Skip to content

mercurius-js/mercurius-gateway

Repository files navigation

@mercurius/gateway

A module to create an Apollo Federation v1 gateway with mercurius.

Quick start

npm i fastify @mercuriusjs/federation @mercuriusjs/gateway

Create the user service

'use strict'

const Fastify = require('fastify')
const { mercuriusFederationPlugin } = require('@mercuriusjs/federation')

const users = {
  1: {
    id: '1',
    name: 'John',
    username: '@john'
  },
  2: {
    id: '2',
    name: 'Jane',
    username: '@jane'
  }
}

const service = Fastify()
const schema = `
  extend type Query {
    me: User
  }

  type User @key(fields: "id") {
    id: ID!
    name: String
    username: String
  }
`

const resolvers = {
  Query: {
    me: () => {
      return users['1']
    }
  },
  User: {
    __resolveReference: (source, args, context, info) => {
      return users[source.id]
    }
  }
}

service.register(mercuriusFederationPlugin, {
  schema,
  resolvers
})

service.listen({ port: 4001 })

Create the post service

'use strict'

const Fastify = require('fastify')
const { mercuriusFederationPlugin } = require('@mercuriusjs/federation')

const posts = {
  p1: {
    pid: 'p1',
    title: 'Post 1',
    content: 'Content 1',
    authorId: 'u1'
  },
  p2: {
    pid: 'p2',
    title: 'Post 2',
    content: 'Content 2',
    authorId: 'u2'
  }
}

const service = Fastify()
const schema = `
  extend type Query {
    topPosts(count: Int): [Post]
  }

  type Post @key(fields: "pid") {
    pid: ID!
    title: String
    content: String
    author: User @requires(fields: "pid title")
  }
`

const resolvers = {
  Query: {
    topPosts: (root, { count = 2 }) => Object.values(posts).slice(0, count)
  },
  Post: {
    __resolveReference: post => {
      return posts[post.pid]
    },
    author: post => {
      return {
        __typename: 'User',
        id: post.authorId
      }
    }
  }
}

service.register(mercuriusFederationPlugin, {
  schema,
  resolvers
})

service.listen({ port: 4002 })

Create the gateway

'use strict'

const Fastify = require('fastify')
const mercuriusGateway = require('@mercuriusjs/gateway')

const gateway = Fastify()
gateway.register(mercuriusGateway, {
  gateway: {
    services: [
      {
        name: 'user',
        url: 'http://localhost:4001/graphql'
      },
      {
        name: 'post',
        url: 'http://localhost:4002/graphql'
      }
    ]
  }
})

gateway.listen({ port: 3000 })

API

mercuriusGateway

Register the gateway in fastify.

const mercuriusGateway = require('@mercurius/gateway')

const schema = ...
const resolvers = ...
const app = Fastify()

app.register(mercuriusGateway, {
  gateway: [
    services: [
      {
        name: 'user',
        url: 'http://localhost:4001/graphql'
      },
      {
        name: 'post',
        url: 'http://localhost:4002/graphql'
      }
    ]
  }
 
})

options

  • all the mercurius plugin options

  • gateway: Object. Run the GraphQL server in gateway mode.

