NGinx Load Balancer / Apache Backend / HTTP 499

Hi All,

I have NGinx configured as a Load Balancer with Apache servers at the backend serving a PHP application. We use 'Pingdom.com' for external application monitoring. I noticed several "outages" rarely lasting more than the minute between checks.

I added some additional logging as detailed here: https://www.nginx.com/resources/admin-guide/logging-and-monitoring

--nginx access log
108.62.115.226 - - [13/Nov/2015:20:11:03 +0000] "GET //login/index.php HTTP/1.1" 499 0 "-" "Pingdom.com_bot_version_1.4_(http://www.pingdom.com/)"rt=30.031 uht="-" urt="-"

On checking each of the Apache backend I found:
108.62.115.226 - - [13/Nov/2015:20:10:33 +0000] "GET //login/index.php HTTP/1.0" 200 6968

The interesting part here is that the Apache backend log entry is 30 seconds *before* the NGinx log entry. I guess the order of the process is
- Nginx passes request to backend but doesn't log it
- Apache deals with it and responds
- However NGinx doesn't receive it
- Pingdom closes the connection which results in the HTTP 499 which is logged in the NGinx log.

The clocks are in sync via NTP and I've manually verified they're not out of time with a `date` on the CLI.

What I'd like to gain some insight into is why NGinx doesn't appear to receive the response. It appears that Apache is responding fine and many of the help articles I've read seems to indicate that the backend is usally at fault but this doesn't appear to be the case here.

Any help would be greatly appreciated.

Thanks!