Currently, my Apollo Server(running on HapiJS) returns HTTP 200 for every request, including failed ones.
I would like the GraphQL server to return HTTP 4xx for unsuccessful requests. The primary reason for it is that I want to set up monitoring for my ELB.
I know that Apollo Server has an engine platform, but I want to implement it using my current infrastructure.
Any ideas of how I could accomplish that? I tried to capture 'onPreResponse' event for my HapiJS server but I couldn't modify status code there.
After reading this answer. Here is a solution by modifying the hapijs plugin graphqlHapi of hapiApollo.ts file.
server.ts:
import { makeExecutableSchema } from 'apollo-server';
import { ApolloServer, gql } from 'apollo-server-hapi';
import Hapi from 'hapi';
import { graphqlHapi } from './hapiApollo';
const typeDefs = gql`
type Query {
_: String
}
`;
const resolvers = {
Query: {
_: () => {
throw new Error('some error');
},
},
};
const schema = makeExecutableSchema({ typeDefs, resolvers });
const port = 3000;
async function StartServer() {
const app = new Hapi.Server({ port });
graphqlHapi.register(app, { path: '/graphql', graphqlOptions: { schema } });
app.ext('onPreResponse', (request: any, h: any) => {
const response = request.response;
if (!response.isBoom) {
return h.continue;
}
return h.response({ message: response.message }).code(400);
});
await app.start();
}
StartServer()
.then(() => {
console.log(`apollo server is listening on http://localhost:${port}/graphql`);
})
.catch((error) => console.log(error));
hapiApollo.ts:
import Boom from 'boom';
import { Server, Request, RouteOptions } from 'hapi';
import { GraphQLOptions, runHttpQuery, convertNodeHttpToRequest } from 'apollo-server-core';
import { ValueOrPromise } from 'apollo-server-types';
export interface IRegister {
(server: Server, options: any, next?: Function): void;
}
export interface IPlugin {
name: string;
version?: string;
register: IRegister;
}
export interface HapiOptionsFunction {
(request?: Request): ValueOrPromise<GraphQLOptions>;
}
export interface HapiPluginOptions {
path: string;
vhost?: string;
route?: RouteOptions;
graphqlOptions: GraphQLOptions | HapiOptionsFunction;
}
const graphqlHapi: IPlugin = {
name: 'graphql',
register: (server: Server, options: HapiPluginOptions, next?: Function) => {
if (!options || !options.graphqlOptions) {
throw new Error('Apollo Server requires options.');
}
server.route({
method: ['GET', 'POST'],
path: options.path || '/graphql',
vhost: options.vhost || undefined,
options: options.route || {},
handler: async (request, h) => {
try {
const { graphqlResponse, responseInit } = await runHttpQuery([request, h], {
method: request.method.toUpperCase(),
options: options.graphqlOptions,
query:
request.method === 'post'
? // TODO type payload as string or Record
(request.payload as any)
: request.query,
request: convertNodeHttpToRequest(request.raw.req),
});
// add our custom error handle logic
const graphqlResponseObj = JSON.parse(graphqlResponse);
if (graphqlResponseObj.errors && graphqlResponseObj.errors.length) {
throw new Error(graphqlResponseObj.errors[0].message);
}
const response = h.response(graphqlResponse);
Object.keys(responseInit.headers as any).forEach((key) =>
response.header(key, (responseInit.headers as any)[key]),
);
return response;
} catch (error) {
// handle our custom error
if (!error.name) {
throw Boom.badRequest(error.message);
}
if ('HttpQueryError' !== error.name) {
throw Boom.boomify(error);
}
if (true === error.isGraphQLError) {
const response = h.response(error.message);
response.code(error.statusCode);
response.type('application/json');
return response;
}
const err = new Boom(error.message, { statusCode: error.statusCode });
if (error.headers) {
Object.keys(error.headers).forEach((header) => {
err.output.headers[header] = error.headers[header];
});
}
// Boom hides the error when status code is 500
err.output.payload.message = error.message;
throw err;
}
},
});
if (next) {
next();
}
},
};
export { graphqlHapi };
Now, when the GraphQL resolver throws an error, the client-side will receive our custom response with Http status code 400 instead of 200 status code with GraphQL errors response.
General from the browser:
Request URL: http://localhost:3000/graphql
Request Method: POST
Status Code: 400 Bad Request
Remote Address: 127.0.0.1:3000
Referrer Policy: no-referrer-when-downgrade
The response body is: {"message":"some error"}
Related
So I have a project using the latest Next js 13, React 18, Urql 3, and using typescript
Currently, I have issues when trying to query the urql from the getstaticprops function. My urql request needs a guest token, and I'm storing the token on session storage(other suggestions ?).
It has no issue when the query is running on the client, but I have it when querying inside the function.
My concern is related to the token reading, so the server cannot read the session storage value.
I'm asking what is the better and simplest way to make this work.
Does use cookies to store guest tokens will make this work?
Or the configuration that doesn't work?
This is my current config for urql.ts
import {
createClient,
ssrExchange,
dedupExchange,
cacheExchange,
fetchExchange,
} from "urql";
import { GRAPH_URL } from "#lib/constant/env";
import type { TypedDocumentNode } from "#urql/core";
const isServerSide = typeof window === "undefined";
const ssrCache = ssrExchange({
isClient: !isServerSide,
});
const client = createClient({
url: GRAPH_URL,
exchanges: [dedupExchange, cacheExchange, ssrCache, fetchExchange],
fetchOptions: () => {
const token = sessionStorage.getItem("accessToken");
return {
headers: {
authorization: token ? `Bearer ${token}` : "",
},
};
},
});
const query = async (
query: TypedDocumentNode<any, object>,
variables?: Record<string, string | string[] | unknown>
) => {
try {
const response = await client.query(query, variables as any).toPromise();
return response;
} catch (error) {
if (error instanceof Error) console.error(error.message);
}
};
const mutation = async (
mutation: TypedDocumentNode<any, object>,
variables?: Record<string, string | string[] | unknown>
) => {
try {
const response = await client
.mutation(mutation, variables as any)
.toPromise();
return response;
} catch (error) {
if (error instanceof Error) console.error(error.message);
}
};
export { client, query, mutation, ssrCache };
And this some of the code for the blog index page
export const getStaticProps = async () => {
await fetchArticlesSummary();
return {
props: {
urqlState: ssrCache.extractData(),
},
revalidate: 600,
};
};
export default withUrqlClient(() => ({
url: GRAPH_URL,
}))(BlogPage);
This is for the fetchArticlesSummary
export const fetchArticlesSummary = async () => {
try {
const {
data: { listArticles },
}: any = await query(getListArticle);
return listArticles.items;
} catch (error) {
return {
notFound: true,
};
}
};
I also doing a setup on _app.tsx
export default function App({ Component, pageProps }: AppProps) {
if (pageProps.urqlState) {
ssrCache.restoreData(pageProps.urqlState);
}
return (
<Provider value={client}>
<Component {...pageProps} />
</Provider>
);
}
Thank you
I have followed urql documentation about server-side configuration and many others but still don't have any solutions.
I am trying to setup front end for graphQl file upload with Apollo-boost-upload. The backend code is based on this link
https://dev.to/dnature/handling-file-uploads-with-apollo-server-2-0-14n7.
It's now reaching the resolver breakpoint after adding the following line in the server.js file
const { apolloUploadExpress } = require("apollo-upload-server");
app.use(apolloUploadExpress({ maxFileSize: 1000000000, maxFiles: 10 }));
And after modifying the schema for the upload type
scalar Upload
Here is the Vue component
<input
type="file"
style="display:none"
ref="fileInput"
accept="image/*"
#change="upload"
>
//Upload method
upload({ target: { files = [] } }) {
if (!files.length) {
return;
}
this.logoImage = files[0];
},
//Dispatching action from vue component
this.$store.dispatch("uploadLogo", { image: this.logoImage });
//Vuex action
const uploadLogo = async (context, payload) => {
context.commit("setLoading", true);
try {
const { data } = await apolloClient.mutate({
mutation: UPLOAD_LOGO,
variables: {file: payload.image},
context: {
hasUpload: true,
},
});
context.commit("setLoading", false);
console.log("Logo:", data.uploadLogo);
} catch (error) {
context.commit("setLoading", false);
console.log(error);
}
};
//Mutation
export const UPLOAD_LOGO = gql`
mutation uploadLogo($file: Upload!) {
uploadLogo(file: $file) {
_id
path
filename
mimetype
user {
_id
}
}
}
`;
// Apolloclient config on main.js
import ApolloClient from "apollo-boost-upload";
import { InMemoryCache } from "apollo-boost";
import VueApollo from "vue-apollo";
// Set up Apollo Client
export const defaultClient = new ApolloClient({
uri: "http://localhost:4000/graphql",
cache: new InMemoryCache({
addTypename: false,
}),
fetchOptions: {
credentials: "include",
},
request: (operation) => {
// if no token in local storage, add it
if (!localStorage.someToken) {
localStorage.setItem("someToken", "");
}
// operation adds the token to authorizatrion header, which is sent o backend
operation.setContext({
headers: {
authorization: "Bearer " + localStorage.getItem("someToken"),
},
});
},
onError: ({ graphQLErrors, networkError }) => {
if (networkError) {
console.log("[networkError]", networkError);
}
if (graphQLErrors) {
for (const error of graphQLErrors) {
console.dir(error);
if (error.name === "AuthenticationError") {
// set auth errir in state
store.commit("setError", error);
// signout user to clear error
store.dispatch("signUserOut");
}
}
}
},
});
Here is the updated typedef (old code commented out) from backend if that helps to identify the issue
const logoUploadTypeDefs = gql`
type File {
_id: ID!
path: String!
filename: String!
mimetype: String!
encoding: String!
user: User
}
# input Upload {
# name: String!
# type: String!
# size: Int!
# path: String!
# }
scalar Upload
type Mutation {
uploadLogo(file: Upload!): File
}
type Query {
info: String
logo: File!
}
`;
Now, the Node app crashes with the following log
I had to change "apollo-upload-server" to "graphql-upload"
change 1:
Commented out "apollo-upload-server" and used "graphql-upload"
// const { apolloUploadExpress } = require("apollo-upload-server");]
const {
graphqlUploadExpress, // A Koa implementation is also exported.
} = require("graphql-upload");
And in the middleware, used this
change 2:
app.use(graphqlUploadExpress());
await apolloServer.start();
instead of old code
app.use(apolloUploadExpress());// Not to be used
await apolloServer.start();
Also, in the resolver, I added this
change 3:
Import Upload from graphql-upload in the resolver file
const { GraphQLUpload } = require("graphql-upload");
....
....
const resolvers = {
// This maps the `Upload` scalar to the implementation provided
// by the `graphql-upload` package.
Upload: GraphQLUpload,
Query: {
....
}
Mutations: {
....
}
}
Refer to Apollo Docs for more details. This fixed the issue of Node crashing with error "Maximum call stack size exceeded at _openReadFs..."
I know some questions about the subject has been opened here and there, but my issue is different :
all the other ones appear in dev mode, in my case it's in production,
a very big percentage of requests pass, a few of them is TypeError: Network request failed - but sometimes for critical requests
it's random, not always the same request. Sometimes it passes, sometimes not.
it appears to three on my projects, one is on AWS the other one on Clever-Cloud, both are projects between 1000 and 5000 users, servers are quite too big for what they do - I think I removed the risk of a server fault. Even if... I can reproduce locally when I don't start the api locally. So it's like the api is not responding, but as I said, I don't think so.
I have no clue where to dig anymore...
