Nginx connection reset

(its probably worth mentioning that I am not an nginx user, or indeed rails, so this is just initial guesses to maybe kick off the thread with some ideas for an answer)

From the details of your log entry, it seems that the external request is being internally forwarded by nginx on server with Host string domain.com" onto a local haproxy running on localhost:82 ?

If that is the case, then really I would look to correlate the log entries from nginx to haproxy, ie find that same request in haproxy log.

Given that I dont know about nginx so I am guessing, I think you need to determine whether this 110 message corresponds to a proxy_connect_timeout or a proxy_read_timeout, the former means that nginx didn't get any response from haproxy (host A sends SYN, your localhost:82 dropped the packet), and the latter that it connected but didn't send any data back (Syn-Syn-ack ack, but no data on the stream).

If the latter, its likely the problem is further back in your web stack, and you should look for the same log entry in memcache or mysql logs.

For example set your slow query log my.conf configuration on mysql and look and see if there is an entry corresponding to your request in that log file. I think my default is in /var/lib/mysql/slow.log but I guess might be some customization.

More generally, in these platforms where you have created a fairly complicated system, it is helpful to have some centralized logging infrastructure for dealing with event correlation. I am currently deploying logstash, for such purposes and there is obviously splunk and logblaze for commercial alternatives.

You should first determine, which layer this fault is occurring in, i.e. at the server level (Linux machine configuration), at PHP (php.ini, php-cgi) level, or at your application level. HTTP errors can be caused by incorrect configuration at any of these layers, so you need to first isolate and identify where exactly the server fault (pun intended) is occurring.

My suggestion is you trace the HTTP request:

1. From the moment it arrives at your server, through the Linux configuration.
2. Then through Nginx configuration (nginx.conf, default.conf).
3. Then through PHP and PHP-CGI.
4. Lastly analyze it once it reaches your application and check what your application does with it for sending the http response.

In this manner you will be able to identify exactly where the problem is, before the good folks at server fault will be able to help you resolve it correctly!

Also post your nginx.conf, default.conf, php-fpm.conf Some detail about your application would also be nice.