Skip to content

Latest commit

 

History

History
279 lines (211 loc) · 11.4 KB

File metadata and controls

279 lines (211 loc) · 11.4 KB

Introduction

CI NPM version js-standard-style

Inspired by the AWSLABS aws-serverless-express library tailor made for the Fastify web framework.

No use of internal sockets, makes use of Fastify's inject function.

Seems faster (as the name implies) than aws-serverless-express and aws-serverless-fastify 😉

👨🏻‍💻Installation

$ npm i @fastify/aws-lambda

Options

@fastify/aws-lambda can take options by passing them with : awsLambdaFastify(app, options)

property description default value
binaryMimeTypes Array of binary MimeTypes to handle []
enforceBase64 Function that receives the response and returns a boolean indicating if the response content is binary or not and should be base64-encoded undefined
serializeLambdaArguments Activate the serialization of lambda Event and Context in http header x-apigateway-event x-apigateway-context false (was true for <v2.0.0)
decorateRequest Decorates the fastify request with the lambda Event and Context request.awsLambda.event request.awsLambda.context true
decorationPropertyName The default property name for request decoration awsLambda
callbackWaitsForEmptyEventLoop See: Official Documentation undefined

📖Example

lambda.js

const awsLambdaFastify = require('@fastify/aws-lambda')
const app = require('./app')

const proxy = awsLambdaFastify(app)
// or
// const proxy = awsLambdaFastify(app, { binaryMimeTypes: ['application/octet-stream'], serializeLambdaArguments: false /* default is true */ })

exports.handler = proxy
// or
// exports.handler = (event, context, callback) => proxy(event, context, callback)
// or
// exports.handler = (event, context) => proxy(event, context)
// or
// exports.handler = async (event, context) => proxy(event, context)

app.js

const fastify = require('fastify')

const app = fastify()
app.get('/', (request, reply) => reply.send({ hello: 'world' }))

if (require.main === module) {
  // called directly i.e. "node app"
  app.listen({ port: 3000 }, (err) => {
    if (err) console.error(err)
    console.log('server listening on 3000')
  })
} else {
  // required as a module => executed on aws lambda
  module.exports = app
}

When executed in your lambda function we don't need to listen to a specific port, so we just export the app in this case. The lambda.js file will use this export.

When you execute your Fastify application like always, i.e. node app.js (the detection for this could be require.main === module), you can normally listen to your port, so you can still run your Fastify function locally.

Usage with NestJS

main.ts

import { NestFactory } from '@nestjs/core';
import {
  FastifyAdapter,
  NestFastifyApplication,
} from '@nestjs/platform-fastify';
import { AppModule } from './app.module';
import awsLambdaFastify, { PromiseHandler } from '@fastify/aws-lambda';
import fastify, { FastifyInstance, FastifyServerOptions } from 'fastify';
import { Context, APIGatewayProxyEvent } from 'aws-lambda';
import { Logger } from '@nestjs/common';

interface NestApp {
  app: NestFastifyApplication;
  instance: FastifyInstance;
}

let cachedNestApp;

async function bootstrapServer(): Promise {

  const serverOptions: FastifyServerOptions = {
      logger: (process.env.LOGGER || '0') == '1',
  };
  const instance: FastifyInstance = fastify(serverOptions);
  const app = await NestFactory.create<NestFastifyApplication>(
      AppModule,
      new FastifyAdapter(instance),
      {
          logger: !process.env.AWS_EXECUTION_ENV ? new Logger() : console,
      },
  );

  const CORS_OPTIONS = {
      origin: '*',
      allowedHeaders: '*',
      exposedHeaders: '*',
      credentials: false,
      methods: ['GET', 'PUT', 'OPTIONS', 'POST', 'DELETE'],
  };

  app.register(require('fastify-cors'), CORS_OPTIONS);

  app.setGlobalPrefix(process.env.API_PREFIX);

  await app.init();

  return {
      app,
      instance
  };
}

export const handler = async (
  event: APIGatewayProxyEvent,
  context: Context,
): Promise<PromiseHandler> => {
    if (!cachedNestApp) {
      const nestApp = await bootstrap();
      cachedNestApp = awsLambdaFastify(nestApp.instance, {
        decorateRequest: true,
      });
    }

  return cachedNestApp(event, context);
};

Notice we are caching the initialized app. In a lambda environment, the handler will be called on each new request. Anything outside the handler may be cached between calls, which lasts as long as AWS has kept the same lambda execution environment up for your function. By storing the initialized app as a variable, we can minimize the cold-start time of our app since it will be constructed once per execution environment spin up.

