Interested in learning more about serverless AWS? Check out my Guide to Building with Serverless AWS (only pay what you want)
I have written a version of this script multiple times over, so I finally made a generic enough version to hopefully fulfill my needs now into the future. There are some prerequisites/assumptions that I make. First is that there is an existing CloudFormation stack with a Lambda function that uses a versioned S3 bucket to grab its source. The template would look something like this
Parameters: SourceBucket: Type: String Description: S3 bucket with source SourceKey: Type: String Description: S3 key with Lambda source code SourceVersion: Type: String Description: S3 version for source code Resources: LambdaFunction: Type: AWS::Lambda::Function Properties: Code: S3Bucket: !Ref SourceBucket S3Key: !Ref SourceKey S3ObjectVersion: !Ref SourceVersion ...
The second is that the code has all dependencies installed and it is all contained in the same directory. (Something similar can be done with more complex projects/codebases and this would be a good starting point for building a script for something like that). The third is that
jq is installed and the AWS CLI is installed and ready to be used.
#!/bin/sh # NOTE: Requires jq and correctly configured aws cli # Change these to the correct values SOURCE_PATH= # i.e. ./src SOURCE_BUCKET_NAME= # i.e. source-bucket SOURCE_KEY= # i.e. my-application/service-name TEMPLATE_PATH= # i.e. ./template.yml STACK_NAME= # i.e. my-stack STACK_SOURCE_VERSION_PARAMETER_NAME= # i.e. SourceVersion zip -r9 deployment-package $SOURCE_PATH aws s3 cp deployment-package.zip s3://$SOURCE_BUCKET_NAME/$SOURCE_KEY S3_VERSION=$(aws s3api get-object --bucket $SOURCE_BUCKET_NAME --key $SOURCE_KEY getobjectoutfile | jq -r '.VersionId') aws cloudformation deploy --template-file $TEMPLATE_PATH --stack-name $STACK_NAME --parameter-overrides $STACK_SOURCE_VERSION_PARAMETER_NAME=$S3_VERSION
After the six variables are properly filled in, the script should properly zip everything up, upload the deployment package to S3, grab the new version ID, and deploy the changes to the CloudFormation template. Voila!
For anyone reading this and wondering how to build all of this infrastructure, I do have some templates that would be a good starting point. For the versioned S3 bucket to hold deployment packages, I suggest using this template. For an API that uses a Lambda function for computing, I suggest using this template. It also contains the correct parameters to use a deployment package from a versioned S3 bucket. For anyone trying to learn AWS and wanting some additional help figuring things out, please don’t hesitate to get in contact with me. I would be more than happy to help out where I can!Categories: aws | ops | databases