Forum: NGINX request_time exstreamly larger than upstream_response_time for some queries when using fastcgi

Hi，
I run python in backend, and use nginx to proxy queries to python by 
using
fastcgi. I found some strange things in my nginx log those days. Some 
queries'
$request_time are up to 40 seconds which their $upstream_response_time 
are less
than 1 second.

Here is my fastcgi related conf.
fastcgi_max_temp_file_size 0;
fastcgi_param PATH_INFO $fastcgi_script_name;
fastcgi_param SCRIPT_NAME "";
fastcgi_pass python_backend;
fastcgi_buffers 1024 8K;

Do you guys know what's the problem?

Chao

Does anybody have a clue about this problem?


________________________________
 From: Tian Chao <tchao.china@yahoo.com>
To: "nginx@nginx.org" <nginx@nginx.org>
Sent: Wednesday, November 21, 2012 1:02 PM
Subject: request_time exstreamly larger than upstream_response_time for 
some queries when using fastcgi


Hi，
I run python in backend, and use nginx to proxy queries to python by 
using
fastcgi. I found some strange things in my nginx log those days. Some 
queries'
$request_time are up to 40 seconds which their $upstream_response_time 
are less
than 1 second.

Here is my fastcgi related conf.
fastcgi_max_temp_file_size 0;
fastcgi_param PATH_INFO $fastcgi_script_name;
fastcgi_param SCRIPT_NAME "";
fastcgi_pass python_backend;
fastcgi_buffers 1024 8K;

Do you guys know what's the problem?

Chao

>  I found some strange things in my nginx log those days. Some queries'
> $request_time are up to 40 seconds which their $upstream_response_time are
> less than 1 second.

$request_time is time since start of the request (first bytes read from
client) till end of the request (last bytes are sent to client and 
logging
happens) - so while the backend can generate the response in 1 second 
there
might be clients with slow connections or even not waiting for the 
response
(closing it) etc.


rr

Hi rr,

Thanks for your explain. Some queries' body sent is just about 1K bytes, 
however the$request_time is up to 40 seconds, so i think the connection 
is tooslow. Is there a way fordiagnosticwhat happened in that time for 
those slow queries?

Thanks,
Chao


________________________________
 From: Reinis Rozitis <r@roze.lv>
To: nginx@nginx.org
Sent: Thursday, November 22, 2012 12:04 PM
Subject: Re: request_time exstreamly larger than upstream_response_time 
for some queries when using fastcgi

> I found some strange things in my nginx log those days. Some queries' 
$request_time are up to 40 seconds which their $upstream_response_time are less 
than 1 second.

$request_time is time since start of the request (first bytes read from 
client) till end of the request (last bytes are sent to client and 
logging happens) - so while the backend can generate the response in 1 
second there might be clients with slow connections or even not waiting 
for the response (closing it) etc.


rr