I have a problem with my production website.
when users enter my website after building they getting a white screen.
I saw on the dev tools that the website trying to load the old bundle.js file and I don't understand why (the cashing is off)
This looks like a problem of a running service worker.
service worker is an API that cache your assets and other files so when a user is offline or having a slow network, he can still see your website.
React creates a service worker for you without your configuration by default, so this is the reason users enter your website and getting your previous bundle.
just put this in your app so next time user will enter your website, the service worker will be terminated.
import { unregister } from './registerServiceWorker';
// ....
unregister();
Service Workers, MDN
Service Workers, Google Web Fundamentals
Assuming you're using webpack to generate the builds and want caching (with service workers) to work as well, you can:
Add hashes to your generated bundles. Then, your webpack config may look like this:
{
output: {
path: path.resolve(__dirname, 'dist'),
filename: "[name].[chunkhash].js",
},
}
Read more about it here.
Do not cache your index.html which contains <script> imports of the updated hash-named bundles.
This will enable you to cache your builds easily, and only fetch new bundles whenever there is an update in the hash-names.
Related
I've split my react application in 3 different projects, using CRA for all of them, auth, X and Y. User is first sent to auth, then I redirect him to either X or Y based on some info.
It works perfectly on PRODUCTION environment (because they run on the same domain), but on dev, X and Y failed to authenticate the user, because they run on different ports (different domains) the data in local storage is not shared between auth, X and Y.
I've tried to find a way to use a reverse proxy (http-proxy) to host the React dev servers on the same domain, but failed too, because the services could not find the assets/static folder, resulting in 404. Also tried http-proxy-middleware, as it is recommended on the CRA docs page, but failed to do so. Is there an easier way that I'm not seeing?
Edit: Found something new, but also failed. Used react-rewired to override CRA scripts, to use PUBLIC_PATH on DEV, but now my bundle.js returns an index.html file.
The following code does redirect to the accordingly react project, but the assets are requested to the wrong path.
const apiProxy = httpProxy.createProxyServer();
app.all("/login/*", function(req, res) {
console.log('redirecting to Login');
apiProxy.web(req, res, {target: servers.login});
});
app.all("/implementacao/*", function(req, res) {
console.log('redirecting to Implementation');
apiProxy.web(req, res, {target: servers.implementation});
});
So I used react-rewired to change the public path
const {
override,
} = require('customize-cra');
module.exports = {
webpack: override(
(config) => {
config.output.publicPath = '/login/';
return config;
},
),
jest: config => {
return config;
},
devServer: configFunction => (proxy, allowedHost) => {
return configFunction(proxy, allowedHost);
},
paths: (paths, env) => {
return paths;
}
};
Now, the assets requests are made correctly to /login/, but nothing the dev server always return an index.html file.
Even with react-app-rewired, to override config, and use publicPath on dev, the assets will not be served from the publicPath.
There is already a pull request on CRA to use PUBLIC_URL in dev mode.
Is there an easier way that I'm not seeing?
Another approach would be to use multiple React Single Page Applications (SPAs) inside one application, see crisp-react. E.g. instead of 3 CRAs in 3 applications/projects have 3 SPAs in one application/project. The backend surely can get data from other backend servers transparently for each SPA.
how do I migrate from a set of existing CRA projects to using crisp-react ?
Background
crisp-react comes with two stock SPAs called ‘First’ and ‘Second’. Both render some explanatory/sample UI.
Migration overview
1.Pick one CRA project and migrate it to the ‘First’ SPA. When finished, you have two CRAs left and two crisp-react SPAs: ‘First’ (renders your UI) and ‘Second’ (still renders the sample UI). Rename the ‘First’ SPA to give it more meaningful name.
2. Pick another CRA and migrate it. When finished, you have one CRA left and two crisp-react SPAs both rendering your UI.
3.Modify crisp-react to add the third SPA and then migrate the remaining CRA to the third SPA.
Migration steps (sketch)
1.1 Follow crisp-react Getting Started.
1.2 The landing page of the First SPA is rendered by crisp-react/client/src/entrypoints/first.tsx
The landing page of the CRA is rendered by src/index.tsx
Replace the content of the former with the latter.
