fast-gateway

A Node.js API Gateway for the masses!

Usage no npm install needed!

<script type="module">
  import fastGateway from 'https://cdn.skypack.dev/fast-gateway';
</script>

README

fast-gateway

NPM version tests

A super fast, framework agnostic Node.js API Gateway for the masses ❤️

Since v2.3.0, AWS Lambda proxying integration is supported via http-lambda-proxy 🔥

Since v3.1.0, WebSockets proxying is supported via faye-websocket 🔥

Also available for Docker: https://hub.docker.com/repository/docker/kyberneees/rproxy

Medium articles:

Install

npm i fast-gateway

Usage

Next we describe two examples proxying HTTP and Lambda downstream services.

For simplicity of reading, both examples are separated, however a single gateway configuration supports all routes configurations.

HTTP Proxying

Gateway

const gateway = require('fast-gateway')
const server = gateway({
  routes: [{
    prefix: '/service',
    target: 'http://127.0.0.1:3000'
  }]
})

server.start(8080)

Remote Service

const service = require('restana')()
service.get('/get', (req, res) => res.send('Hello World!'))

service.start(3000)

Lambda Proxying

Gateway

npm i http-lambda-proxy

const gateway = require('fast-gateway')
const server = gateway({
  routes: [{
    prefix: '/service',
    target: 'my-lambda-serverless-api',
    proxyType: 'lambda',
    proxyConfig: {
      region: 'eu-central-1'
    }
  }]
})

server.start(8080)

You might also want to read: Setting AWS Credentials in Node.js

Lambda Implementation

const serverless = require('serverless-http')
const json = require('serverless-json-parser')
const query = require('connect-query')

const service = require('restana')()
service.use(query())
service.use(json())

// routes
service.get('/get', (req, res) => {
  res.send({ msg: 'Go Serverless!' })
})
service.post('/post', (req, res) => {
  res.send(req.body)
})

// export handler
module.exports.handler = serverless(service)

Configuration options explained

{
  // Optional server instance. Any HTTP framework that supports the following signature is compatible:
  // - server[HTTP_METHOD](pattern, [middleware1, middleware2,], handler)
  // 
  // Known compatible frameworks: Restana, Express.js
  // If omitted, restana is used as default HTTP framework
  server, 
  // Optional restana library configuration (https://www.npmjs.com/package/restana#configuration)  
  // Please note that if "server" is provided, this settings are ignored.
  restana: {},
  // Optional global middlewares in the format: (req, res, next) => next() 
  // Default value: []
  middlewares: [],
  // Optional global value for routes "pathRegex". Default value: '/*'
  pathRegex: '/*',
  // Optional global requests timeout value (given in milliseconds). Default value: '0' (DISABLED)
  // Ignored if proxyType = 'lambda'
  timeout: 0,
  // Optional "target" value that overrides the routes "target" config value. Feature intended for testing purposes.
  targetOverride: "https://yourdev.api-gateway.com",
  // Optional "Proxy Factory" implementation, allows the integration of custom proxying strategies.
  // Default value: require('fast-proxy-lite/lib/proxy-factory')
  proxyFactory: ({ proxyType, opts, route }) => {...}

  // HTTP proxy
  routes: [{
    // Optional proxy type definition. Supported values: http, http-legacy, lambda
    // Modules:
    // - http: fast-proxy-lite
    // - http-legacy: fast-proxy
    // - lambda: http-lambda-proxy
    // Default value: http
    proxyType: 'http'
    // Optional proxy library configuration: 
    // - fast-proxy-lite: https://www.npmjs.com/package/fast-proxy-lite#options
    // - fast-proxy: https://www.npmjs.com/package/fast-proxy#options
    // - http-lambda-proxy: https://www.npmjs.com/package/http-lambda-proxy#options
    // Default value: {}
    proxyConfig: {},
    // Optional proxy handler function. Default value: (req, res, url, proxy, proxyOpts) => proxy(req, res, url, proxyOpts)
    proxyHandler: () => {},
    // Optional path matching regex. Default value: '/*'
    // In order to disable the 'pathRegex' at all, you can use an empty string: ''
    pathRegex: '/*',
    // Optional service requests timeout value (given in milliseconds). Default value: '0' (DISABLED)
    // This setting apply only when proxyType = 'http'
    timeout: 0,
    // route prefix
    prefix: '/public',
    // Optional documentation configuration (unrestricted schema)
    docs: {
      name: 'Public Service',
      endpoint: '/api-docs',
      type: 'swagger'
    },
    // Optional "prefix rewrite" before request is forwarded. Default value: ''
    prefixRewrite: '',
    // Optional "url rewrite" hook. If defined, the prefixRewrite setting is ignored. 
    urlRewrite: (req) => req.url,
    // Remote HTTP server URL to forward the request. 
    // If proxyType = 'lambda', the value is the name of the Lambda function, version, or alias.
    target: 'http://localhost:3000',
    // Optional HTTP methods to limit the requests proxy to certain verbs only
    // Supported HTTP methods: ['GET', 'DELETE', 'PATCH', 'POST', 'PUT', 'HEAD', 'OPTIONS', 'TRACE']
    methods: ['GET', 'POST', ...], 
    // Optional route level middlewares. Default value: []
    middlewares: [],
    // Optional proxy lifecycle hooks. Default value: {}
    hooks: {
      async onRequest (req, res) {
      //   // we can optionally reply from here if required
      //   res.end('Hello World!')
      //
      //   // we can optionally update the request query params from here if required
      //   req.query.category = 'js'
      //
      //   return true // truthy value returned will abort the request forwarding
      },
      onResponse (req, res, stream) {  
        // do some post-processing here
        // ...
      }

