All requests are being handled by a single worker

[ghost]

I have uWS cluster setup with worker threads as given in example section. My code is as follows

const createWorker = () => {
    let w = new Worker(__filename);
    w.on('online', () => {
        log.info('Worker is online');
    });
    w.on('exit', (code) => {
        log.info('Worker died with code : ' + code );
        new Worker(__filename);
    });
    w.on('error', (e) => {
        log.error('Worker Error : ', e);
        new Worker(__filename);
    });
    w.on( 'disconnect' , (worker) => {
        log.info('Worker with id ' + worker.id + ' got disconnected');
    });
}
if (isMainThread) {
    log.info('Main Thread');
    cpus()
    .forEach(() => {
        log.info('Forking')
        createWorker();
    });
} else {
    log.info('Starting Server')
    app.listen(port, (token) => {
        log.info('Server Started ', port);
    });
    process.on('SIGTERM', () => {
        log.info('SIGTERM');
    });
    process.on('SIGINT', () => {
        log.info('SIGINT');
    });
}

I generate a random unique id for every worker and I have my logger configured to add instance id to every log that it prints

[2020-05-31T20:05:50.764] [56c10bf2] [INFO] cluster.js - Main Thread
[2020-05-31T20:05:50.764] [56c10bf2] [INFO] cluster.js - Forking
[2020-05-31T20:05:50.767] [56c10bf2] [INFO] cluster.js - Forking
[2020-05-31T20:05:50.768] [56c10bf2] [INFO] cluster.js - Forking
[2020-05-31T20:05:50.769] [56c10bf2] [INFO] cluster.js - Forking
[2020-05-31T20:05:50.848] [56c10bf2] [INFO] cluster.js - Worker is online
[2020-05-31T20:05:50.855] [56c10bf2] [INFO] cluster.js - Worker is online
[2020-05-31T20:05:50.865] [56c10bf2] [INFO] cluster.js - Worker is online
[2020-05-31T20:05:50.895] [56c10bf2] [INFO] cluster.js - Worker is online
[2020-05-31T20:05:51.939] [10884b91] [INFO] cluster.js - Starting Server
[2020-05-31T20:05:51.940] [10884b91] [INFO] cluster.js - Server Started  8082
[2020-05-31T20:05:51.918] [c947bd34] [INFO] cluster.js - Starting Server
[2020-05-31T20:05:51.918] [c947bd34] [INFO] cluster.js - Server Started  8082
[2020-05-31T20:05:51.976] [53fc459e] [INFO] cluster.js - Starting Server
[2020-05-31T20:05:51.976] [53fc459e] [INFO] cluster.js - Server Started  8082
[2020-05-31T20:05:51.985] [fbe94ad3] [INFO] cluster.js - Starting Server
[2020-05-31T20:05:51.985] [fbe94ad3] [INFO] cluster.js - Server Started  8082

When I just trigger some requests concurrently using load test all the requests end up being served by a single worker.

[2020-05-31T20:06:03.957] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:03.965] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:03.967] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:03.968] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:04.035] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:04.037] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:04.040] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:04.121] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:04.122] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:04.123] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:04.166] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:04.168] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:04.187] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:04.189] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:04.206] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:04.208] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:04.230] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:04.231] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:04.259] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:04.261] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:04.270] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:04.298] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:04.301] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:04.311] [c947bd34] [INFO] middleware.js - middleware triggered
[2020-05-31T20:06:04.312] [c947bd34] [INFO] middleware.js - middleware triggered
.....

Just truncated the logs for few requests and I verified that all the 10000 requests that I triggered were served by a single worker. I hope its not desired behaviour.

[zaksmok]

What tool did you use to test? Mostly it's issue with sticking keep-alive session. When you use multiple clients everything will be fine.

[ghost]

@zaksmok i just found the issue. Its already been raised here before. I am running it in mac machine and it requires different port for every worker. So only one worker gets successfully created out of four and eventually handles all the request. I hope it would not occur in linux machine which i have not tried yet

[zaksmok]

I tested on WSL (Debian 10), Debian 10 and Ubuntu 20. Everthing worked fine on these machines so it can be osx issue

[ghost]

It is a Linux feature

[ghost]

Yup got it @alexhultman-2fa i just got confused with nodejs cluster. It works with same port for all the workers in mac. I think it just creates a master listening on port and transfers the request to worker using round robin methodology. Here we create four independent worker threads which listen to port directly

[ghost]

Since mac is just my development machine and my server is linux it should not be a problem and i closed the issue for the very reason