Code is not being executed on one core even though it is online - pm2

Question

Right after the startup of the app, everything seems fine, all four cores are online running in the cluster mode. Two cores have the same name, remaining two differ. Here is the ecosystem.json file:
{ 'apps': [
{
'name': 'CHS',
'script': './main.js',
'out_file': '../logs/appname.log',
'error_file': '../logs/appname.log',
'log_date_format' : 'YYYY-MM-DD HH:mm:ss Z',
'rotateModule': true,
'compress': true,
'dateFormat': 'YYYY-MM-DD',
'max_size': '10M',
'retain': 300,
'log_date_format' : 'YYYY-MM-DD HH:mm:ss Z',
'merge_logs': true,
'env': {
'NODE_ENV': 'production',
'METEOR_SETTINGS': '{ ... }'
},
'env_production': {
'NODE_ENV': 'production'
},
'instances': '1',
'exec_mode': 'cluster'
},
{
'name': 'ORC',
'script': './main.js',
'out_file': '../logs/appname.log',
'error_file': '../logs/appname.log',
'log_date_format' : 'YYYY-MM-DD HH:mm:ss Z',
'rotateModule': true,
'compress': true,
'dateFormat': 'YYYY-MM-DD',
'max_size': '10M',
'retain': 300,
'merge_logs': true,
'env': {
'NODE_ENV': 'production',
'METEOR_SETTINGS': '{ ... }'
},
'env_production': {
'NODE_ENV': 'production'
},
'instances': '1',
'exec_mode': 'cluster'
},
{
'name': 'WCR',
'script': './main.js',
'out_file': '../logs/appname.log',
'error_file': '../logs/appname.log',
'log_date_format' : 'YYYY-MM-DD HH:mm:ss Z',
'rotateModule': true,
'compress': true,
'dateFormat': 'YYYY-MM-DD',
'max_size': '10M',
'retain': 300,
'merge_logs': true,
'env': {
'NODE_ENV': 'production',
'METEOR_SETTINGS': '{ ... }'
},
'instances': '2',
'exec_mode': 'cluster'
}
]
}
Few seconds after the startup, CHS core does not execute any code for some reason (without any errors), while other three works fine. I have tried to reorder cores in ecosystem.json, add "pmx": false (per some online suggestion, to avoid some parts in pm2 itself), rename cores, and some other things but issue was the same. I was trying to make sure there is nothing in the app code that makes this issue happens, searched errors in the logs, and this is only I have (in pm2 logs):
2020-04-17T14:16:49: PM2 error: (node:3720) [DEP0007] DeprecationWarning: worker.suicide is depricated. Please use worker.exitedAfterDisconnect.
2020-04-17T14:17:01: PM2 error: Error: write ENOTSUP
at ChildProcess.target._send (internal/child_process.js:692:20)
at ChildProcess.target.send (internal/child_process.js:576:19)
at senderHelper (internal/cluster/utils.js:25:15)
at send (internal/cluster/master.js:357:10)
at handle.add (internal/cluster/master.js:329:5)
at SharedHandle.add (internal/cluster/shared_handle.js:29:3)
at queryServer (internal/cluster/master.js:318:10)
at Worker.onmessage (internal/cluster/master.js:250:5)
at ChildProcess.onInternalMessage (internal/cluster/utils.js:42:8)
at emitTwo (events.js:131:20)
2020-04-17T14:17:01 PM2 log: App name:CHS id:0 disconnected
Also, at the end this happens only on the one windows VM instance (on the second one no issues at all). These instances are identical based on the configurations. PM2 version was 4.2.3 but then I have tried with 3.5.1 and I got the same issue. Does anybody have any idea how to troubleshoot this issue?
Lately it turned out that issue is related with UDP, and probably some unsupported features in node for windows, like this:
https://nodejs.org/api/cluster.html#cluster_event_listening

Answer 1

The root cause was that UDP does not work in a cluster mode on PM2. Actually we were using SNMP which relies on UDP under the hood, and it was not clear we have such an issue considering the logs. Spinning up a new core in a fork mode, running SNMP resolves the issue.