teamster v1.3.0

teamster

The twelve-factor app's processes are disposable, meaning they can be started or stopped at a moment's notice. This facilitates fast elastic scaling, rapid deployment of code or config changes, and robustness of production deploys.

Processes shut down gracefully when they receive a SIGTERM signal from the process manager. For a web process, graceful shutdown is achieved by ceasing to listen on the service port (thereby refusing any new requests), allowing any current requests to finish, and then exiting.

— The Twelve-Factor App

Teamster is a Twelve-Factor-compliant, Unix-y worker process manager for Node. Its primary use is to facilitate the painless running and graceful shutdown of HTTP servers, but it has many other potential use cases.

Clusters running with teamster listen for the Unix signal SIGTERM, and then attempt to shut down their worker processes gracefully. This is useful in both single-worker and multiple-worker situations, as in both cases it's desirable for an HTTP server to finish serving any requests in progress before exiting.

When a teamster master process receives the SIGTERM signal, it tells all of the worker processes to stop accepting new connections, serve their requests already in progress, and then exit.

Install

npm install teamster --save

Usage

Run a function

Teamster can run a single function for you in a worker or workers. Simply pass #run a function as the first argument.

require('teamster').run(function work() {
  console.log('I am a worker!');
}, options);

Run an HTTP server

Although you could easily do this with #run, teamster can also run an HTTP server for you with #runServer. The first argument to #runServer should be a function with the standard Node request handler signature, which includes Express apps.

require('teamster').runServer(function handleRequest(req, res) {
  res.end('I am served from a worker!');
}, options);

Note that #runServer will either run on a port specified in the optional options object argument, or any available port.

Options

Both #run and #runServer accept an optional second options argument.

Signals

Teamster works by responding to Unix signals. Typically, you'll only want to send signals to teamster's master process, and it will forward signals onto the worker processes as appropriate.

Caveat

Node is very fast, and it's unlikely that you need to be run a process for each CPU. It's more likely you'll have unused workers and unnecessarily high memory usage, unless you've done testing and are sure you'll benefit from running more than a single worker process. Even with a single worker, however, teamster is useful, as it will take care of graceful worker shutdowns for you.

Thanks, Heroku

While I created and maintain this project, it was done while I was an employee of Heroku on the Human Interfaces Team, and they were kind enough to allow me to open source the work. Heroku is awesome.