lambda-aws

This repository is the starting point for a Infrastructure as Code pipeline.

AWS Lambda Function

The repository contains a AWS Lambda Function written in Python3 that processes AWS Kinesis Data Stream logs and writes the processed results into a AWS DynamoDB table.

Example of input before processing:

{
    "id": 33,
    "message": "This function will count the number of occurrences of the word error."
}

Example of input after processing:

{
    "id": 33,
    "message": "This function will count the number of occurrences of the word error.",
    "errors": 1
}

Continuous Integration and Continuous Deployment

Each time a new Pull Request is created (and/or a new push on master), the IaC pipeline (here implemented as GitHub action) starts and deploys the whole infrastructure using Terraform:

• An AWS Kinesis Data Stream named application-logs-stream
• An AWS DynamoDB table named application-logs-table
• An AWS Lambda function named kinesis-to-dynamodb

The pipeline takes care of:

• Job Lint: Linting Lambda Python3 code.
• Job Terraform-deploy: Zipping the file lambda.py into a function.zip that will be moved to the Terraform repository and uploaded as AWS Lambda source file. The result file will also be pushed on a new branch starting from the HEAD (it will be then deleted only for Pull Requests).
• Init Terraform backend on a AWS s3 bucket.
• Applying Terraform resources in two steps:
  1. Firstly the AWS Kinesis Data Stream and the AWS DynamoDB table.
  2. Then the Lambda function.
• Job Kinesis-test: Test the whole workflow by installing and configuring the AWS Kinesis Agent on the GitHub runner, using repository secrets divided into a repository environment. Then it runs a test Python application that produces 3 log lines that will be provided to the AWS Kinesis Data Stream using the AWS Kinesis Agent.
• Job Terraform-destroy: eventually, the pipeline will clean up the infrastructure by destroying all Terraform generated resources and, in the case of a Pull Request, remove the branch that contains the function.zip since it will be created again once merged.

Terraform destroy should not be executed when the pipeline refers to master/release but for convenience, it is destroyed all the times.

Additionally, a Jenkinsfile example to deploy the workflow is provided in the repository.

To see more about the Terraform code, visit the below repository.

References

• Python Application that produces logs -> kinesis-application
• Terraform workflow -> terraform-lambda-aws