cluster: worker doesn't react on message

[relloccate]

Hi there.

I have a problem:
The Worker doesn't react on "message" event from the Primary with: type: "module" enabled in package.json
But the Primary is react on "message" event from the worker

Without type: "module" both works.

win10 x64 (node v14.17.5 and v16.7.0)
Guys, any fixes? Thanks.

Example:

import cluster from 'cluster';

if (cluster.isMaster) {
    const worker = cluster.fork();

    worker.on('message', message => {
        // will log the message
        console.log(message);
    });

    worker.send('From Master');
} else {
    process.on('message', message => {
        // will not log the message
        console.log(message);
    });

    process.send('From Worker');
}

Doesn't work ONLY with enabled node.js imports.

[PurnashisHazra]

Could you create a repo which replicates this problem? Can I take this issue up?

[relloccate]

Could you create a repo which replicates this problem? Can I take this issue up?

The example above, just a minimal and illustrative example. Pretty simple to test on your PC. Sorry, I did not understand why you need a repo.

[PurnashisHazra]

Just with some tests and all for easy debugging . Nevermind, I'll take up this issue

[Linkgoron]

This is probably a duplicate of #37782 and #39140

e.g. this works:

import cluster from 'cluster';

if (cluster.isMaster) {
    const worker = cluster.fork();
    worker.on('message', message => {
        console.log(message);
    });
    setTimeout(() => {
        worker.send('From Master');
    }, 1000);
} else {
    process.on('message', message => {
        console.log(message);
    });

    process.send('From Worker');
}

tl;dr of the other issues, there's a minor race condition (read: bug) because modules are now loaded asynchronously and when starting a module messages are missed as you don't register in time to get them. The safest way to start without a race condition is to get a ping from the child and then you know you're safe. e.g. POC:

import cluster from 'cluster';
import { once } from 'events';

if (cluster.isMaster) {
    const worker = cluster.fork();

    let firstMessage = true
    worker.on('message', message => {
        if (firstMessage) {
            firstMessage = false;
        } else {
            console.log(message);
        }
    });
    await once(worker, 'message');
    worker.send('From Master');
} else {
    process.on('message', message => {
        console.log(message);
    });

    process.send('IM READY');
    process.send('From Worker');
}

[relloccate]

Yes, i saw #37782 after.
"modules are now loaded asynchronously" - With babel or no ESM it works correctly without any timeouts.
Anyway thank you, i did it and it worked.

My Solution:

import cluster from 'cluster';

if (cluster.isMaster) {
    const worker = cluster.fork();

    worker.on('message', message => {
        if (message === 'getPayload') {
            worker.send({ payload: 'something' });
        }
    });
} else {
    process.on('message', ({ payload }) => {
        console.log(payload);
    });

    process.send('getPayload');
}

[relloccate]

Dupe of #37782

[Linkgoron]

Yes, i saw #37782 after.
"modules are now loaded asynchronously" - With babel or no ESM it works correctly without any timeouts.

Just to clarify - when I said modules what I meant was that ESM modules (as in type: module or .mjs) are loaded asynchronously (e.g. for top-level await etc), while using CommonJS is still synchronous - not that all code is now loaded asynchronously.