1.3 The first CRA consists of React components: src/App.tsx and others you added. Copy the components to crisp-react/client/src/components/from-first-cra/
1.4 Ensure crisp-react client app compiles: From crisp-react/client/ execute: yarn compile
1.5 Ensure crisp-react client app builds: From crisp-react/client/ execute: yarn build
1.6 Ensure crisp-react client looks ok without backend data: see client Usage Scenarios.
1.7 Get the backend (e.g. Express) involved: see backend Usage Scenarios.
1.8 Milestone reached: browser can be pointed to backend (Express) and get from it html files and bundles - which results in the first SPA rendering initial UI (to the extent possible without data supplied via API enpoints).
1.9 Decide how the the first SPA will get data from API. 3 basic choices here:
- the API endpoints are implemented in Express so you can retire your backend servers
- Express does expose API endpoints but acts as a reverse proxy getting data from your backend servers
- Express knows nothing about API and data supplied by backend servers that are queried directly by the components inside the first SPA.
2.1 Second SRA
as above
...
When I make a new production build of React app, new services worker installs but goes to skipwaiting state. So I have to hard reset cache on every build that is not good on live site.
I can't tell to every client, "please clear your cache to see latest". If I remove registerServiceWorker() from index file then build production do not detect my routes so it redirects to 404.
I used react-rewired-app boilerplate for my project and faced issue in that. Now I resolved that with a bunch of code in config-overrides.js file
config.plugins.forEach((obj) => {
if (obj.config) {
if (obj.config.clientsClaim) {
obj.config.skipWaiting = true
}
}
});
We created a Linux Web App in Microsoft Azure. The application is static written with React (html and Javascript).
We copied the code into the wwwroot folder, but the application only showing only hostingstart.html and when we try to get page index.html we have this error:
Cannot GET /index.html
We tried with a sample of Azure in GitHub (https://github.com/Azure-Samples/html-docs-hello-world) but the error is the same.
The url is this: https://consoleadmin.azurewebsites.net/index.html
Last week the application was running correctly.
We forget to do something?
MAY 2020 - You don't have to add any javascript files or config files anywhere. Let me explain.
I was facing this exact same issue and wasted 6 hours trying everything including the most popular answer to this question. While the accepted answer is a nice workaround (but requires more work than just adding the index.js file), there's something a simpler than that.
You see, when you just deploy an Azure Web App (or App Service as it is also called), two things happen:
The web app by default points to opt/startup/hostingstart.html
It also puts a hostingstart.html in home/site/wwwroot
When you deploy your code, it replaces hostingstart.html in home/site/wwwroot but the app is still pointing to opt/startup/hostingstart.html. If you want to verify this, try deleting opt/startup/hostingstart.html file and your web app will throw a "CANNOT GET/" error.
So how to change the default pointer? It's simpler than it looks:
Go to Configuration tab on your web app and add the following code to startup script:
pm2 serve /home/site/wwwroot --no-daemon
If this web app is a client-side single-page-app and you're having issues with routing, then add --spa to the above command as follows:
pm2 serve /home/site/wwwroot --no-daemon --spa
This will tell the web app to serve wwwroot folder. And that's it.
Image for reference:
Screenshot explaination
PS: If you only set the startup script without deploying your code, it will still show the hostingstart.html because by default that file lies in the wwwroot folder.
Ok you are gonna love this. This happened to me today also. Same exact thing.
I am pretty sure the azure team flipped a switch somewhere and we fell through a crack.
I found this obscure answer with no votes and it did the trick (with a little extra finagling)
BONUS! this also fixed my router issues I was having only on the deployed site (not local):
Credit: #stormwild: Default documents not serving on node web app hosted on Azure
From #stormwild's post see here:
https://blogs.msdn.microsoft.com/waws/2017/09/08/things-you-should-know-web-apps-and-linux/#NodeHome
Steps:
Go to your azure portal, select your app service and launch ssh
In ssh terminal, navigate via command line to /home/site/wwwroot
create index.js there with the following code:
var express = require('express');
var server = express();
var options = {
index: 'index.html'
};
server.use('/', express.static('/home/site/wwwroot', options));
server.listen(process.env.PORT);
NOTE: Be sure to run npm install --save express also in this folder else your app service will crash on startup
Be sure to restart your app service if it doesn't do so automagically
A workaround, I changed the webapp stack to PHP 7
Another solution would be to add a file called ecoysystem.config.js right next to your index.html file.
module.exports = {
apps: [
{
script: "npx serve -s"
}
]
};
This will tell pm2 to associate all requests to index.html as your app service starts up.
