TypeError: Failed to fetch dynamically imported module

[IPWright83]

Describe the bug

Since switching to Vite we noticed a new production issue, where sometimes users are encountering an error if we deploy while they have an active session:

TypeError: Failed to fetch dynamically imported module

I believe this is because if any code is modified in an area that Vite would turn into a dynamic module, then the file hash changes, however when they try to visit an area that would trigger the dynamic load, those files no longer exist so they hit the error message above.

Quoting from https://stackoverflow.com/a/74057337/21061

When you dynamically import a route/component, during build it creates a separate chunk. By default, chunk filenames are hashed according to their content – Overview.abc123.js. If you don't change the component code, the hash remains the same. If the component code changes, the hash changes too - Overview.32ab1c.js. This is great for caching.

Now this is what happens when you get this error:

• You deploy the application
• Your Home chunk has a link to /overview route, which would load Overview.abc123.js
• Client visits your site
• You make changes in your code, not necessarily to the Overview component itself, but maybe to some children components that Overview imports.
• You deploy changes, and Overview is built with a different hash now - Overview.32ab1c.js
• Client clicks on /overview link - gets the Failed to fetch dynamically imported module error, because Overview.abc123.js no longer exists

That is why the errors correlate with deployments. One way to fix it is to not use lazy loaded routes, but that's not a great solution when you have many heavy routes - it will make your main bundle large

What I expect to happen, is not to encounter any errors if the users session remains active during a deployment.

I have been unable to come up with a good workaround (specifically for me using React ErrorBoundary is the best I can do so far with a re-direct similar to https://stackoverflow.com/a/74861436/21061 which is a mitigation and provides quite a poor user experience flashing an error message).

Reproduction

https://github.com/IPWright83/vite-dynamic-import

Steps to reproduce

The above repository has been set up to mimick a production deployment as obviously that is a much more complicated set-up. It leverages React.lazy to force a dynamic module and uses a setTimeout to provide a delay with which to simulate a user navigation to a page requiring a module. In a real production scenario I don't believe React.lazy is required.

• Clone the above repository

• npm install

• npm run build

• open a 2nd terminal

• Terminal 1 npx serve dist (starts a web server)

• Browser open the URL (usually localhost:3000)

• Text Editor modify src/Foo.jsx changing the string "Foo" to something else (within 30s of launching the page - increase the setTimeout in App.jsx if this is not long enough)

• Terminal 2 npm run build

Wait... 30s after loading the page you should see a blank page render with errors in the browser console:

If you were to reload the page, you can see that Foo-b53985a6.js has been renamed to Foo-535d5a10.js (or similar new hash)

System Info

❯ npx envinfo --system --npmPackages '{vite,@vitejs/*}' --binaries --browsers

  System:
    OS: Linux 5.15 Ubuntu 20.04.5 LTS (Focal Fossa)
    CPU: (12) x64 Intel(R) Core(TM) i7-10750H CPU @ 2.60GHz
    Memory: 18.36 GB / 31.10 GB
    Container: Yes
    Shell: 5.0.17 - /bin/bash
  Binaries:
    Node: 16.14.0 - ~/.nvm/versions/node/v16.14.0/bin/node
    Yarn: 1.22.18 - ~/.nvm/versions/node/v16.14.0/bin/yarn
    npm: 8.3.1 - ~/.nvm/versions/node/v16.14.0/bin/npm
  Browsers:
    Chrome: 109.0.5414.74
    Chromium: 109.0.5414.74
    Firefox: 109.0

### Used Package Manager

pnpm

### Logs

_No response_

### Validations

- [X] Follow our [Code of Conduct](https://github.com/vitejs/vite/blob/main/CODE_OF_CONDUCT.md)
- [X] Read the [Contributing Guidelines](https://github.com/vitejs/vite/blob/main/CONTRIBUTING.md).
- [X] Read the [docs](https://vitejs.dev/guide).
- [X] Check that there isn't [already an issue](https://github.com/vitejs/vite/issues) that reports the same bug to avoid creating a duplicate.
- [X] Make sure this is a Vite issue and not a framework-specific issue. For example, if it's a Vue SFC related bug, it should likely be reported to [vuejs/core](https://github.com/vuejs/core) instead.
- [X] Check that this is a concrete bug. For Q&A open a [GitHub Discussion](https://github.com/vitejs/vite/discussions) or join our [Discord Chat Server](https://chat.vitejs.dev/).
- [X] The provided reproduction is a [minimal reproducible example](https://stackoverflow.com/help/minimal-reproducible-example) of the bug.

