pmalouin / master-process Goto Github PK

The purpose of this module is to reload a node.js application with no downtime by using the cluster capabilities.

Read more here.

npm i master-process --save

Use this code at the very beginning of your node.js application:

if (cluster.isMaster && 					//if is a master
    typeof v8debug !== 'object' &&			//not in debug mode
    process.env.NODE_ENV !== 'test') {      //not in test mode

  var mp = require('master-process');
  mp.init();
  return;
}

The application itself is run as a worker process.

The master process handles the special SIGHUP signal to create a new worker and once the new worker is listening it closes the old one.

Use this signal to tell the master process that you have updated the application and it should reload it.

The number of workers can be controlled with the WORKERS environment variable. The default is 1.

WORKERS=AUTO sets the number of workers equals to the number of cores.

The master-process does not handle application crashes and restarts. Once a worker crash the master process itself will crash, the service manager should take care of restarting the application.

If the master process detects that the version of the master-process module has changed it will quit with exit code 1. The service manager should take care of restarting the application.

The master process watch by default the behavior of the worker. If the process is taking too much resources it will load a new worker.

I use this special signal to profile the underlying application (check v8profiler). The master process pauses-resume the CPU/Mem monitoring and pass the signal to the worker.

If process.env.PORT starts with an / (slash) master-process will assume you are going to listen on a unix socket and it will take care of few things:

• cleaning the socket if exists on start up, otherwise the worker will fail with EADDRINUSE.
• cleaning the socket on exit.

Use DEBUG=master-process to debug this module.

Every worker receives these additional environment variables:

• PPID: The parent process id.
• RELOAD_INDEX: The number of times that the process has been reload with the SIGHUP signal.
• WORKER_INDEX: The index of the worker, useful when using more than one worker with WORKERS=AUTO or WORKERS=X.

• cluster-master.

MIT 2015 - Jose F. Romaniello