I can give you my API.js service file, maybe you'll find what's wrong ?
import URI from 'urijs';
import { Platform } from 'react-native';
import NetInfo from '#react-native-community/netinfo';
import { getUserToken, wipeData } from '../utils/data';
import { SCHEME, MW_API_HOST } from '../config';
import deviceInfoModule from 'react-native-device-info';
import { capture } from '../utils/sentry';
const unauthorisedHandler = (navigation) => {
wipeData();
navigation.reset({ index: 0, routes: [{ name: 'Auth' }] });
};
const checkNetwork = async (test = false) => {
const isConnected = await NetInfo.fetch().then((state) => state.isConnected);
if (!isConnected || test) {
await new Promise((res) => setTimeout(res, 1500));
return false;
}
return true;
};
class ApiService {
host = MW_API_HOST;
scheme = SCHEME;
getUrl = (path, query) => {
return new URI().host(this.host).scheme(this.scheme).path(path).setSearch(query).toString();
};
execute = async ({ method = 'GET', path = '', query = {}, headers = {}, body = null }) => {
try {
const config = {
method,
headers: {
'Content-Type': 'application/json',
Accept: 'application/json',
appversion: deviceInfoModule.getBuildNumber(),
appdevice: Platform.OS,
currentroute: this.navigation?.getCurrentRoute?.()?.name,
...headers,
},
body: body ? JSON.stringify(body) : null,
};
const url = this.getUrl(path, query);
console.log('url: ', url);
const canFetch = await checkNetwork();
if (!canFetch) return;
let response;
// To try to avoid mysterious `TypeError: Network request failed` error
// that throws an error directly
// we try catch and try one more time.
try {
response = await fetch(url, config);
} catch (e) {
if (e?.toString().includes('Network request failed')) {
// try again
await new Promise((res) => setTimeout(res, 250));
console.log('try again because Network request failed');
response = await fetch(url, config);
} else {
throw e;
}
}
if (!response.ok) {
if (response.status === 401) {
const token = await getUserToken();
if (token) unauthorisedHandler(API.navigation);
return response;
}
}
if (response.json) return await response.json();
return response;
} catch (e) {
capture(e, { extra: { method, path, query, headers, body } });
return { ok: false, error: "Sorry, an error occured, technical team has been warned." };
}
};
executeWithToken = async ({ method = 'GET', path = '', query = {}, headers = {}, body = null }) => {
const token = await getUserToken();
if (token) headers.Authorization = token;
return this.execute({ method, path, query, headers, body });
};
get = async (args) => this.executeWithToken({ method: 'GET', ...args });
post = async (args) => this.executeWithToken({ method: 'POST', ...args });
put = async (args) => this.executeWithToken({ method: 'PUT', ...args });
delete = async (args) => this.executeWithToken({ method: 'DELETE', ...args });
}
const API = new ApiService();
export default API;
Talking with experts here and there, it seems that it's normal : internet network is not 100% reliable, so sometimes, request fail, for a reason that we can't anticipate (tunnel, whatever).
I ended up using fetch-retry and I still have a few of those, but much less !
So we're creating a React-Native app using Apollo and GraphQL. I'm using JWT based authentication(when user logs in both an activeToken and refreshToken is created), and want to implement a flow where the token gets refreshed automatically when the server notices it's been expired.
The Apollo Docs for Apollo-Link-Error provides a good starting point to catch the error from the ApolloClient:
onError(({ graphQLErrors, networkError, operation, forward }) => {
if (graphQLErrors) {
for (let err of graphQLErrors) {
switch (err.extensions.code) {
case 'UNAUTHENTICATED':
// error code is set to UNAUTHENTICATED
// when AuthenticationError thrown in resolver
// modify the operation context with a new token
const oldHeaders = operation.getContext().headers;
operation.setContext({
headers: {
...oldHeaders,
authorization: getNewToken(),
},
});
// retry the request, returning the new observable
return forward(operation);
}
}
}
})
However, I am really struggling to figure out how to implement getNewToken().
My GraphQL endpoint has the resolver to create new tokens, but I can't call it from Apollo-Link-Error right?
So how do you refresh the token if the Token is created in the GraphQL endpoint that your Apollo Client will connect to?