[sapphi-red]

This is not something specific to Vite and happens with Webpack, too.

https://mitchgavan.com/code-splitting-react-safely/
http://dimaip.github.io/2020/04/25/deploying-apps-with-code-splitting/

I don't have any idea to solve this on a bundler side.

[awacode21]

We are experiencing the same issues with a vite+vue SSR application.
we are experiencing thousands of those a day. also multiple days after deployment... we also saw this after our very first go-live, where no "previous version" was online. We are completely desperate since one week on how to fix this.
Other strange thing is, sometimes we get this error logged in Sentry, and when we try to check if the file really is not available, we are able to access it. Others are not working

[IPWright83]

Thanks @sapphi-red, is it Vite or Rollup generating these hashes (assuming that's what they are) on the filenames? I wonder if there's a way to suppress that behaviour and if it's required for cache busting whether there's any alternative approaches.

[victorlmneves]

@sapphi-red both generate hashes when running the build. If your product it's a SPA you can try to use a service worker together with an interceptor (if you use Axios) to manage that. I did it on Vue 2 where it's basically stores on the localStorage a version of the app and then compares if the one that you access it's the same or not and if not it triggers a reload of the browser in order to get the updated version

[IPWright83]

@victorlmneves could you provide a bit of a fuller explanation/ code snippets maybe of tha so I understand the concepts a bit more? I'd be quite interested in exploring what that might look like.

[victorlmneves]

@IPWright83
https://dev.to/voodu/vue-3-pwa-service-worker-12di
https://medium.com/@FezVrasta/service-worker-updates-and-error-handling-with-react-1a3730800e6a

And here is an example that I did some years ago

import { register } from "register-service-worker"

if (process.env.NODE_ENV === "production") {
  register(`${process.env.BASE_URL}service-worker.js`, {
    ready() {
      console.log(
        "App is being served from cache by a service worker.\n" +
          "For more details, visit https://goo.gl/AFskqB"
      )
    },
    registered() {
      console.log("Service worker has been registered.")
    },
    cached() {
      console.log("Content has been cached for offline use.")
    },
    updatefound() {
      console.log("New content is downloading.")
    },
    updated() {
      console.log("New content is available please refresh.")
      caches.keys().then(names => {
        for (let name of names) {
          console.log(name)
          caches.delete(name)
        }
      })
      setTimeout(() => {
        window.location.reload()
      }, 1000)
    },
    offline() {
      console.log(
        "No internet connection found. App is running in offline mode."
      )
    },
    error(error) {
      console.error("Error during service worker registration:", error)
    }
  })
}

And another one using Axios interceptors

axios.interceptors.response.use((resp) => {
  let v = resp.headers['vers'] || 'default'

  if (v !== localStorage.getItem('vers') && resp.config.method == 'get') {
    localStorage.setItem('vers', v)
    window.location.reload() // For new version, simply reload on any get
  }

  return Promise.resolve(resp)
})

[sapphi-red]

@IPWright83
You can remove these hashes from the filename by configuring entryFileNames and chunkFileNames.

import { defineConfig } from 'vite'

export default defineConfig({
  build: {
    rollupOptions: {
      entryFileNames: '[name].js',
      chunkFileNames: '[name].js'
    }
  }
})

But yes this is recommended for cache busting.

I don't know any alternatives other than the ones described in the articles I linked above.

[victorlmneves]

@IPWright83 by the way, have you tried to access directly the file that is displayed on the console that wasn't possible to import? I'm asking because on our project we get those errors randomly but when we access them directly the file exist

[IPWright83]

I have @victorlmneves. Is there any way to adopt an approach more like this #5825 @sapphi-red? Anything after the queryString can still cache-bust, but as the file name is constant you'll get the new payload. I think that example is a little primitive (I think JS needs transforming too). Could you see any issues with this approach?

[sapphi-red]

I guess it's possbile by using a plugin.
I think that will lead to runtime errors sometimes because the dynamic import in old version files can resolve to new version files.

For example,
first version deployed

// foo.js
export const foo = 'foo'
export const bar = 'bar'

// a file depends on foo.js
import('./foo.js').then(mod => {
  if (mod.bar.startsWith('b')) {
    console.log(mod.foo)
  }
})

second version deployed

// foo.js
export const foo = 'foo'

// a file depends on foo.js
import('./foo.js').then(mod => {
  console.log(mod.foo)
})

[victorlmneves]

@IPWright83
Meanwhile, I also saw someone handling the problem this way

router.onError((error, to) => {
  if (error.message.includes('Failed to fetch dynamically imported module')) {
    window.location = to.fullPath
  }
})

[IPWright83]

@victorlmneves yeah, that works for vue-router and is in the StackOverflow link I shared. Unfortunately we're using react-router through various indirection connected-react-router & react-router-config. So even if it did expose an onError handler which I've been unable to find, we'd struggle calling it :(

[mikemklee]

Any updates on this?
Also running into the same issue with a vue 2 + vite app, when using dynamic imports.

The vue router hack could work for some imports, but we have dynamic imports outside of the router, which i don't think the hack would be able to cover.

Besides, forcing a reload because an error occurred doesn't seem like the best UX, so wondering if there's a solution/workaround that is more subtle and "behind the scenes" 🤔

[yakobe]

While I understand that we should try to solve this problem ourselves, it seems like this must be quite a fundamental problem that developers will have when creating apps with vite? Especially when continuous deployment comes into the mix, because the files update very often and trigger lots of errors. For now I am back to "nightly deployments" to try and minimize disruption 😬.

It would be great if there was a standard solution somewhere in the docs. Preferably one without making the users manually "opt-in" to a page reload. Even nicer would be if vite could somehow handle it for us so we don't have to do anything. Eg catch the error and refresh the files without a hard reload of the page. That would greatly improve the DX 🙂.

Thanks for all the great work. Vite is awesome.

[victorlmneves]

@yakobe in my case, it's not about outdated chunks. I'm getting this with existing files that is even worst to figure out the problem and try to fix it, especially because the only way I'm able to reproduce the issue it's shutting down the node server (I have SSR app) :/

[lzacharkow]

For what it's worth, an ad-blocker was the root of this issue for me. Disabling it resolved my issue.

[lmiller1990]

A lot of Cypress users run into this: cypress-io/cypress#25913

I don't think I can fix it on the Cypress end. I can spend time (several days of development resources) to patch Vite if needed, but it looks like the cause isn't entirely clear.

Some observations:

1. Happens mostly with large bundles - like MUI (Material UI).
2. Especially happens on CI with Docker + many instances running (eg, race condition/resource issue)?

Any tips -- maybe from someone more familiar with Vite - on debugging this?

I will inspect Cypress more first and keep the limitations in mind.

[benadamstyles]

We are just hitting this now too, but it's not just when the user has a browser tab open. It can also happen much later, if the user returns to our app after we have deployed. We're trying to work out if it's caused by the index.html page being cached for too long, rather than an issue with the JS file hashes themselves. Maybe it helps someone who is experiencing this? My reasoning goes like this:

• Let's say you have index.html which references index.abc.js which dynamically imports moduleX.123.js.
• You update your code and deploy it, so now you have index.def.js which dynamically imports moduleX.456.js. This should not cause any "module not found" issues because all the file hashes are new, even the entry index JS file.
• However, if your index.html file has been cached for longer, it still contains the reference to index.abc.js, which dynamically imports the now deleted moduleX.123.js. Hence you get an error.

I'm still confirming this but I wondered if it matches anyone else's experience/setup?

[AlfonsoUceda]

Happened to me after upgrading from vite 4.0.1 to 4.1.4

[adamkpurdy]

@benadamstyles I had the same issue. I removed the hashing option in the config like another user mentioned.

[aakash14goplani]

I am facing same issue with SvelteKit + Vite.

[eifr-vault]

Not sure if related but I'm getting the same error when using storybook with a react component that imports a module from an index.ts file (exports all others)

[namu254]

ad-blocker

same

[lmiller1990]

I reproduced this in CI only, you can follow my progress as I debug it in Cypress: cypress-io/cypress#25913 (comment) I suspect the same core issue. I think it's a race condition or resources related issue.

[aethr]

I think there are possibly two separate issues being described by different users/teams in this thread:

1. failed to fetch after a new deployment where the old version was deleted from the server

• in this case, attempting to load the resource referenced by the error is expected to fail, it has been deleted

2. failed to fetch that occurs somewhat randomly

• in this case, attempting to load the resource succeeds

For people experiencing issue 1., the solution is to organise your deployments so that old versions aren't deleted immediately. Doing a deployment of new app code doesn't mean that all users running your app in their browser will update immediately (unless you force it somehow).

For those users who still have an older version running in their browser (lets say index.123.js), navigating to routes that load chunks dynamically will always reference the chunk that was current when they loaded your app (ie home.123.js), and this is correct. You don't want the user getting half of their chunks with new app code and half with old, this would lead to very hard to debug issues. If you delete those chunks while the user is still navigating around the app, you will hit this error when they change routes and try to load a chunk they haven't loaded yet.

For users that load the app after a deployment, index.html will reference a new index.456.js chunk that references all the other chunks that are consistent with that version (ie home.456.js if there are also changes in that chunk). The only difference is which version was referenced by index.html when the user loaded the tab.

If you keep a manifest of each deployment, and your instrumentation/logging includes the "version" of the app users are using, you should be able to figure out how long old versions are still used. My team has seen users still operating in the same tab they loaded a month ago without updating, so we keep all our built assets deployed for a long time. If you want to force users to reload when a new version is deployed, that's a whole piece of engineering in itself.

[lmiller1990]

According to quite a bit of research in cypress-io/cypress#25913 at least one class of error is related to CI on GH Actions specifically - not sure if this helps anyone, but worth noting in case it does.

[azz0r]

Theres no quick work around for this? I'm seeing this after switching to vite from webpack but only on certain route files - despite them all following the same react lazy pattern.

[liamvdv]

see https://vitejs.dev/guide/build#load-error-handling for a (user visible) fix. you may include a dynamic import early on in your app and trigger the dynamic import error and then issue a re-render.

[AndrewJHart]

This actually just started happening to us - we did not get the errors before with vite even though the routes were lazy loaded. Thanks for all the useful info here to everyone who contributed.

[zuhaib10]

@liamvdv this solved my blank page issue when the new build was pushed. Thanks

[SecretCastle]

Is it also a way to add cache-control: no-store, max-age=0 in the response header of the proxy server or nginx?

[liamvdv]

@SecretCastle No, see this example:

Let's say, to reduce the initial bundle size, you choose to bundle by page. Let's walk through what that means:

1. User requests /. Nginx is instructed to serve the same JS for every path, since it is a Single Page Application. The Browser loads index.js which contains a client side router [e.g. react-dom-router / NEXT.js implicitly]. The router mounts and sees that you want to see component A for '/', which triggers a request to 'chunk-component-a.js'. Let's assume thats going OK.
2. Now, the user now scrolls on Page '/' and after some time finds some interesting thing and clicks a button, initiating a redirect to '/abc'. The router knows to now show you component B. While the user was happily searching through Page '/', you redeployed and re-bundled your js.
3. Because you bundled by page, the router now fetch the JS chunk for component B. That's called lazy loading -- it's loaded when needed. The router now loads 'chunk-component-b.js' from your server.
4. Sadly, chunk-component-b.js is not known, because you renamed component B to component Z. The webserver serving the JS only knowns about 'chunk-component-z.js' after the redeployment. Consequently, the lazy load of '/abc' fails.

Note: For ease of understanding I ignored that Vite does not produce chunks in the form of chunk-[component-name].js. It uses the first few characters of the sha256 - a hash of content of the file that is to be lazy loaded. This means that any change to your file will result in a different chunk filename and thus lead to stale references in code loaded prior to deployment.

