serverless-multi-lambda-layers v1.0.0

PinataFarms multi-lambda-layers Serverless Plugin

This plugin allows managing node.js lambda dependencies separately and putting each of them into a separate lambda layer.

Objective - be able to easily keep each lambda dependencies in a separate layer.

Why?

There are two main points:

• Keeping lambdas lean (attaching small layers only with needed dependencies)
• Reducing probability of reaching the hard limit of 250MB lambda size (unzipped layers + lambda code)

Thanks to the above-mentioned points we are able to live-patch lambdas in production via the AWS console as lambda code sizes stay below the live edit limit.

How it works

Every time when I mention lambdas directory you should keep in mind that this directory could be overwritten via the plugin configuration (see section below).

This plugin expects you to maintain a similar directory structure for your Serverless service:

$ tree
├── lambdas
│   ├── lambdaA
│   │   ├── handler.js
│   │   ├── package-lock.json
│   │   └── package.json
│   └── lambdaB
│       ├── handler.js
│       ├── package-lock.json
│       └── package.json
├── package-lock.json
├── package.json
└── serverless.yml

The main service package.json file has only Serverless Framework dependencies (e.g. plugins) and npm scripts in it.

All lambdas functions are put into lambdas directory, each inside their own directory with their own package.json files where their dependencies are defined.

This plugin has 2 hooks:

• before:package:createDeploymentArtifacts - runs before creation of deployment artifacts (zip archives)
• after:package:createDeploymentArtifacts - runs after creation of deployment artifacts (zip archives)

before:package:createDeploymentArtifacts

The plugin goes over functions block in your serverless.yml file, sees lambdas with handlers pointing at lambdas directory (e.g. handler: lambdas/lambdaA/handler.hello) and processes each of them:

• removes node_modules directory if present
• installs dependencies (npm i --production)
• creates layer/nodejs directory structure
• moves node_modules directory into that layer/nodejs directory

after:package:createDeploymentArtifacts

The plugin goes over the same lambdas as in the previous hook and processes each of them:

• moves node_modules directory from layer/nodejs directory to a lambda directory
• removes layer/nodejs directory

Installation

$ npm install @pinatafarms/serverless-multi-lambda-layers

Configuration

1. Add @pinatafarms/serverless-multi-lambda-layers to plugins in your serverless.yml file

   plugins:
     - "@pinatafarms/serverless-multi-lambda-layers"

2. By default, all your lambdas should be in lambdas directory. You could change this directory by adding such configuration to your serverless.yml file

   custom:
     serverless-multi-lambda-layers:
       lambdaDir: my-lambda-dir

Example

Directory structure

$ tree
├── my-lambda-dir
│   ├── lambdaA
│   │   ├── handler.js
│   │   ├── package-lock.json
│   │   └── package.json
│   └── lambdaB
│       ├── handler.js
│       ├── package-lock.json
│       └── package.json
├── package-lock.json
├── package.json
└── serverless.yml

serverless.yml

service: example-service

frameworkVersion: '2'

plugins:
  - serverless-multi-lambda-layers

custom:
  serverless-multi-lambda-layers:
    lambdaDir: my-lambda-dir

provider:
  name: aws
  runtime: nodejs12.x
  stage: ${opt:stage, 'qa'}
  region: ${opt:region, 'us-east-1'}

package:
  individually: true

functions:
  lambdaA:
    handler: my-lambda-dir/lambdaA/handler.hello
    package:
      exclude:
        - '**/*'
      include:
        - my-lambda-dir/lambdaA/handler.js
  lambdaB:
    handler: my-lambda-dir/lambdaB/handler.hello
    package:
      exclude:
        - '**/*'
      include:
        - my-lambda-dir/lambdaB/handler.js