      // if proxyType= 'http', other options allowed https://www.npmjs.com/package/fast-proxy-lite#opts
    }
  }]
}

onResponse hooks default implementation

For developers reference, default hooks implementation are located in lib/default-hooks.js file.

The "GET /services.json" endpoint

Since version 1.3.5 the gateway exposes minimal documentation about registered services at: GET /services.json

Example output:

[  
   {  
      "prefix":"/public",
      "docs":{  
         "name":"Public Service",
         "endpoint":"/swagger.json",
         "type":"swagger"
      }
   },
   {  
      "prefix":"/admin"
   }
]

NOTE: Please see docs configuration entry explained above.

WebSockets support

WebSockets proxying is supported since v3.1.0. Main considerations:

  • The faye-websocket module dependency require to be installed:
    npm i faye-websocket
  • WebSockets middlewares are not supported.
  • WebSocketRoute configuration definition:
    interface WebSocketRoute {
  proxyType: 'websocket';
  // https://github.com/faye/faye-websocket-node#initialization-options
  proxyConfig?: {}; 
  prefix: string;
  target: string;
  // https://github.com/faye/faye-websocket-node#subprotocol-negotiation
  subProtocols?: []; 
  hooks?: WebSocketHooks;
}

interface WebSocketHooks {
  onOpen?: (ws: any, searchParams: URLSearchParams) => Promise<void>;
}
  • The / route prefix is considered the default route.

Configuration example:

gateway({
  routes: [{
    // ... other HTTP or WebSocket routes
  }, {
    proxyType: 'websocket',
    prefix: '/echo',
    target: 'ws://ws.ifelse.io'
  }]
}).start(PORT)

Timeouts and Unavailability

We can restrict requests timeouts globally or at service level using the timeout configuration.

You can also define endpoints specific timeout using the property timeout of the request object, normally inside a middleware:

req.timeout = 500 // define a 500ms timeout on a custom request.

NOTE: You might want to also check https://www.npmjs.com/package/middleware-if-unless

Circuit Breakers

By using the proxyHandler hook, developers can optionally intercept and modify the default gateway routing behavior right before the origin request is proxied to the remote service. Therefore, connecting advanced monitoring mechanisms like Circuit Breakers is rather simple.

Please see the demos/circuitbreaker.js example for more details using the opossum library.

Rate Limiting

Rate limiting, as well many other gateway level features can be easily implemented using fast-gateway:

const rateLimit = require('express-rate-limit')
const requestIp = require('request-ip')

gateway({
  middlewares: [
    // first acquire request IP
    (req, res, next) => {
      req.ip = requestIp.getClientIp(req)
      return next()
    },
    // second enable rate limiter
    rateLimit({
      windowMs: 1 * 60 * 1000, // 1 minutes
      max: 60, // limit each IP to 60 requests per windowMs
      handler: (req, res) => res.send('Too many requests, please try again later.', 429)
    })
  ],

  // your downstream services
  routes: [{
    prefix: '/public',
    target: 'http://localhost:3000'
  }, {
    // ...
  }]
})

In this example we have used the express-rate-limit module.

Hostnames support

We can also implement hostnames support with fast-gateway, basically we translate hostnames to prefixes:

...

