500 Internal Server Error shows in random moments from random endpoint methods

[nabiedma]

Summary

So, basically we have deployed our finished project to Vercel. Locally the project works great. We are using Next.js 14.1 (but with Pages Router) and using Lucia Auth for our Authentication with a MongoDB database. In localhost, the project runs great and without errors. But once deployed into Vercel (have double checked to have the correct .env variables), when logging into the project, 1 of 2 endpoints that are being called from the /api routes "crashes" for no reason. Taking a look at the logs, they are not very useful since they only show:
500 Internal Server Error
Node.js process exited with exit status: 129. The logs above can help with debugging the issue.

And there is no more log info. Last week it was showing more information, couldn't get a copy or screenshot of that but it was also telling something about a response from node that was being called after it was already ended (response.end())

Example

No response

Steps to Reproduce

The endpoints that give errors (sometimes is one, sometimes is the other one) are coded liked this:

/api/v1/users/getAllUsersButAdmins

import connectDB from "@/lib/db";
import UserService from "@/db/services/user";
import type { NextApiRequest, NextApiResponse } from "next";

export default async function handler(
  req: NextApiRequest,
  res: NextApiResponse
) {
  if (req.method === "GET") {
    try {
      await connectDB();
      const userService = new UserService();
      const users = await userService.getAllUsersButAdmins();

      if (!users) res.status(404).json({ message: "Users not found" });
      res.status(200).json(users);
    } catch (e) {
      console.error(e);
    }
  }
}

/api/v1/gyms

import connectDB from "@/lib/db";
import type { NextApiRequest, NextApiResponse } from "next";
import GymService from "@/db/services/gym";

export default async function handler(
  req: NextApiRequest,
  res: NextApiResponse
) {
  if (req.method === "GET") {
    try {
      await connectDB();
      const gymService = new GymService();
      const gyms = await gymService.getAllGyms();

      if (!gyms) res.status(404).json({ message: "Gyms not found" });
      res.status(200).json(gyms);
    } catch (e) {
      console.error(e);
    }
  }
}

I have already tried changing the final responses for something like:

res.status(200).json(gyms);
res.end();

But it is the same thing. The crazy thing is that this is also happening in random moments. Maybe I have navigated to other pages of the project, and when I come back to the main page, some endpoint crashes. And locally they work smoothly and correctly.
Maybe it's also useful to say that we are using Node 20.9 locally to run the project.
Have searched for other similar posts but they do not have answers from Vercel team or they don't get to any solution.

Thanks in advance for the time!

[QuiiBz]

Hi, could you try to redeploy your application and let us know if you're still having issues? We've recently fixed an edge-case regarding applications that returned a connection: keep-alive header, but I'm not sure if your issue is related.

You mentioned that the applications crashes with an exit code of 129, which means your code raised an uncaught exception.

[nabiedma]

Hey @QuiiBz, thanks for writing a reply.
Just redeployed it, and tested, but the same keeps happening. Something weird is I do not have any log tracked now 🤔 In other moments I could see instantly the logs of the error and other type of logs. But now I don't see anything.

Will check again later. But unfortunately this edge case you have mentioned don't seem to solve this strange issue. Thanks!

[nabiedma]

On the logs page, now I am getting this response:

Uncaught Exception: Error [ERR_STREAM_WRITE_AFTER_END]: write after end
at ServerResponse.end (node:_http_outgoing:1047:15)
at K.r.end (/var/task/node_modules/next/dist/compiled/next-server/pages-api.runtime.prod.js:20:13470)
at /var/task/node_modules/next/dist/compiled/next-server/pages-api.runtime.prod.js:20:14284
at K.r.send (/var/task/node_modules/next/dist/compiled/next-server/pages-api.runtime.prod.js:20:14293)
at K.r.json (/var/task/node_modules/next/dist/compiled/next-server/pages-api.runtime.prod.js:20:14376)
at l (/var/task/.next/server/pages/api/v1/auth/login.js:1:5237)
at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
at async K (/var/task/node_modules/next/dist/compiled/next-server/pages-api.runtime.prod.js:20:16545)
at async U.render (/var/task/node_modules/next/dist/compiled/next-server/pages-api.runtime.prod.js:20:16981)
at async r3.runApi (/var/task/node_modules/next/dist/compiled/next-server/server.runtime.prod.js:17:41752) {
code: 'ERR_STREAM_WRITE_AFTER_END'
}
Node.js process exited with exit status: 129. The logs above can help with debugging the issue.

If it's useful, it's happening on the:
/api/v1/gyms GET endpoint

[nabiedma]

Still waiting for a reply...

We tried changing all the Axios request to the API from axios to fetch(), but it's still the same. We keep receiving 500 errors from random endpoints, sometimes at the beginning of the application or anywhere.
Is there a way we can have a better log of this? This is really frustrating.

Thanks

[luthfirobbaniy]

Hello @nabiedma, I also encountered a this similar "random" error and have solved it (even though it's not Next.js). This is the solution that worked for me:

Is your MongoDB server located far away from your deployed app in Vercel? If this true, You can try to move them as close as possible so that your server and app have closer locations (to reduce latency). You can refer to this link about how to change the region from the Vercel dashboard. Also add "regions" to your vercel.json configuration file.

// vercel.json
{
  "regions": ["sfo1"] // use your preferred region code
}

[QuiiBz]

Could you share a minimal reproduction? Looking at the error (ERR_STREAM_WRITE_AFTER_END), it seems to be caused by something in your code, more specifically at api/v1/auth/login.js:1:5237.