This project automates the provisioning of scalable and resilient public cloud infrastructure on AWS using Terraform. It is designed to deploy a simple Python-based RESTful API, supported by a robust backend infrastructure that includes an Application Load Balancer (ALB), Auto Scaling EC2 instances, load testing with Locust, and an RDS PostgreSQL database.
The FastAPI Python code is available in this repository.
The Locust Python code is available in this repository.
Reasons for choosing North Virginia:
- Maturity and Service Range: As one of AWS's oldest and most mature regions, North Virginia offers a comprehensive array of services. This is crucial for our project which relies on various AWS services like EC2, RDS, and ALB.
- Cost-Effectiveness: Due to its scale, North Virginia often offers competitive pricing, which is beneficial for managing the project budget effectively.
- Rich Community and Ecosystem: The strong AWS user community in this region is an excellent resource for support, best practices, and even talent acquisition.
- Data Center Density: The high density of data centers in this region ensures robust infrastructure support. This aspect is vital for the high availability and resilience required by our application.
- VPC: A custom Virtual Private Cloud (VPC) is set up with a CIDR block of
10.0.0.0/16. - Subnets: Includes both public (
10.0.1.0/24,10.0.2.0/24) and private (10.0.3.0/24,10.0.4.0/24) subnets spread across two availability zones for high availability.
- Configured to distribute incoming traffic across EC2 instances in private subnets, ensuring efficient load handling and fault tolerance.
- EC2 instances are deployed within the private subnets, running a Python RESTful API.
- Auto Scaling is configured to automatically adjust the number of instances based on load with AWS CloudWatch Metric Alarms, ensuring efficient resource utilization.
- Application running on EC2 instances is monitored using AWS CloudWatch logs.
- Auto-scaling group with a minimal size of 2 to facilitate load balancer testing.
- Auto-scaling group launches new EC2 instance when CPU utilization reaches 70% or ALB Request Count Per Target reaches 200. It takes around 3 minutes of activity to start up-scaling
- Auto-scaling group ends EC2 instance when ALB Request Count Per Target reaches 160. It takes around 15 minutes of inactivity to start down-scaling.
- Auto-scaling thresholds were defined to allow the demonstration of its operation.
- A MySQL database instance is provisioned in the private subnets, offering secure and scalable database services.
- Configured for multi-AZ deployments for high availability and automated backups for data durability.
- Database's user and password are managed with AWS Secrets Manager.
- Defined for ALB, EC2 with auto-scaling instances, EC2 for Locust instance, and RDS to ensure secure access control.
- The ALB security group allows HTTP/HTTPS traffic, whereas the EC2 security group permits traffic from the ALB. The RDS security group allows database connections from EC2 instances. The Locust security group allows all traffic.
- Internet Gateways are set up in each public subnet to enable outbound internet access to the load balancer.
- NAT Gateways are set up in each private subnet to enable the load balancer to redirect traffic to resources in the private subnets.
- Route Tables are configured for both public and private subnets to control network routing.
- All resources are defined and managed using Terraform, providing a reliable and repeatable process for infrastructure deployment.
- The infrastructure's configuration is modularized for better organization and easier maintenance.
To deploy this infrastructure, ensure you have Terraform installed and configured with AWS credentials of an account with access to S3 bucket. Follow the steps below:
- Make sure you have Terraform (v1.6.3) installed and configured with AWS credentials
- Initialize Terraform:
terraform init - Plan the deployment:
terraform plan - Apply the configuration:
terraform apply -auto-approve - Wait for AWS to finish launching instances. This may take a few minutes.
- Copy the load balancer's DNS from terminal outputs and paste it into the browser
- Test API endpoints
- Copy the ec2-locust instance's DNS from terminal outputs and paste it into the browser
- Run a load test with 100 users and a spawn rate of 100 to see autoscaling launch a new instance. This may take a few (a bit more) minutes.
- Destroy the infrastructure:
terraform destroy -auto-approve