// binding hostnames to prefixes
const hostnames2prefix = [{
  prefix: '/api',
  hostname: 'api.company.tld'
}]
// instantiate hostnames hook, this will prefix request urls according to data in hostnames2prefix
const hostnamesHook = require('fast-gateway/lib/hostnames-hook')(hostnames2prefix)

// separately instantiate and configure restana application
const app = restana()
const server = http.createServer((req, res) => {
  hostnamesHook(req, res, () => {
    return app(req, res)
  })
})

// gateway configuration
gateway({
  server: app, // injecting existing restana application
  routes: [{
    prefix: '/api',
    target: 'http://localhost:3000'
  }]
})

...

Afterwards:
curl --header "Host: api.company.tld:8080" http://127.0.0.1:8080/api-service-endpoint

You can optionally npm install micromatch and benefit from patterns support:

const hostnames2prefix = [{
  prefix: '/admin',
  hostname: '*.admin.company.tld'
}, {
  prefix: '/services',
  hostname: [
    'services.company.tld', 
    '*.services.company.tld'
  ]
}]

For more details, please checkout the basic-hostnames.js demo.

Gateway level caching

Caching support is provided by the http-cache-middleware module. https://www.npmjs.com/package/http-cache-middleware

Why?

Because caching is the last mile for low latency distributed systems!

Enabling proper caching strategies at gateway level will drastically reduce the latency of your system, as it reduces network round-trips and remote services processing.
We are talking here about improvements in response times from X ms to ~2ms, as an example.

We use the http-cache-middleware module to support gateway level caching. Read more about it: https://github.com/jkyberneees/http-cache-middleware

Setting up gateway level cache available for all services

Single node cache (memory):

// cache middleware
const cache = require('http-cache-middleware')()
// enable http cache middleware
const gateway = require('fast-gateway')
const server = gateway({
  middlewares: [cache],
  routes: [...]
})

Memory storage is recommended if there is only one gateway instance and you are not afraid of losing cache data.

Multi nodes cache (redis):

// redis setup
const CacheManager = require('cache-manager')
const redisStore = require('cache-manager-ioredis')
const redisCache = CacheManager.caching({
  store: redisStore,
  db: 0,
  host: 'localhost',
  port: 6379,
  ttl: 30
})

// cache middleware
const cache = require('http-cache-middleware')({
  stores: [redisCache]
})

// enable http cache middleware
const gateway = require('fast-gateway')
const server = gateway({
  middlewares: [cache],
  routes: [...]
})

Required if there are more than one gateway instances

How to cache remote services endpoints response?

https://github.com/jkyberneees/http-cache-middleware#enabling-cache-for-service-endpoints

How to invalidate caches?

https://github.com/jkyberneees/http-cache-middleware#invalidating-caches

Custom cache keys

Cache keys are generated using: req.method + req.url, however, for indexing/segmenting requirements it makes sense to allow cache keys extensions.
Unfortunately, this feature can't be implemented at remote service level, because the gateway needs to know the entire lookup key when a request reaches the gateway.

For doing this, we simply recommend using middlewares on the service configuration:

routes: [{
  prefix: '/users',
  target: 'http://localhost:3000',
  middlewares: [(req, res, next) => {
    req.cacheAppendKey = (req) => req.user.id // here cache key will be: req.method + req.url + req.user.id
    return next()
  }]
}]

In this example we also distinguish cache entries by user.id, very common case!

Disable cache for custom endpoints

You can also disable cache checks for certain requests programmatically:

routes: [{
  prefix: '/users',
  target: 'http://localhost:3000',
  middlewares: [(req, res, next) => {
    req.cacheDisabled = true
    return next()
  }]
}]

Related projects

Benchmarks

Benchmark scripts can be found in benchmark folder.

Laptop: MacBook Pro 2016, 2,7 GHz Intel Core i7, 16 GB 2133 MHz LPDDR3
wrk -t8 -c50 -d20s http://127.0.0.1:8080/service/get

  • fast-gateway: 18069.77 reqs/secs
  • k-fastify-gateway: 9763.61 reqs/secs

Sponsors

  • (INACTIVE) Kindly sponsored by ShareNow, a company that promotes innovation!

Support / Donate 💚

You can support the maintenance of this project:

Breaking Changes

v3.x

  • The fast-proxy-lite module is used by default to support http proxy type 🔥. This means, no undici or http2 are supported by default.
  • The old fast-proxy module is available under the http-legacy proxy type, but the module is not installed by default.
  • Proxy configuration is now generalized under the proxyConfig property.