The example given in the the Apollo Error Link documentation is a good starting point but assumes that the getNewToken() operation is synchronous.
In your case, you have to hit your GraphQL endpoint to retrieve a new access token. This is an asynchronous operation and you have to use the fromPromise utility function from the apollo-link package to transform your Promise to an Observable.
import React from "react";
import { AppRegistry } from 'react-native';
import { onError } from "apollo-link-error";
import { fromPromise, ApolloLink } from "apollo-link";
import { ApolloClient } from "apollo-client";
let apolloClient;
const getNewToken = () => {
return apolloClient.query({ query: GET_TOKEN_QUERY }).then((response) => {
// extract your accessToken from your response data and return it
const { accessToken } = response.data;
return accessToken;
});
};
const errorLink = onError(
({ graphQLErrors, networkError, operation, forward }) => {
if (graphQLErrors) {
for (let err of graphQLErrors) {
switch (err.extensions.code) {
case "UNAUTHENTICATED":
return fromPromise(
getNewToken().catch((error) => {
// Handle token refresh errors e.g clear stored tokens, redirect to login
return;
})
)
.filter((value) => Boolean(value))
.flatMap((accessToken) => {
const oldHeaders = operation.getContext().headers;
// modify the operation context with a new token
operation.setContext({
headers: {
...oldHeaders,
authorization: `Bearer ${accessToken}`,
},
});
// retry the request, returning the new observable
return forward(operation);
});
}
}
}
}
);
apolloClient = new ApolloClient({
link: ApolloLink.from([errorLink, authLink, httpLink]),
});
const App = () => (
<ApolloProvider client={apolloClient}>
<MyRootComponent />
</ApolloProvider>
);
AppRegistry.registerComponent('MyApplication', () => App);
You can stop at the above implementation which worked correctly until two or more requests failed concurrently. So, to handle concurrent requests failure on token expiration, have a look at this post.
Update - Jan 2022
you can see basic React JWT Authentication Setup from: https://github.com/bilguun-zorigt/React-GraphQL-JWT-Authentication-Example
I've also added the safety points to consider when setting up authentication on both the frontend and backend on the Readme section of the repository. (XSS attack, csrf attack etc...)
Original answer - Dec 2021
My solution:
Works with concurrent requests (by using single promise for all requests)
Doesn't wait for error to happen
Used second client for refresh mutation
import { setContext } from '#apollo/client/link/context';
async function getRefreshedAccessTokenPromise() {
try {
const { data } = await apolloClientAuth.mutate({ mutation: REFRESH })
// maybe dispatch result to redux or something
return data.refreshToken.token
} catch (error) {
// logout, show alert or something
return error
}
}
let pendingAccessTokenPromise = null
export function getAccessTokenPromise() {
const authTokenState = reduxStoreMain.getState().authToken
const currentNumericDate = Math.round(Date.now() / 1000)
if (authTokenState && authTokenState.token && authTokenState.payload &&
currentNumericDate + 1 * 60 <= authTokenState.payload.exp) {
//if (currentNumericDate + 3 * 60 >= authTokenState.payload.exp) getRefreshedAccessTokenPromise()
return new Promise(resolve => resolve(authTokenState.token))
}
if (!pendingAccessTokenPromise) pendingAccessTokenPromise = getRefreshedAccessTokenPromise().finally(() => pendingAccessTokenPromise = null)
return pendingAccessTokenPromise
}
export const linkTokenHeader = setContext(async (_, { headers }) => {
const accessToken = await getAccessTokenPromise()
return {
headers: {
...headers,
Authorization: accessToken ? `JWT ${accessToken}` : '',
}
}
})
export const apolloClientMain = new ApolloClient({
link: ApolloLink.from([
linkError,
linkTokenHeader,
linkMain
]),
cache: inMemoryCache
});
If you are using JWT, you should be able to detect when your JWT token is about to expire or if it is already expired.
Therefore, you do not need to make a request that will always fail with 401 unauthorized.
