not that lot of work.
For this part 1, we'll provision our API Gateway with Terraform, and for parts 2 and 3:

Part 2: coding the backend with Serverless Framework
Part 3: securing the API with Amazon Cognito

Tech Stack

  • AWS: Most popular Cloud provider. You need an account to follow this article properly;
  • AWS API Gateway: AWS managed API Gateway that will expose our rest endpoints;
  • AWS Lambda: serverless functions on AWS working as our backend;
  • AWS SNS: AWS Simple Notification Service that, among other types of notifications, allows us to send SMS for a phone number;
  • Terraform: IaC (Infrastructure as Code) tool that allows us to provision cloud resources supporting several cloud providers;
  • Serverless Framework: a Framework for support building and deploying serverless functions grouped as a Serverless Service, allowing also the provisioning of resources needed for these functions;
  • NodeJS: JS runtime where our JavaScript lambda functions going to be running;
  • JavaScript: Of course, the programming language we'll write our lambda.

Why Terraform, since Serverless Framework itself can provision AWS resources?

Well, it's very common to have an API Gateway serving different backend services.
Ex: endpoint X points to a serverless function, endpoint Y points to Java Spring Boot Microservice, and endpoint Z points to third-party API.
If the API Gateway gets provisioned by the serverless Service, its infrastructure code going to be highly coupled to this service, and then you need to implement the other types of integrations using this service as a proxy (creating more lambdas and increasing the infrastructure cost and complexity).
Also, if you destroy this Service, the API Gateway will be destroyed too.

So, hands-on!

Terraform AWS API Gateway provisioning

If you haven't Terraform installed, do it:
When provisioning resources, the docs help a lot:
Nobody should try to memorize the code for a bunch of services AWS provides.

First, we need to set the cloud provider configurations.
Let's create the file like this:

provider “aws” {
    region = “us-east-1    version = “2.61    access_key = "your-access-key”
    secret_key = "your-secret-key” 

We are telling Terraform that our provider is AWS, and we want to provision resources on us-east-1 region. We're also setting AWS credentials having the roles and policies needed to manage API Gateway (we could set the credentials as environment variables as well). It's important to set the version of the terraform AWS plugin to avoid upgrades with breaking changes issues. Now, we can start terraform in this project running in the terminal:

$ terraform init

You should see the message Terraform has been successfully initialized!.
To finally provision an API Gateway, let's create another .tf file as follows:

resource "aws_api_gateway_rest_api" "my_api_gateway" {
  name        = "My API"
  description = "AWS Rest API example with Terraform"
  endpoint_configuration {
    types = ["REGIONAL"]
  • resource: provision a cloud resource to be managed by Terraform;
  • aws_api_gateway_rest_api: the type of resource;
  • my_api_gateway: a name to reference this resource in this project.

    The content inside the block refers to attributes of the API Gateway. You can look at AWS API Gateway if you are not familiar.

We can now execute the command below to check what Terraform is going to do after we apply the provisioning plan:

$ terraform plan

Among other output messages, you should see:

Terraform will perform the following actions:
  # aws_api_gateway_rest_api.my_api_gateway will be created
  + resource "aws_api_gateway_rest_api" "my_api_gateway" {
      + api_key_source           = "HEADER"
      + arn                      = (known after apply)
      + created_date             = (known after apply)
      + description              = "AWS Rest API example with Terraform"
      + execution_arn            = (known after apply)
      + id                       = (known after apply)
      + minimum_compression_size = -1
      + name                     = "My API"
      + root_resource_id         = (known after apply)
      + endpoint_configuration {
          + types = [
              + "REGIONAL",
Plan: 1 to add, 0 to change, 0 to destroy.

As we can see in the last line, Terraform going to add 1 resource, change none, and destroy none.
It's not guaranteed that this plan will be the same applied. So we could use $ terraform plan -out, but this plan is pretty simple.
The command to apply the changes is:

$ terraform apply

Then the exact plan to be executed will be presented again. You can answer yes.
Going to AWS Console and looking for API Gateway Resource (region us-east-1), we can see the API created:

However, clicking on it, there are no API resources paths:

Let's define that our API base path going to be <api-url>/my-api/v1. Back to Terraform files, we add this code:

resource "aws_api_gateway_rest_api" "my_api_gateway" {

resource "aws_api_gateway_resource" "my_api" {
  rest_api_id =
  parent_id   = aws_api_gateway_rest_api.my_api_gateway.root_resource_id
  path_part   = "my-api"

resource "aws_api_gateway_resource" "v1" {
  rest_api_id =
  parent_id   =
  path_part   = "v1"
  • aws_api_gateway_resource: the type of resource is API gateway resource;
  • rest_api_id: the id of the API Gateway that owns this resource;
  • parent_id: the id of the parent path, which is the root "/" of the API Gateway;
  • path_part: the path part we want to create.

Note that for the "v1" resource, the parent_id is the id of the "my_api" resource, thus the complete path going to be "/my-api/v1".
Running apply again:

$ terraform apply

We should see on AWS Console three API resources ("/", "my-api" and "v1"):

That's it for this post. In part 2 we're going to implement the backend with Serverless and in part 3 we'll implement authentication with Amazon Cognito:

← Back to home