In essence, the index.js contains a map of lazy imports('some-file.js') statements to files on you server (example.com/chunk-xxxx.js). That map becomes stale in redeployments because the chunk filenames for components differ.

That's also why choosing explicit chunk filenames can relieve the problem, as you can choose chunk filenames that will remain correct between deployments.

Caches won't help when redeployments change your filename -- but they do make it worse, as they prolong the span in which errors occur if you do not invalidate the cache.

The full solution to this problem is to keep past deployment chunk-xxxx.js files around. Because they are shorted sha256 hashes, it serves like a content-addressable filesystem and thus new files only overwrite old deployment files if the content of the files are the same, in which case we do not care.

Hope that helps everyone wrap their head around the problem. Have a nice day!

[SecretCastle]

@liamvdv thanks. Your explanation is excellent. I understand that in this case, the essence is that the user did not refresh the page and the request was still for the historical bundle. I don't know if my understanding is correct.

[geakstr]

I saw this error when vitest started to fail on some of the tests, I cleaned up vite cache (rm -rf node_modules/.vite) and now it works again

[ColiZei]

I am using a aws infrastructure with vue3, vue router, and vite.

The workaround from vite (https://vitejs.dev/guide/build#load-error-handling) does not solve the problem.
Every time the refresh is triggered, the error does not disappear and i am not able to navigate with the router ...
Blank page and refresh on navigation click.

Any progress / help?

Thx!

[ilijaag]

This happens on Angular 17 with SSR when new deployment happens. Builder @angular-devkit/build-angular:application which uses Vite under the hood. Any updates on this?

[leungk712]

@ilijaag

https://vitejs.dev/guide/build#load-error-handling

I don't use Angular but this seems to be the solution from what ppl above have noted. This would go in the Angular file equivalent to main.tsx/index.tsx file (for React)

[JasonLuo2024]

The errors is actually kinda misleading. If you open the files that failed to be imported, some may contain "Cannot find static asset".

{"url":"/_nuxt/9Ima4rhO.js","statusCode":404,"statusMessage":"Cannot find static asset /_nuxt/9Ima4rhO.js","message":"Cannot find static asset /_nuxt/9Ima4rhO.js","stack":""}

In my case, adding "routeRules" to ''nuxt.config.ts'' resolves the issue.

routeRules: {
"/**": { isr: true },
},

[kylebehse]

For me it was the server config. It was not recognizing .mjs extentions as javascript. .mjs is a new javascript extention that represents modules. So I updated my .htaccess to include this:
AddType application/javascript .mjs

[HaNdTriX]

We had a similar issue. It was caused by a workbox precache manifest and the cache eviction functionality of google chrome. In our app the service worker was caching opague responses, leading to exceeded storage quota.

This lead to the eviction of the dynamic module resulting in the error above.

If you are using service workers or had been using service workers:

• check if the service worker is installed. If it is remove it and try to reproduce.
• check your storage quota
• check if you have opaque responses in your cache storage
• update your browser (the latest version added some additional information regarding cache eviction)

hope this helps!

[ilijaag]

Quick update guys, for Angular with es-build, the issue was forgotten `"outputHashing": "all" in the angular.json (or project.json) for the production build. Without that the main.js was not with hashing in the output build and then it was cached when new deploy was done and it was looking in chunk files with the old hashings.

[dhust]

A bad autocomplete in VSCode put this in one of my files.
import { USERNAME } from '$env/static/private';

[ritute]

Was running into this too, and found that Vite now emits a vite:preloadError event and they suggest a simple solution for this issue here to reload when this happens:

https://vitejs.dev/guide/build#load-error-handling

But it also seems like this issue has to do with your hosting service. In our case, we're using Heroku which seemingly isn't keeping older versions around, and perhaps there's a way to configure this to keep them for some time as an alternative workaround.

[pratikpatel-aurochs]

@IPWright83 Meanwhile, I also saw someone handling the problem this way

router.onError((error, to) => {
  if (error.message.includes('Failed to fetch dynamically imported module')) {
    window.location = to.fullPath
  }
})

🚀