You can simplify the implementation this way:
const REFRESH_TOKEN_LEGROOM = 5 * 60
export function getTokenState(token?: string | null) {
if (!token) {
return { valid: false, needRefresh: true }
}
const decoded = decode(token)
if (!decoded) {
return { valid: false, needRefresh: true }
} else if (decoded.exp && (timestamp() + REFRESH_TOKEN_LEGROOM) > decoded.exp) {
return { valid: true, needRefresh: true }
} else {
return { valid: true, needRefresh: false }
}
}
export let apolloClient : ApolloClient<NormalizedCacheObject>
const refreshAuthToken = async () => {
return apolloClient.mutate({
mutation: gql```
query refreshAuthToken {
refreshAuthToken {
value
}```,
}).then((res) => {
const newAccessToken = res.data?.refreshAuthToken?.value
localStorage.setString('accessToken', newAccessToken);
return newAccessToken
})
}
const apolloHttpLink = createHttpLink({
uri: Config.graphqlUrl
})
const apolloAuthLink = setContext(async (request, { headers }) => {
// set token as refreshToken for refreshing token request
if (request.operationName === 'refreshAuthToken') {
let refreshToken = localStorage.getString("refreshToken")
if (refreshToken) {
return {
headers: {
...headers,
authorization: `Bearer ${refreshToken}`,
}
}
} else {
return { headers }
}
}
let token = localStorage.getString("accessToken")
const tokenState = getTokenState(token)
if (token && tokenState.needRefresh) {
const refreshPromise = refreshAuthToken()
if (tokenState.valid === false) {
token = await refreshPromise
}
}
if (token) {
return {
headers: {
...headers,
authorization: `Bearer ${token}`,
}
}
} else {
return { headers }
}
})
apolloClient = new ApolloClient({
link: apolloAuthLink.concat(apolloHttpLink),
cache: new InMemoryCache()
})
The advantage of this implementation:
If the access token is about to expire (REFRESH_TOKEN_LEGROOM), it will request a refresh token without stopping the current query. Which should be invisible to your user
If the access token is already expired, it will refresh the token and wait for the response to update it. Much faster than waiting for the error back
The disadvantage:
If you make many requests at once, it may request several times a refresh. You can easily protect against it by waiting a global promise for example. But you will have to implement a proper race condition check if you want to guaranty only one refresh.
after checking this topic and some others very good on internet, my code worked with the following solution
ApolloClient,
NormalizedCacheObject,
gql,
createHttpLink,
InMemoryCache,
} from '#apollo/client';
import { setContext } from '#apollo/client/link/context';
import jwt_decode, { JwtPayload } from 'jwt-decode';
import {
getStorageData,
setStorageData,
STORAGE_CONTANTS,
} from '../utils/local';
export function isRefreshNeeded(token?: string | null) {
if (!token) {
return { valid: false, needRefresh: true };
}
const decoded = jwt_decode<JwtPayload>(token);
if (!decoded) {
return { valid: false, needRefresh: true };
}
if (decoded.exp && Date.now() >= decoded.exp * 1000) {
return { valid: false, needRefresh: true };
}
return { valid: true, needRefresh: false };
}
export let client: ApolloClient<NormalizedCacheObject>;
const refreshAuthToken = async () => {
const refreshToken = getStorageData(STORAGE_CONTANTS.REFRESHTOKEN);
const newToken = await client
.mutate({
mutation: gql`
mutation RefreshToken($refreshAccessTokenRefreshToken: String!) {
refreshAccessToken(refreshToken: $refreshAccessTokenRefreshToken) {
accessToken
status
}
}
`,
variables: { refreshAccessTokenRefreshToken: refreshToken },
})
.then(res => {
const newAccessToken = res.data?.refreshAccessToken?.accessToken;
setStorageData(STORAGE_CONTANTS.AUTHTOKEN, newAccessToken, true);
return newAccessToken;
});
return newToken;
};
const apolloHttpLink = createHttpLink({
uri: process.env.REACT_APP_API_URL,
