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:
- 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/cluster.pid')
May be later: Once cluster is started with forever
forever start -c coffee server.coffee
it can be reloaded with
forever stop --killSignal SIGHUP -c coffee server.coffee