Discarded requests whilst render worker process is restarting

[NadhifRadityo]

Verify canary release

• I verified that the issue exists in the latest Next.js canary release

Provide environment information

Operating System:
      Platform: win32
      Arch: x64
      Version: Windows 10 Pro
    Binaries:
      Node: 18.12.1
      npm: N/A
      Yarn: N/A
      pnpm: N/A
    Relevant Packages:
      next: 13.4.13-canary.6
      eslint-config-next: 13.4.12
      react: 18.2.0
      react-dom: 18.2.0
      typescript: 5.1.6
    Next.js Config:
      output: N/A

Which area(s) of Next.js are affected? (leave empty if unsure)

No response

Link to the code that reproduces this issue or a replay of the bug

https://codesandbox.io/p/sandbox/nextjs-discarded-requests-whilst-render-worker-process-is-restarting-y6vq47

To Reproduce

1. Create an appDir API route that simulates a long request

// file: app/test/route.ts
import { NextRequest, NextResponse } from "next/server";

export const dynamic = "force-dynamic";

export async function GET(request: NextRequest) {
	await new Promise(res => setTimeout(res, 2000));
	return new NextResponse(`${Date.now()}`);
}

2. Build and run the project.

npm run build && npm run start

2. Flood the API routes with requests. In this case, I am using Apache benchmark, you could use Apache JMeter or curl with loops instead.

ab -n 100 -c 10 http://localhost:3000/test

3. While it's still requesting, find the PID of the render worker and kill it. (I run npm run start in debug mode on VSCode so that I can see the PID tree directly)

taskkill /f /pid <PID>

4. Errors logged out in the app console. Such as, but not limited to:

TypeError: fetch failed
    at Object.fetch (node:internal/deps/undici/undici:11118:11)
    at async invokeRequest (E:\Projects\Web\nextsocketissue\node_modules\next\dist\server\lib\server-ipc\invoke-request.js:21:12)
    at async invokeRender (E:\Projects\Web\nextsocketissue\node_modules\next\dist\server\lib\router-server.js:226:29)
    at async handleRequest (E:\Projects\Web\nextsocketissue\node_modules\next\dist\server\lib\router-server.js:419:24)
    at async requestHandler (E:\Projects\Web\nextsocketissue\node_modules\next\dist\server\lib\router-server.js:436:13) {
  cause: Error: read ECONNRESET
      at TCP.onStreamRead (node:internal/stream_base_commons:217:20) {
    errno: -4077,
    code: 'ECONNRESET',
    syscall: 'read'
  }
}

Error: aborted
    at connResetException (node:internal/errors:717:14)
    at Socket.socketCloseListener (node:_http_client:455:19)
    at Socket.emit (node:events:523:35)
    at TCP.<anonymous> (node:net:335:12) {
  code: 'ECONNRESET'
}

Note:

1. You may need to repeat step 3 multiple times before getting an error in the console. This step requires perfect timing between Apache benchmark requesting and you killing the render worker process.
2. You also may get a different error message depending on which next.js version you use. I get a different one with next@latest and next@canary. But the error messages essentially say the same thing, ECONNRESET.

Describe the Bug

This bug is caused by the render worker restarting whilst it is still rendering the page. The restart mechanism works fine*, but it suffers when there is a request happening.

I originally discuss the issue in #51605 (#51605 (comment), #51605 (comment)), but figured it would be better creating a new one because I think this is a different problem.

Errors generated in this issue are also consistent with worker process restarting due to high memory usage. #49929 (comment)

* in my machine, with a small project. I noticed in a bigger project, sometimes the restart mechanism is broken, and I need to hard restart the whole app.

Expected Behavior

Next.js could rerender the request again after the render worker restarted, but this would not be possible if the response had already been sent partially. An easier solution would be to drop the request altogether and display an internal server error page instead.

Which browser are you using? (if relevant)

No response

How are you deploying your application? (if relevant)

next start, locally

[ThijsZijdel]

"next": "13.4.17",
Error appears after about 4-5 small dev file changes inside localhost and won't go away.
Removing .next and re-running start dev fixes it temporary.

TypeError: fetch failed
    at Object.fetch (node:internal/deps/undici/undici:11118:11)
    at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
    at async invokeRequest (../node_modules/.pnpm/next@13.4.17_@babel+core@7.22.10_react-dom@18.2.0_react@18.2.0/node_modules/next/dist/server/lib/server-ipc/invoke-request.js:17:12)
    at async invokeRender (../node_modules/.pnpm/next@13.4.17_@babel+core@7.22.10_react-dom@18.2.0_react@18.2.0/node_modules/next/dist/server/lib/router-server.js:253:29)
    at async handleRequest (../node_modules/.pnpm/next@13.4.17_@babel+core@7.22.10_react-dom@18.2.0_react@18.2.0/node_modules/next/dist/server/lib/router-server.js:446:24)
    at async requestHandler (../node_modules/.pnpm/next@13.4.17_@babel+core@7.22.10_react-dom@18.2.0_react@18.2.0/node_modules/next/dist/server/lib/router-server.js:463:13)
    at async Server.<anonymous> (../node_modules/.pnpm/next@13.4.17_@babel+core@7.22.10_react-dom@18.2.0_react@18.2.0/node_modules/next/dist/server/lib/start-server.js:117:13) {
  cause: Error: connect ECONNREFUSED ::1:50682
      at TCPConnectWrap.afterConnect [as oncomplete] (node:net:1300:16) {
    errno: -61,
    code: 'ECONNREFUSED',
    syscall: 'connect',
    address: '::1',
    port: 50682
  }

Reverting back to "next": "13.4.13" removes the error fully.

[balazsorban44]

Hi, we no longer rely on render workers (they got removed). Please upgrade to next@latest and open a new issue if you still experience this. Thank you!

[github-actions]

This closed issue has been automatically locked because it had no new activity for 2 weeks. If you are running into a similar issue, please create a new issue with the steps to reproduce. Thank you.