How to properly handle SIGUSR2 Signal sent by nodemon

[Kannaj]

Hello ,

When nodemon sends a sigusr2 signal to node , i'm handling it like the below

    process.on('SIGUSR2',function(){
        console.log('SIGUSR2 recieved')
        socketcluster.killWorkers();
        socketcluster.killBrokers();
        process.exit(1)
      })

root        50  0.0  0.0      0     0 ?        Z    07:58   0:00 [node] <defunct>
root        60  0.8  0.0      0     0 ?        Z    07:58   0:01 [node] <defunct>
root        65  0.7  0.0      0     0 ?        Z    07:58   0:01 [node] <defunct>
root        70  0.5  0.0      0     0 ?        Z    07:58   0:00 [node] <defunct>
root        75  3.0  0.0      0     0 ?        Z    07:58   0:05 [node] <defunct>
root        91  0.0  0.0      0     0 ?        Z    08:00   0:00 [node] <defunct>
root        96  1.9  0.0      0     0 ?        Z    08:00   0:01 [node] <defunct>
root       101  1.7  0.0      0     0 ?        Z    08:00   0:01 [node] <defunct>
root       106  1.2  0.0      0     0 ?        Z    08:00   0:00 [node] <defunct>
root       111  7.2  0.0      0     0 ?        Z    08:00   0:04 [node] <defunct>

The problem is that i dont think the worker/broker processes are properly terminated before the master process exits. This is creating zombie processes. Is there a better way to handle the SIGUSR2?

[MegaGM]

Nodemon and the zombie processes. I faced the same problem some time ago.
After I examined code relative to the USERSIG2

socketcluster/index.js

Line 466 in bd70b52

if (this.options.rebootOnSignal) {

I decided to set up the rebootOnSignal: false option. Now there are no zombie processes anymore.
I consider this way as a workaround, since nodemon reloads all the processes, including master, but on my PC it takes about 500ms so I just don't care.

Though I'm okay with the workaround, I'd like to know "the right way" of doing things anyway. @Someone tell us please :3

[happilymarrieddad]

I'm not sure if this helps, but I have an admin channel that all my servers listen to and when they receive a "reboot" command it calls the killworkers and killbrokers functions. I think the problem you are facing is those are functions and so they run async and the process.exit() is called right after which means the killing doesn't finish. I'm using redis and the brokers to handle it. I hope that makes sense. Let me know if you need more clarification and/or if this is a solution you are interested in. We have quite a large cluster and it works great for us. Thanks!