Deploy FastAPI Application to AWS Lambda with Docker Documentation

Why use Lambda

Pros

• Pay per use. No charge incurred if there is no activity. Cheaper for most use case.
• Easy integration with other AWS services. Can invoke using function's url or other AWS activities.
• Can configure memory and /tmp storage.
• Can use docker (max of 10 GB of project size).
• Built-in function URL.

Limitation

• Can only be run for max 15 minutes. If your operation takes longer, use other services like ECS.
• Cannot receive large files >6MB directly. Must use workaround such as upload to S3 first and invoke the function based on S3 activity. Discussion on that: https://www.reddit.com/r/aws/comments/yds4u0/aws_lambdas_function_url_size_limit/
• Hard to test code locally.

Steps

1. Setup project

   • Wrap fastapi

     • Run pip install mangum
     • Wrap fastapi with mangum. Refer sample code: main.py

   • Create dockerfile. It is not the normal dockerfile but it is built for for Lambda environment

     • Dockerfile for Lambda

   • Create aws.yml file for Github Action (You need to complete steps 2, 3, 4 first)

     • Create a file .github/workflow/aws.yml
     • Refer aws.yml. Follow the instruction and dont forget to change ENV variables to the correct value

2. Create a Github repository

   This repository will store your project's code

   • In the repository's setting, navigate to Secrets
   • Insert AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY. You can get the key from your aws account/dedicated production account. The account need to have programmatic access. Refer https://docs.aws.amazon.com/powershell/latest/userguide/pstools-appendix-sign-up.html

3. Create AWS ECR Repository

   • Go to https://ap-southeast-1.console.aws.amazon.com/ecr/repositories?region=ap-southeast-1
   • Create new repository. Make sure the repo is private

4. Create Lambda function

   • Go to https://ap-southeast-1.console.aws.amazon.com/lambda/home?region=ap-southeast-1#/create/function
   • Select Container image
   • Insert function name
   • For Container Image URI, select any image (doesn't really matter because we will overwrite this later). If you don't have any images, refer https://docs.aws.amazon.com/AmazonECR/latest/userguide/docker-push-ecr-image.html to push image manually.
   • For architecture, select x86_64. You could pick arm64, it's cheaper but you need to build a docker image for arm64. Currently, we just gonna create the x86_64 one

5. Create Function URL

   • Go to Lambda function's configuration tab
   • Go to Function URL tab
   • Set relevant CORS information

Resources

• https://github.com/gbdevw/python-fastapi-aws-lambda-container
• https://medium.com/analytics-vidhya/python-fastapi-and-aws-lambda-container-3e524c586f01
• https://www.youtube.com/watch?v=V-ac_ZvdAW4 (To apply API KEY authorization.)

ARCHIVE

5. Create API Gateway No longer relevant since AWS introduced built-in function url

   • Go to https://ap-southeast-1.console.aws.amazon.com/apigateway/main/precreate?region=ap-southeast-1

   • Click Build on REST API

   • Give it a name and leave everything as default and click Create API

     Here, the path (/, /ping) is called as resources

   • In / resource, create an ANY method Integration type: Lambda Function, Use Lambda Proxy integration: ✅, Select relevant region and the lambda function you just created. Click Save.

   • In / resource, create a new resource. Configure as proxy resource: ✅ You can give it any name, default also ok. Click Create resource.

   • In the new resource (probably /{proxy+}), create an ANY method. For integration type, select Lambda Function Proxy Select relevant region and the lambda function you just created. Click Save

   • Click Deploy API. Select or create a stage. You can name it prod or staging or whatever Click Deploy