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

Read more here.

• Node 6.x
• Node 8.x
• Node 10.x
• Node 12.x
• Node 14.x

npm i master-process --save

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

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

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

The master-process module uses the cluster module to run the user application in cluster mode. There are two types of processes involved in a Node cluster:

• master process,
• worker processes.

The worker processes are used to run your application. All worker processes in a cluster will serve requests on a single server port or UNIX domain socket (see cluster documentation for how this is achieved).

The master process handles forking the required number of workers as well as:

• handling the SIGHUP signal on the master process to reload the cluster (new workers are created and old workers are destroyed once the new ones are ready to service requests). Use this signal to tell the master process that you have updated the application and it should reload it.

• handling the SIGTERM signal to cleanly shut down all workers and exit the cluster.

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

WORKERS=MAX sets the number of workers equals to the number of cores (as returned by os.cpus().length) WORKERS=AUTO sets the number of workers equals to the number of cores - 1 (or a single worker if single core)

If a worker exits unexpectedly, master-process will attempt to replace it with a new worker. Similarly if the worker crashes or is killed by the operating system it will also be replaced.

To avoid avoid excessive resource usage in case newly-started workers keep crashing there is a WORKER_THROTTLE environment variable that is used to throttle how often a given worker is restarted:

• if a worker has been running for less than WORKER_THROTTLE when it crashes there will be a delay before a replacement worker is created.

• if a worker has been running for longer than WORKER_THROTTLE then the replacement worker is started immediately.

The default value is WORKER_THROTTLE=1s.

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. Here are the environment variables that can be used to control the process monitoring and their respective defaults:

MEM_MONITOR_FAILURES=10
CPU_MONITOR_FAILURE=10
MAX_MEMORY_ALLOWED_MB=1200
MAX_CPU_ALLOWED=95

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