Socket.IO and Node.js Core Cluster - node.js

Question

Is it possible to use Socket.IO with Node's core cluster (not the outdated module)?
I can fork multiple workers and it seems to work fine; however, when opening a connection I get the error: solve: warn - client not handshaken client should reconnect
Here's the relevant code snippet (with a few simple things like expressjs config removed):
if ( cluster.isMaster ) {
for ( var i = 0; i < numCPUs; i++ ) {
cluster.fork();
}
} else {
app.get('/', function (req, res) {
res.sendfile( __dirname + '/public/html/index.html' );
});
io.configure( function() {
var RedisStore = require('socket.io').RedisStore,
opts = { host: 'localhost', port: 8888 };
io.set('store', new RedisStore( { redisPub: opts, redisSub: opts, redisClient: opts } ));
});
app.listen( 8888 );
io.sockets.on('connection', function (socket) {
socket.emit( 'some', 'data' );
});
}
I've tried with and without using RedisStore and with the trick on this site (which I believe is obsolete now): http://www.danielbaulig.de/socket-ioexpress/
I've also looked at the code at http://www.ranu.com.ar/2011/11/redisstore-and-rooms-with-socketio.html, although I don't see how that code is any different than using MemoryStore.
All of my test connections are using Websockets (RFC 6455). This works fine if I set numCPUs to equal 1.
Node.js version 0.6.17
Socket.io version 0.9.5
Expressjs version 2.5.9
Update - include console output (note, on this attempt the connection did ultimately work, although it threw the same errors):
info - socket.io started
info - socket.io started
info - socket.io started
info - socket.io started
info - socket.io started
debug - served static content /socket.io.js
debug - client authorized
info - handshake authorized 17644195072103946664
debug - setting request GET /socket.io/1/websocket/17644195072103946664
debug - set heartbeat interval for client 17644195072103946664
debug - websocket writing 7:::1+0
warn - client not handshaken client should reconnect
info - transport end (error)
debug - set close timeout for client 17644195072103946664
debug - cleared close timeout for client 17644195072103946664
debug - cleared heartbeat interval for client 17644195072103946664
debug - discarding transport
debug - client authorized
info - handshake authorized 16098526291524652257
debug - setting request GET /socket.io/1/websocket/16098526291524652257
debug - set heartbeat interval for client 16098526291524652257
debug - websocket writing 7:::1+0
warn - client not handshaken client should reconnect
info - transport end (error)
debug - set close timeout for client 16098526291524652257
debug - cleared close timeout for client 16098526291524652257
debug - cleared heartbeat interval for client 16098526291524652257
debug - discarding transport
debug - client authorized
info - handshake authorized 13419993801561067603
debug - setting request GET /socket.io/1/websocket/13419993801561067603
debug - set heartbeat interval for client 13419993801561067603
debug - client authorized for
debug - websocket writing 1::
debug - websocket writing 5:::{"some":"data","args":[11354]}
debug - websocket writing 5:::{"some":"data","args":[36448]}
This is how the console output ends on a failure (fails about 9 times out of 10):
info - transport end by forced client disconnection
debug - websocket writing 0::
info - transport end (booted)
debug - set close timeout for client 1639301251431944437
debug - cleared close timeout for client 1639301251431944437
debug - cleared heartbeat interval for client 1639301251431944437
debug - discarding transport
debug - got disconnection packet
debug - got disconnection packet
Update - Added links to possible tickets on github:
https://github.com/LearnBoost/socket.io/issues/881
https://github.com/LearnBoost/socket.io/issues/438

Answer 1

It seems the issue may be related to the node module that Socket.IO comes installed with.
When I installed redis (npm install hiredis redis) and created the clients for the RedisStore using the redis module, everything suddenly worked perfectly. I've been running for over an hour with ~500 concurrent connections and haven't seen a single error, and every node process is being utilized.
hiredis#0.1.14
redis#0.7.2
Running Redis 2.6rc3 on port 6379.
Update: In Node.s 0.8, it looks like the cluster library should be considerably more mature, so the above code/issues will probably become obsolete: http://nodejs.org/docs/v0.7.8/api/cluster.html