    • gateway.services: Service[] | Function An array of GraphQL services that are part of the gateway or a Function that returns a Promise that resolves to an array of services. Required.
      • service.name: A unique name for the service. Required.
      • service.url: The URL of the service endpoint. It can also be an Array of URLs and in which case all the requests will be load balanced throughout the URLs. Required.
      • service.mandatory: Boolean Marks service as mandatory. If any of the mandatory services are unavailable, gateway will exit with an error. (Default: false)
      • service.useSecureParse: Boolean Marks if the service response needs to be parsed securely using secure-json-parse. (Default: false)
      • service.rewriteHeaders: Function A function that gets the original headers as a parameter and returns an object containing values that should be added to the headers
      • service.setResponseHeaders: Function A function that gets reply as a parameter and can be used to set headers for the federated response to be sent to the client.
      • service.initHeaders: Function or Object An object or a function that returns the headers sent to the service for the initial _service SDL query.
      • service.connections: The number of clients to create. (Default: 10)
      • service.agent: An optional, fully configured undici agent/pool instance to use to perform network requests. If used, you must set all connections options on the instance as the request related options from the service configuration will not be applied.
      • service.bodyTimeout: The timeout after which a request will time out, in milliseconds. (Default: 30e3 - 30 seconds)
      • service.headersTimeout: The amount of time the parser will wait to receive the complete HTTP headers, in milliseconds. (Default: 30e3 - 30 seconds)
      • service.keepAliveMaxTimeout: The maximum allowed keepAliveTimeout. (Default: 5e3 - 5 seconds)
      • service.maxHeaderSize: The maximum length of request headers in bytes. (Default: 16384 - 16KiB)
      • service.keepAlive: The amount of time pass between the keep-alive messages sent from the gateway to the service, if undefined, no keep-alive messages will be sent. (Default: undefined)
      • service.wsUrl: The url of the websocket endpoint
      • service.wsConnectionParams: Function or Object
        • wsConnectionParams.connectionInitPayload: Function or Object An object or a function that returns the connection_init payload sent to the service.
        • wsConnectionParams.reconnect: Boolean Enable reconnect on connection close (Default: false)
        • wsConnectionParams.maxReconnectAttempts: Number Defines the maximum reconnect attempts if reconnect is enabled (Default: Infinity)
        • wsConnectionParams.connectionCallback: Function A function called after a connection_ack message is received.
        • wsConnectionParams.failedConnectionCallback: Function A function called after a connection_error message is received, the first argument contains the message payload.
        • wsConnectionParams.failedReconnectCallback: Function A function called if reconnect is enabled and maxReconnectAttempts is reached.
        • wsConnectionParams.rewriteConnectionInitPayload: Function A function that gets the original connection_init payload along with the context as a parameter and returns an object that replaces the original connection_init payload before forwarding it to the federated service
      • service.collectors: Object
        • collectors.collectHeaders: boolean Adds to context the collectors.responseHeaders object in which are stored the response headers from federated services.
        • collectors.collectStatutsCodes: boolean Adds to context the collectors.statusCodes object in which are stored the status codes of the response from federated services.
        • collectors.collectExtensions: boolean Adds to context the collectors.extensions object in which are stored the extensions field of the response from federated services.
    • gateway.retryServicesCount: Number Specifies the maximum number of retries when a service fails to start on gateway initialization. (Default: 10)
    • gateway.retryServicesInterval: Number The amount of time(in milliseconds) between service retry attempts in case a service fails to start on gateway initialization. (Default: 3000)
    • gateway.pollingInterval: Number The amount of time (milliseconds) between polling the services for schema updates. If not specified, the gateway will not poll for schema updates. When gateway.services is a function, the list of services is updated with the result of the function every gateway.pollingInterval milliseconds.

Hooks

Hooks are registered with the fastify.graphqlGateway.addHook method and allow you to listen to specific events in the GraphQL request/response lifecycle. You have to register a hook before the event is triggered, otherwise the event is lost.

By using hooks you can interact directly with the GraphQL lifecycle of Mercurius gateway. There are GraphQL Request and Subscription hooks:

Notice: these hooks are only supported with async/await or returning a Promise.

Lifecycle

The schema of the internal lifecycle of Mercurius gateway.

On the right branch of every section there is the next phase of the lifecycle, on the left branch there is the corresponding GraphQL error(s) that will be generated if the parent throws an error (note that all the errors are automatically handled by Mercurius).

Gateway lifecycle

How the gateway lifecycle works integrated with the Mercurius lifecycle.

Incoming GraphQL Request
  │
  └─▶ Routing
           │
  errors ◀─┴─▶ preParsing Hook
                  │
         errors ◀─┴─▶ Parsing
                        │
               errors ◀─┴─▶ preValidation Hook
                               │
                      errors ◀─┴─▶ Validation
                                     │
                            errors ◀─┴─▶ preExecution Hook
                                            │
                                   errors ◀─┴─▶ Execution
                                                  │
                                                  └─▶ preGatewayExecution Hook(s) (appends errors only)
                                                         │
                                                errors ◀─┴─▶ GatewayExecution(s)
                                                               │
                                                      errors ◀─┴─▶ Resolution (once all GatewayExecutions have finished)
                                                                     │
                                                                     └─▶ onResolution Hook

Gateway subscription lifecycle

How the gateway subscription lifecycle works integrated with the Mercurius lifecycle.