Usage with NestJS/GraphQL

In addition to the above, when using NestJS with GraphQL in a lambda environment, you will need to copy your schema.gql into your dist folder rather than relying on the autoSchemaFile option since this tries to write to /src directory inside a running lambda function and you are only allowed to write to /tmp directory. You can modify your build script in your package.json to acheive this:

    "build": "nest build && yarn copyschema",
    "copyschema": "cp src/schema.gql dist/schema.gql"

You will also need to modify your app.module.ts file:

app.module.ts

@Module({
  imports: [
    GraphQLModule.forRootAsync<MercuriusDriverConfig>({
      imports: [XYZModule],
      useFactory: () => {
        const schemaModuleOptions: Partial<GqlModuleOptions> = {};

        if (process.env.NODE_ENV !== 'production' || process.env.IS_OFFLINE) {
          schemaModuleOptions.autoSchemaFile = join(
            process.cwd(),
            'src/schema.gql',
          );
        } else {
          schemaModuleOptions.typePaths = ['dist/*.gql'];
        }

        return {
          graphiql: true,
          context: (req: any) => ({
            XYZLoader: createXYZLoader(XYZService),
            headers: req.headers,
          }),
          ...schemaModuleOptions,
        };
      },
      driver: MercuriusDriver,
    }),
    XYZModule,
  ],
  controllers: [AppController],
  providers: [AppService],
})

📣Hint

Lambda arguments

The original lambda event and context are passed via Fastify request and can be used like this:

app.get('/', (request, reply) => {
  const event = request.awsLambda.event
  const context = request.awsLambda.context
  // ...
})

If you do not like it, you can disable this by setting the decorateRequest option to false.

Alternatively the original lambda event and context are passed via headers and can be used like this, if setting the serializeLambdaArguments option to true:

app.get('/', (request, reply) => {
  const event = JSON.parse(decodeURIComponent(request.headers['x-apigateway-event']))
  const context = JSON.parse(decodeURIComponent(request.headers['x-apigateway-context']))
  // ...
})

Lower cold start latency

Since AWS Lambda now enables the use of ECMAScript (ES) modules in Node.js 14 runtimes, you could lower the cold start latency when used with Provisioned Concurrency thanks to the top-level await functionality.

We can use this by calling the fastify.ready() function outside of the Lambda handler function, like this:

import awsLambdaFastify from '@fastify/aws-lambda'
import app from './app.js'
export const handler = awsLambdaFastify(app)
await app.ready() // needs to be placed after awsLambdaFastify call because of the decoration: https://github.com/fastify/aws-lambda-fastify/blob/master/index.js#L9

Here you can find the approriate issue discussing this feature.

⚡️Some basic performance metrics

@fastify/aws-lambda (decorateRequest : false) x 56,892 ops/sec ±3.73% (79 runs sampled)

@fastify/aws-lambda x 56,571 ops/sec ±3.52% (82 runs sampled)

@fastify/aws-lambda (serializeLambdaArguments : true) x 56,499 ops/sec ±3.56% (76 runs sampled)

serverless-http x 45,867 ops/sec ±4.42% (83 runs sampled)

aws-serverless-fastify x 17,937 ops/sec ±1.83% (86 runs sampled)

aws-serverless-express x 16,647 ops/sec ±2.88% (87 runs sampled)

Fastest is @fastify/aws-lambda (decorateRequest : false), @fastify/aws-lambda

⚠️Considerations

  • For apps that may not see traffic for several minutes at a time, you could see cold starts
  • Stateless only
  • API Gateway has a timeout of 29 seconds, and Lambda has a maximum execution time of 15 minutes. (Using Application Load Balancer has no timeout limit, so the lambda maximum execution time is relevant)
  • If you are using another web framework (Connect, Express, Koa, Restana, Sails, Hapi, Fastify, Restify) or want to use a more generic serverless proxy framework, have a look at: serverless-http

🎖Who is using it?

locize is using @fastify/aws-lambda
localistars is using @fastify/aws-lambda

The logos displayed in this page are property of the respective organisations and they are not distributed under the same license as @fastify/aws-lambda (MIT).