Multiple upstream faye websocket servers with nginx - nginx

Question

I would like to cluster my faye websocket server. A single server works well but I want to be prepared to scale.
My first attempt is to start a few thin servers on different sockets and then add them to the upstream for my server in nginx.
The bayeux messages are distributed amongst the clusters but chrome dev tools shows about 16
websocket 101 connections with connection closed in the frame tab of the networking panel:
Connection Close Frame (Opcode 8)
Connection Close Frame (Opcode 8, mask)
And a whole bunch of /meta/connect and /meta/handshake on the server side for all of the faye instances.
An excerpt:
D, [2013-11-15T15:34:50.215631 #5344] DEBUG -- : {"channel"=>"/meta/connect", "clientId"=>"q7odwfbovudiw87dg0jke3xbrg51tui", "connectionType"=>"callback-polling", "id"=>"p"}
D, [2013-11-15T15:34:50.245012 #5344] DEBUG -- : {"channel"=>"/meta/connect", "clientId"=>"ckowb5vz9pnbh7jwomc8h0qsk8t0nus", "connectionType"=>"callback-polling", "id"=>"r"}
D, [2013-11-15T15:34:50.285460 #5344] DEBUG -- : {"channel"=>"/meta/handshake", "version"=>"1.0", "supportedConnectionTypes"=>["callback-polling"], "id"=>"u"}
D, [2013-11-15T15:34:50.312919 #5344] DEBUG -- : {"channel"=>"/meta/handshake", "version"=>"1.0", "supportedConnectionTypes"=>["callback-polling"], "id"=>"w"}
D, [2013-11-15T15:34:50.356219 #5344] DEBUG -- : {"channel"=>"/meta/handshake", "version"=>"1.0", "supportedConnectionTypes"=>["callback-polling"], "id"=>"y"}
D, [2013-11-15T15:34:50.394820 #5344] DEBUG -- : {"channel"=>"/meta/handshake", "version"=>"1.0", "supportedConnectionTypes"=>["callback-polling"], "id"=>"10"}
Starting the thin servers (each in its own terminal):
thin start -C config/thin/development.yml -S /tmp/faye.1.sock
thin start -C config/thin/development.yml -S /tmp/faye.2.sock
thin start -C config/thin/development.yml -S /tmp/faye.3.sock
My thin config:
---
environment: development
timeout: 30
log: log/thin.log
pid: tmp/pids/thin.pid
max_conns: 1024
max_persistent_conns: 100
require: []
wait: 30
# socket: /tmp/faye.sock
daemonize: false
rackup: config.ru
My NGINX config:
upstream thin_cluster {
server unix:/tmp/faye.1.sock fail_timeout=0;
server unix:/tmp/faye.2.sock fail_timeout=0;
server unix:/tmp/faye.3.sock fail_timeout=0;
}
server {
# listen 443;
server_name ~^push\.mysite\.dev(\..*\.xip\.io)?$;
charset UTF-8;
tcp_nodelay on;
# ssl on;
# ssl_certificate /var/www/heypresto/certificates/cert.pem;
# ssl_certificate_key /var/www/heypresto/certificates/key.pem;
# ssl_protocols TLSv1 SSLv3;
location / {
proxy_pass http://thin_cluster;
proxy_redirect off;
proxy_http_version 1.1;
proxy_set_header Upgrade $http_upgrade;
proxy_set_header Connection "upgrade";
break;
}
}
Thought it was too good to be true to put more upstreams for websockets and it would just work, but it seems so close...
EDIT: I've realised that this is probably not going to work and I should probably create N servers (push1.mysite.dev, push2.mysite.dev etc) and have the backend tell the frontend which one to connect.
Still If there are any thoughts out there...