FATAL ERROR when connecting inside Node.js worker thread (Nitro plugin)

[DrJume]

Environment Information

• OS [e.g. Mac, Arch, Windows 10]: Mac
• Node Version [e.g. 8.2.1]: 22.14.0
• NPM Version [e.g. 5.4.2]: 11.3.0 (pnpm v10.8.1)
• C++ Toolchain [e.g. Visual Studio, llvm, g++]:
• confluent-kafka-javascript version [e.g. 2.3.3]: 1.2.0

Steps to Reproduce

1. setup basic Nitro dev environment (https://nitro.unjs.io)
2. connect a producer inside a Nitro plugin (it is run on server start)
3. npm run dev

confluent-kafka-javascript Configuration Settings

Additional context

• node-rdkafka has the same issue and they have an open PR:
  Fix worker_threads support Blizzard/node-rdkafka#1110

• existing similar issue in Nitro: Issue with serialport in dev nitrojs/nitro#3147

FATAL ERROR: v8::HandleScope::CreateHandle() Cannot create a handle without a HandleScope
----- Native stack trace -----

 1: 0x1041d70a0 node::OnFatalError(char const*, char const*) [/Users/drjume/.local/bin/node]
 2: 0x1043d7128 v8::Utils::ReportApiFailure(char const*, char const*) [/Users/drjume/.local/bin/node]
 3: 0x10457d054 v8::internal::HandleScope::Extend(v8::internal::Isolate*) [/Users/drjume/.local/bin/node]
 4: 0x1043ec394 v8::Object::GetCreationContextChecked() [/Users/drjume/.local/bin/node]
 5: 0x1040e1f4c node::InternalMakeCallback(v8::Isolate*, v8::Local<v8::Object>, v8::Local<v8::Function>, int, v8::Local<v8::Value>*, node::async_context, v8::Local<v8::Value>) [/Users/drjume/.local/bin/node]
 6: 0x12470cf14 Nan::Callback::Call_(v8::Isolate*, v8::Local<v8::Object>, int, v8::Local<v8::Value>*, Nan::AsyncResource*) const [/Users/drjume/Documents/Dev/.../node_modules/.pnpm/@confluentinc+kafka-javascript@1.2.0/node_modules/@confluentinc/kafka-javascript/build/Release/confluent-kafka-javascript.node]
 7: 0x124709200 Nan::Callback::Call(int, v8::Local<v8::Value>*) const [/Users/drjume/Documents/Dev/.../node_modules/.pnpm/@confluentinc+kafka-javascript@1.2.0/node_modules/@confluentinc/kafka-javascript/build/Release/confluent-kafka-javascript.node]
 8: 0x12470a2a4 NodeKafka::Callbacks::EventDispatcher::Flush() [/Users/drjume/Documents/Dev/.../node_modules/.pnpm/@confluentinc+kafka-javascript@1.2.0/node_modules/@confluentinc/kafka-javascript/build/Release/confluent-kafka-javascript.node]
 9: 0x104de010c uv__async_io [/Users/drjume/.local/bin/node]
10: 0x104df3920 uv__io_poll [/Users/drjume/.local/bin/node]
11: 0x104de0674 uv_run [/Users/drjume/.local/bin/node]
12: 0x1040e2508 node::SpinEventLoopInternal(node::Environment*) [/Users/drjume/.local/bin/node]
13: 0x104220154 node::NodeMainInstance::Run() [/Users/drjume/.local/bin/node]
14: 0x104196364 node::Start(int, char**) [/Users/drjume/.local/bin/node]
15: 0x194ff2b4c start [/usr/lib/dyld]

FATAL ERROR: HandleScope::HandleScope Entering the V8 API without proper locking in place
----- Native stack trace -----