Incoming GraphQL Websocket subscription data
  │
  └─▶ Routing
           │
  errors ◀─┴─▶ preSubscriptionParsing Hook
                  │
         errors ◀─┴─▶ Subscription Parsing
                        │
               errors ◀─┴─▶ preSubscriptionExecution Hook
                              │
                     errors ◀─┴─▶ Subscription Execution
                                    │
                                    │
                                    └─▶ preGatewaySubscriptionExecution Hook(s)
                                            │
                                   errors ◀─┴─▶ Gateway Subscription Execution(s)
                                                  │
                                      wait for subscription data
                                                  │
                   subscription closed on error ◀─┴─▶ Subscription Resolution (when subscription data is received)
                                                        │
                                                        └─▶ onSubscriptionResolution Hook
                                                              │
                                              keeping processing until subscription ended
                                                              │
                               subscription closed on error ◀─┴─▶ Subscription End (when subscription stop is received)
                                                                    │
                                                                    └─▶ onSubscriptionEnd Hook

GraphQL Request Hooks

It is pretty easy to understand where each hook is executed by looking at the lifecycle definition.

preGatewayExecution

In the preGatewayExecution hook, you can modify the following items by returning them in the hook definition:

  • document
  • errors

This hook will only be triggered in gateway mode. When in gateway mode, each hook definition will trigger multiple times in a single request just before executing remote GraphQL queries on the federated services.

Note, this hook contains service metadata in the service parameter:

  • name: service name
fastify.graphqlGateway.addHook('preGatewayExecution', async (schema, document, context, service) => {
  const { modifiedDocument, errors } = await asyncMethod(document)

  return {
    document: modifiedDocument,
    errors
  }
})

Manage Errors from a request hook

If you get an error during the execution of your hook, you can just throw an error. The preGatewayExecution hook, which will continue execution of the rest of the query and append the error to the errors array in the response.

fastify.graphqlGateway.addHook('preGatewayExecution', async (schema, document, context, service) => {
  throw new Error('Some error')
})

Add errors to the GraphQL response from a hook

The preGatewayExecution hook support adding errors to the GraphQL response.

fastify.graphqlGateway.addHook('preGatewayExecution', async (schema, document, context) => {
  return {
    errors: [new Error('foo')]
  }
})

Note, the original query will still execute. Adding the above will result in the following response:

{
  "data": {
    "foo": "bar"
  },
  "errors": [
    {
      "message": "foo"
    }
  ]
}

GraphQL Subscription Hooks

It is pretty easy to understand where each hook is executed by looking at the lifecycle definition.

preGatewaySubscriptionExecution

This hook will only be triggered in gateway mode. When in gateway mode, each hook definition will trigger when creating a subscription with a federated service.

Note, this hook contains service metadata in the service parameter:

  • name: service name
fastify.graphqlGateway.addHook('preGatewaySubscriptionExecution', async (schema, document, context, service) => {
  await asyncMethod()
})

Manage Errors from a subscription hook

If you get an error during the execution of your subscription hook, you can just throw an error and Mercurius will send the appropriate errors to the user along the websocket.`

fastify.graphqlGateway.addHook('preSubscriptionParsing', async (schema, source, context) => {
  throw new Error('Some error')
})

GraphQL Application lifecycle Hooks

There is one hook that you can use in a GraphQL application.

onGatewayReplaceSchema

When the Gateway service obtains new versions of federated schemas within a defined polling interval, the onGatewayReplaceSchema hook will be triggered every time a new schema is built. It is called just before the old schema is replaced with the new one.

It has the following parameters:

  • instance - The gateway server FastifyInstance (this contains the old schema).
  • schema - The new schema that has been built from the gateway refresh.
fastify.graphqlGateway.addHook('onGatewayReplaceSchema', async (instance, schema) => {
  await someSchemaTraversalFn()
})

If this hook throws, the error will be caught and logged using the FastifyInstance logger. Subsequent onGatewayReplaceSchema hooks that are registered will not be run for this interval.

Collectors

Collectors gather additional information about the response from the services that are part of the gateway and adds them to the context.collectors object.

Depending on the configuration it may have the following fields:

  • responseHeaders
  • statusCodes
  • extensions

Each collector stores data in the same format:

  {
    "foo": { // query id
      "service": "bar", // name of the service
      "data": {
         "foo": "bar" // example: statusCode: 500
      }
    }
  }

It is possible to access and manipulate context.collectors via onResolution Hook

  app.graphql.addHook('onResolution', async function (execution, context) {
    console.log(context.collectors)

    // {
    //   statusCodes: {
    //     topPosts: {
    //       service: 'post',
    //       data: {
    //         statusCode: 500
    //       }
    //     },
    //     me: {
    //       service: 'user',
    //       data: {
    //         statusCode: 404
    //       }
    //     }
    //   },
    //   responseHeaders: {...},
    //   extensions: {...},
    // }

    execution.collectors = context.collectors // add collectors to response
  })