});
const apolloAuthLink = setContext(async (request, { headers }) => {
if (request.operationName !== 'RefreshToken') {
let token = getStorageData(STORAGE_CONTANTS.AUTHTOKEN);
const shouldRefresh = isRefreshNeeded(token);
if (token && shouldRefresh.needRefresh) {
const refreshPromise = await refreshAuthToken();
if (shouldRefresh.valid === false) {
token = await refreshPromise;
}
}
if (token) {
return {
headers: {
...headers,
authorization: `${token}`,
},
};
}
return { headers };
}
return { headers };
});
client = new ApolloClient({
link: apolloAuthLink.concat(apolloHttpLink),
cache: new InMemoryCache(),
});
A much simpler solution is using RetryLink. retryIf supports async operations so one could do something like this:
class GraphQLClient {
constructor() {
const httpLink = new HttpLink({ uri: '<graphql-endpoint>', fetch: fetch })
const authLink = setContext((_, { headers }) => this._getAuthHeaders(headers))
const retryLink = new RetryLink({
delay: { initial: 300, max: Infinity, jitter: false },
attempts: {
max: 3,
retryIf: (error, operation) => this._handleRetry(error, operation)
}})
this.client = new ApolloClient({
link: ApolloLink.from([ authLink, retryLink, httpLink ]),
cache: new InMemoryCache()
})
}
async _handleRetry(error, operation) {
let requiresRetry = false
if (error.statusCode === 401) {
requiresRetry = true
if (!this.refreshingToken) {
this.refreshingToken = true
await this.requestNewAccessToken()
operation.setContext(({ headers = {} }) => this._getAuthHeaders(headers))
this.refreshingToken = false
}
}
return requiresRetry
}
async requestNewAccessToken() {
// get new access token
}
_getAuthHeaders(headers) {
// return headers
}
}
For some reason no error shows up in the server console when I start my hapi server with nodemon and navigate to http://localhost:3000/hapi-ext-fetch and this makes debugging very difficult. Here is my code:
var Hapi = require('hapi');
var Joi = require('joi');
var fetch = require('isomorphic-fetch');
var debugMode = { debug: { request: [ 'error', 'request-internal' ] }};
var server = new Hapi.Server(debugMode);
server.connection({ port: 3000 });
var myPlugin = {
register: function (server, options, next) {
server.route([
{
method: 'GET',
path: '/{name}',
handler: function ( request, reply ) {
throw new Error('this error isnt shown!');
},
config: {
validate: {
params: {
name: Joi.string().min(3).max(10)
}
}
}
}
]);
next();
}
};
myPlugin.register.attributes = {
name: 'myPlugin',
version: '1.0.0'
};
server.register([
{
register: myPlugin,
routes: {
prefix: '/test'
}
}
], function() {
server.ext( 'onPreResponse', ( request, reply ) => {
if ( typeof request.response.statusCode !== 'undefined' ) {
return reply.continue();
}
fetch('http://localhost:3000/test/whatever')
.then(function(result) {
reply(result);
})
.catch(function(err) {
reply('error on server side: ' + err.stack);
});
});
server.start((err) => {
if (err) {
throw err;
}
console.log('Server running at:', server.info.uri);
});
});
I'm using hapi 13.0.0
Can't say I totally understand your use case here and if this question will be helpful to other people. But what you're trying to do it seems is:
Send a request to /hapi-fetch-ext
Have that request 404
And then in an onPreResponse go fetch another route /test/whatever
Hope to see the "this error isn't shown error"
Not sure if you're aware but this is going to cause an infinite cycle of requests (your fetch will cause another onPreResponse and so on and so on). So you should probably only go fetch on a 404:
server.ext( 'onPreResponse', ( request, reply ) => {
if (request.response.isBoom && request.response.output.statusCode === 404) {
return fetch('http://localhost:3000/test/whatever')
.then(function(result) {
reply(result);
})
.catch(function(err) {
reply('error on server side: ' + err.stack);
});
}
return reply.continue();
});