Easy cluster management + zero downtime deploys & reloads


Basic cluster management in one line:

return if require('cluster_master').runMaster()
# worker's code here 

By default it'll spawn os.cpus().length workers and respawn every dead worker.

Out of box you'll get safe zero downtime reload:

  • send SIGHUP to process
  • cluster will try to spawn new worker
  • if new worker is ok then cluster will gracefully close one running worker & 'll reload all old workers
  • if new forker failed it won't destroy old workers

For more stuff see source.

Oops! Seems like forever still does not have proper support for specifying kill signal.

However you can specify pid file for ClusterMaster and send hups yourself:

return if require('cluster_master').runMaster(pidfile: 'pids/')

May be later: Once cluster is started with forever

forever start -c coffee

it can be reloaded with

forever stop --killSignal SIGHUP -c coffee