什么是nginx的$ request_time？为什么它比php-fpm执行时间多得多？

Question

Background : It is the objective of my project is to create a very fast responding website. There are only very few pages, many of them are simply 302 redirects.

In Nginx , I'm logging here $request_time .

In PHP , I'm logging the microtime of the start of the request and just before it exits.

php-fpm (php 5.3.27) and nginx (1.4.4) run on the same machine, there is no database (just writing to a Beanstalkd queue), there are no complicated nginx configurations.

Problem : There is a massive discrepancy between the PHP execution time and Nginx's $request_time. The $request_time is generally 0.5 seconds but during some hours it's more like 3 seconds on average. PHP's execution time is always between 0.008 seconds and 0.02 seconds (using PhalconPHP ).

Question : Why could there be such a big discrepancy? Perhaps I don't fully understand what $request_time is, or maybe my webservers have some problematic configuration? I'd be happy to provide more information about the environment.

Answer 1

$request_time is the time from the first byte sent, to the time everything is closed up and logging has been completed. If you look at your nginx logs and the microtime you are logging, how close are they in terms of starting times? For the ones where the request_time was closer to 3 seconds, and the php execution time was say 0.02 seconds, if you extrapolate the request start time and compare that to the microtime inside php, are they close, or did nginx need to wait a second or 2 (perhaps for a php process to free up, etc..). It would probably be interesting to also log the $upstream_response_time to see how that compares.