I wanted to deploy a node lambda on AWS using CDK. The NodejsFunction seemed like an improvement over using the standard Function construct: the CDK SDK would bundle your code for you, in a docker container running the same version of node used in the lambda. Instead of defining a code param you define an entry file that is used for the compilation process. This seems like a great improvement over copy/pasting your local nodejs application (including the local node_modules!) into the lambda.

However, I ran into trouble using the new NodejsFunction with my folder structure:

├── aws
│   ├── bin
│   │   └── app.ts
│   ├── cdk.context.json
│   ├── cdk.json
│   ├── jest.config.js
│   ├── lib
│   │   └── app-stack.ts
│   ├── package-lock.json
│   ├── package.json
│   └── tsconfig.json
├── scraper
│   ├── index.js
│   ├── package-lock.json
│   ├── package.json

Here are some of the errors I ran into:

• Error: Expected depsLockFilePath: /Users/mike/Projects/app/aws/package-lock.json to be under projectRoot: /Users/mike/Projects/app/scraper (../aws/package-lock.json)
• Error: Cannot find entry file at ../../scraper/index.js
• Error: Failed to bundle asset AppStack/lambdaName/Code/Stage, bundle output is located at /Users/mike/Projects/app/aws/cdk.out/bundling-temp-bf573289455d3d7a0be3d679feccc06baffacf40ec876dfab54c1a694020c731-error: Error: docker exited with status 1

What I found is that if you are creating a NodejsFunction and it’s not in the same folder as the CDK application (which I find to be the most ergonomic) you have to define more than just entry:

const lambdaAppDir = path.resolve(__dirname, '../../scraper')
const sourceLocationParams: aws_lambda_nodejs.NodejsFunctionProps = {
  projectRoot: lambdaAppDir,
  entry: path.join(lambdaAppDir, 'index.js'),
  depsLockFilePath: path.join(lambdaAppDir, 'package-lock.json'),
}

const scrapeFunction = new aws_lambda_nodejs.NodejsFunction(this, "scraper", {
  ...sourceLocationParams,

  runtime: aws_lambda.Runtime.NODEJS_18_X,
  ...
})