Nginx server configuration for Thin, Faye & Redis - ruby-on-rails

Question

/etc/nginx/nginx.conf looks like:
user deploy;
worker_processes 5;
error_log logs/error.log;
events {
worker_connections 1024;
use epoll;
}
http {
include mime.types;
default_type application/octet-stream;
sendfile on;
keepalive_timeout 65;
upstream foreman4000 {
server x.x.x.x:4000;
server x.x.x.x:4001;
server x.x.x.x:4002;
server x.x.x.x:4003;
server x.x.x.x:4004;
}
server {
listen 80;
server_name x.x.x.x; #server IP
access_log /opt/nginx/foreman4000.access.log;
location / {
proxy_pass http://foreman4000;
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $remote_addr;
}
}
}
Here I use gem foreman, which uses upstart to manage all process and start all servers with one command
I created Procfile in the main directory of the project which contains:
redis: redis-server
thin: bundle exec thin start -p $PORT
faye: rackup faye.ru -E production -s thin
Added to Gemfile:
gem 'foreman'
gem 'thin'
gem "foreman-export-daemontools", "~> 0.0.1"
Ran bundle install locally to edit Gemfile.lock
Deployed project on the server.
Started Nginx
deploy#dcards101:/opt/nginx/conf$ sudo /etc/init.d/nginx stop [ OK ]
deploy#dcards101:/opt/nginx/conf$ sudo /etc/init.d/nginx srart [ OK ]
Exported data from Procfile to Upstart
deploy#dcards101:/var/www/cards/current$ rvmsudo foreman export upstart -a cards -u root
Started application
deploy#dcards101:/var/www/cards/current$ rvmsudo start cards
Now everything had to be good but what i see on the server is only
502 Bad Gateway
nginx/1.0.15
Logs say:
2012/07/17 17:22:30 [error] 11593#0: *148 no live upstreams while connecting to upstream, client: x.x.x.x, server: x.x.x.x, request: "GET / HTTP/1.1", upstream: "http://foreman4000/", host: "x.x.x.x"
Please help with anything you can. Server -- Ubuntu 10 LTS.

Answer 1

got the same error solved it this way:
first install nginx_tcp_proxy_module
( I followed this tutorial but changed it to use passenger and thin with nginx)
than add the tcp part to your nginx.conf:
tcp {
upstream websockets {
## node processes
server 12.34.56.78:9292;
check interval=300 rise=2 fall=5 timeout=1000;
}
server {
listen 9200;
server_name domain.org;
tcp_nodelay on;
proxy_pass websockets;
}
}
doesn´t work on port 80 for me
after that I still get empty responses from faye/privat_pub but there was an extremly trivial solution:
RAILS_ENV=production bundle exec rackup private_pub.ru -s thin -E production
look private_pub - Issue #29
Now everything works except chrome how fires 2 times
(and I need an deamon-process for the rackup)
hope it helps you too

Answer 2

I think your problem is that you put your app-server and the faye server in the same upstream!
If I get the method of upstream and foreman right, your first visitor get the app the second faye and so on. ( maybe I`m wrong because I don´t know foreman .. but if foreman shares all available servers to all services, that might be your problem )
I wood say try capistrano instead of foreman .. so you have full control which server starts where .. because at my my host http don`t work for private_pub (because of nginx) so I had to install the nginx_tcp_proxy_module to get the tcp block working in my nginx.conf
or just try server by server via ssh to find the error