 1: 0x1048170a0 node::OnFatalError(char const*, char const*) [/Users/drjume/.local/bin/node]
 2: 0x104a17128 v8::Utils::ReportApiFailure(char const*, char const*) [/Users/drjume/.local/bin/node]
 3: 0x104a17f50 v8::HandleScope::HandleScope(v8::Isolate*) [/Users/drjume/.local/bin/node]
 4: 0x104721668 node::InternalCallbackScope::InternalCallbackScope(node::Environment*, v8::Local<v8::Object>, node::async_context const&, int, v8::Local<v8::Value>) [/Users/drjume/.local/bin/node]
 5: 0x104721b90 node::InternalMakeCallback(node::Environment*, v8::Local<v8::Object>, v8::Local<v8::Object>, v8::Local<v8::Function>, int, v8::Local<v8::Value>*, node::async_context, v8::Local<v8::Value>) [/Users/drjume/.local/bin/node]
 6: 0x104721fd0 node::InternalMakeCallback(v8::Isolate*, v8::Local<v8::Object>, v8::Local<v8::Function>, int, v8::Local<v8::Value>*, node::async_context, v8::Local<v8::Value>) [/Users/drjume/.local/bin/node]
 7: 0x11cff4f14 Nan::Callback::Call_(v8::Isolate*, v8::Local<v8::Object>, int, v8::Local<v8::Value>*, Nan::AsyncResource*) const [/Users/drjume/Documents/Dev/.../node_modules/.pnpm/@confluentinc+kafka-javascript@1.2.0/node_modules/@confluentinc/kafka-javascript/build/Release/confluent-kafka-javascript.node]
 8: 0x11cff1200 Nan::Callback::Call(int, v8::Local<v8::Value>*) const [/Users/drjume/Documents/Dev/.../node_modules/.pnpm/@confluentinc+kafka-javascript@1.2.0/node_modules/@confluentinc/kafka-javascript/build/Release/confluent-kafka-javascript.node]
 9: 0x11cff22a4 NodeKafka::Callbacks::EventDispatcher::Flush() [/Users/drjume/Documents/Dev/.../node_modules/.pnpm/@confluentinc+kafka-javascript@1.2.0/node_modules/@confluentinc/kafka-javascript/build/Release/confluent-kafka-javascript.node]
10: 0x10542010c uv__async_io [/Users/drjume/.local/bin/node]
11: 0x105433920 uv__io_poll [/Users/drjume/.local/bin/node]
12: 0x105420674 uv_run [/Users/drjume/.local/bin/node]
13: 0x104722508 node::SpinEventLoopInternal(node::Environment*) [/Users/drjume/.local/bin/node]
14: 0x104860154 node::NodeMainInstance::Run() [/Users/drjume/.local/bin/node]
15: 0x1047d6364 node::Start(int, char**) [/Users/drjume/.local/bin/node]
16: 0x194ff2b4c start [/usr/lib/dyld]

[constb]

Sadly, it seems, the problem persists. I hit that same bump when running tests using vitest – it still crashes with the same error.

[DrJume]

I created a workaround using pnpm dependency patching: I patched nitro to run the dev server inside a childProcess instead of a worker:

diff --git a/dist/core/index.mjs b/dist/core/index.mjs
index f7a6143b63977ec90dcc6a093a2e6466992ef71f..ba6e550451410f628672c4b3ad9b3e4aa2d3fa14 100644
--- a/dist/core/index.mjs
+++ b/dist/core/index.mjs
@@ -151,6 +151,27 @@ const NitroDefaults = {
   }
 };
 
+import { fork } from 'node:child_process'
+
+function ChildProcess(path, options) {
+  const child = fork(path, {
+    ...options,
+  })
+
+  return Object.assign(child, {
+    postMessage: (message) => {
+      child.send(message)
+    },
+    terminate: () => {
+      return new Promise((resolve) => {
+        child.kill()
+        resolve()
+      })
+    },
+    _exitCode: undefined,
+  })
+}
+
 async function resolveAssetsOptions(options) {
   for (const publicAsset of options.publicAssets) {
     publicAsset.dir = resolve(options.srcDir, publicAsset.dir);
@@ -2295,7 +2316,7 @@ class NodeDevWorker {
       this.close(`worker entry not found in "${workerEntryPath}".`);
       return;
     }
-    const worker = new Worker(workerEntryPath, {
+    const worker = ChildProcess(workerEntryPath, {
       env: {
         ...process.env,
         NITRO_DEV_WORKER_ID: String(this.#id)
@@ -2353,6 +2374,7 @@ class NodeDevWorker {
           if (process.env.DEBUG) {
             consola.warn(`force closing dev worker...`);
           }
+          resolve()
         }, gracefulShutdownTimeoutSec * 1e3);
         this.#worker?.on("message", (message) => {
           if (message.event === "exit") {
diff --git a/dist/presets/_nitro/runtime/nitro-dev.mjs b/dist/presets/_nitro/runtime/nitro-dev.mjs
index 3848eb7467469e42fda8091d52b5808bfbb50825..c911215f0da090913bf226fca510abb9d5104bc8 100644
--- a/dist/presets/_nitro/runtime/nitro-dev.mjs
+++ b/dist/presets/_nitro/runtime/nitro-dev.mjs
@@ -8,7 +8,7 @@ import { scheduledTasks, tasks } from "#nitro-internal-virtual/tasks";
 import { Server } from "node:http";
 import { join } from "node:path";
 import nodeCrypto from "node:crypto";
-import { parentPort, threadId } from "node:worker_threads";
+import { threadId } from "node:worker_threads";
 import wsAdapter from "crossws/adapters/node";
 import {
   defineEventHandler,
@@ -20,6 +20,10 @@ import {
 if (!globalThis.crypto) {
   globalThis.crypto = nodeCrypto;
 }
+const parentPort = {
+  on: process.on.bind(process),
+  postMessage: process.send.bind(process),
+}
 const { NITRO_NO_UNIX_SOCKET, NITRO_DEV_WORKER_ID } = process.env;
 trapUnhandledNodeErrors();
 parentPort?.on("message", (msg) => {

In the current nitro version it's only possible to set a server preset (worker/child_process) for the production build. This patch patches the default dev preset to use a Node child_process.