Very helpful information here: https://burkeholland.github.io/posts/static-site-azure/
I have a web application in React that I needed to implement a contact form. The application is created using create-react-app and the server folder added. For the form I used sendgrid mail. Does the server work on port 4567, how do the app build to work on the domain? It is a one-page application.
Thx, it is important.
When running in production, a React app is simple HTML, CSS, and JavaScript. These files are sent from your server to a client when requested in the same way that requests/responses are handled for any web page. There are a few steps that need to be done before your React app is ready for production
1: Create a Production Build
First you need to create a production build of your app. This process takes all of your separate .js or .jsx files and puts them together into a single minified file, and the same for .css. Then your index.html is updated to include a link to the CSS and script to the JS. This is done so that only three files will need to be sent rather than the 10s or 100s that exist in development.
If you used create-react-app to start your application, you can use the command:
npm run build
to do this. Otherwise, you need to have webpack installed, and then run:
node_modules/.bin/webpack --config webpack.prod.js --mode production
(which you might want to add as a script to package.json).
See React: Optimizing Performance for more.
2. Serve your Application
Now your server should have a route for your application and when it receives a request on that route, the server should respond by sending index.html from your client/build/ directory (where client/ is the directory of the React app).
Here is an example with Node/Express as the server (in app.js):
const path = require('path');
app.get('*', (req, res) => {
res.sendFile(path.join(__dirname), 'client', 'build', 'index.html');
});
Note that this is just the way to send a static file using Node and can easily be done with any server.
Additional
You mentioned you want to submit forms with your application. If your routes for receiving POST requests match the routes that the forms are on (e.g. form is on /form and server listens for POST on /form) you can just use the default HTML form submission. However this is not a great way to do things when using React because then routing will be controlled by your server rather than by React. Instead you should use some sort of AJAX method to submit the form.
Since your server is now serving your React app (rather than React serving itself as in development), you can just make relative requests and those requests will be made to your server. For example the request (using the fetch API):
const models = await fetch('/api/models');
Will be made to your_host/api/models by default.
in the package.json add
"proxy": "http://localhost:4567"
Using Next.js - building this
I am porting my app to the next.js framework. Currently, I am re-creating the authentication system, otherwise, this iteration of my app is pretty bare-bones. All of a sudden, after adding my protected routes (Higher Order Components) - not sure if that is related - I started getting this error along with super clunky loading (obviously).
The connection to http://localhost:3000/_next/webpack-hmr was interrupted while the page was loading.
Otherwise, everything works as expected.
I have no idea how to even begin to troubleshoot this sort of thing. Does anyone have any ideas of how I might get a bit more info/insight on this problem? Guidance on how to debug? My next move is to start disconnecting things until it goes away I guess. Any help would be appreciated! Thanks
As per the resolution suggested in https://github.com/zeit/next.js/issues/9776, You can unregister the service worker if you are using CRA.
The following code shows the way to remove registered service worker:
import { unregister } from './serviceWorker'
// ...
unregister();
If you have already deployed the registered one, first you to need build this code and deploy again it will deploy with unregistered one. The reason for doing this is because the service worker is registered in your users' browsers if you've used register before. unregister removes it entirely, though. If you build you app again, the main JS bundle will get a new hash, the users will download it, and unregister will remove it for them.
Have had the same issue.
I followed recommendations from various sources I researched and none of them worked. In my case it was an issue in next.config.js file.
More specifically, in my next.config.js file I was including an async generateBuildId function but I left it blank in order to revisit the algorithm at a later stage since the project was in its early stage. After returning a valid buildID from this function the warning disappeared.
This usually happens when you are developing the app
The issue might be service work that is getting register to the browser again and again in development mode as the browsers reload fully,
open the Dev tool and check whether a service worker is a registered or not.
if it registers then unregister it, and update your next.config.js so that the service worker will not register again in the browser
I am using "next-pwa" module, below is the check which i have done
pwa: {
disable: process.env.NODE_ENV === 'development',
register: true,
scope: '/',
dest: 'public',
swSrc: 'service-worker.js',
},
My two cents: this happened to me using Firefox, connecting to localhost;
same error:
The connection to http://localhost:9009/__webpack_hmr
was interrupted while the page was loading.
Being specific I was running start-storybook -p 9009 -s public, but that's actually unrelated to the issue.
I solved this clearing Cached Web Content and Site Data at Firefox preferences: