Production Use

[ghost]

Thank you the research.

I see a small typo bug.
It should be
var worker = workers[worker_index(connection.remoteAddress, num_processes)];

The scripts works well for me.

[elad]

Thanks! Fixed.

[ghost]

I want to use this code in production.
Do you have any recommendation/addition?

[elad]

Yes, there's a caveat, see joyent/node issue #7784. I would wait a bit before using it in production, unless the caveat is insignificant for your scenario.

In some cases the master process seems to not get the connection event causeing the client to hang. If the client retries, it usually succeeds. However this is still unreliable.

If this won't be fixed in the next couple of days I will rewrite the example in this repository using a different approach.

Here's what I have in mind:

• Start multiple node.js instances on different ports
• Use nginx upstream with ip_hash as suggested in http://socket.io/docs/using-multiple-nodes/
• Use cluster for respawning
• Add an example of communicating between the node.js processes when broadcasting messages