# Bad perfomance with nginx and php-fpm **URL:** \ **Category:** NGINX\ **Created:** [November 17, 2011, 6:28pm UTC](https://www.ruby-forum.com/t/bad-perfomance-with-nginx-and-php-fpm/213972 "2011-11-17T18:28:53Z")\ **Posts on this page:** 7\ **Page:** 1
**Author:** ![luislavena](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/l/0ea827/32.png) [@luislavena](https://www.ruby-forum.com/u/luislavena)\ **Post date:** [November 17, 2011, 6:28pm UTC](https://www.ruby-forum.com/t/bad-perfomance-with-nginx-and-php-fpm/213972/1 "2011-11-17T18:28:53Z")
Hello, i’m trying to install nginx with php-fpm. It works, but it’s not really fast and i have problems with multiple connections. Versions: nginx: nginx version: nginx/1.0.10 PHP 5.3.8-1~dotdeb.2 (fpm-fcgi) (built: Aug 25 2011 13:36:54) Configs: VHost: […] location ~ .php$ { fastcgi\_index index.php; include /etc/nginx/fastcgi\_params; fastcgi\_param SCRIPT\_FILENAME /web$fastcgi\_script\_name; fastcgi\_pass unix:/tmp/foo.socket; } […] php5-fpm: […] listen = ‘/tmp/foo.socket’ user = foo group = foo pm = dynamic pm.max\_children = 10 pm.start\_servers = 2 pm.min\_spare\_servers = 1 pm.max\_spare\_servers = 3 chroot = /var/www/foo/ chdir = /web/ […] fastcgi\_params: […] fastcgi\_connect\_timeout 60; fastcgi\_send\_timeout 180; fastcgi\_read\_timeout 180; fastcgi\_buffer\_size 128k; fastcgi\_buffers 4 256k; fastcgi\_busy\_buffers\_size 256k; fastcgi\_temp\_file\_write\_size 256k; fastcgi\_intercept\_errors on; […] I installed a fresh Wordpress without any plugins on the page and tried ApacheBench with -n 1000 -c 10 on it: Time taken for tests: 25.860 seconds Complete requests: 1000 Failed requests: 875 (Connect: 0, Receive: 0, Length: 875, Exceptions: 0) The same page on a standard apache setup spawns 10x php5-cgi: Time taken for tests: 104.929 seconds Complete requests: 1000 Failed requests: 0 Not sure what’s wrong ☹ Thanks Dieter ---
**Author:** ![komi](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/k/4bbf92/32.png) [@komi](https://www.ruby-forum.com/u/komi)\ **Post date:** [November 17, 2011, 6:47pm UTC](https://www.ruby-forum.com/t/bad-perfomance-with-nginx-and-php-fpm/213972/2 "2011-11-17T18:47:27Z")
How about the logs file? Regards, Joe On Thu, Nov 17, 2011 at 11:50 PM, Dieter K. ---
**Author:** ![komi](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/k/4bbf92/32.png) [@komi](https://www.ruby-forum.com/u/komi)\ **Post date:** [November 17, 2011, 6:54pm UTC](https://www.ruby-forum.com/t/bad-perfomance-with-nginx-and-php-fpm/213972/3 "2011-11-17T18:54:52Z")
2011/11/17 Dieter K. [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid) > Configs: > php5-fpm: > […] > listen = ‘/tmp/foo.socket’ > user = foo > group = foo > pm = dynamic can you test by setting pm = static and pm.max\_children to something a little bit hight than 10 (12 or 15). Just to ensure the problem does not come from the dynamic PM. ---
**Author:** ![komi](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/k/4bbf92/32.png) [@komi](https://www.ruby-forum.com/u/komi)\ **Post date:** [November 17, 2011, 11:02pm UTC](https://www.ruby-forum.com/t/bad-perfomance-with-nginx-and-php-fpm/213972/4 "2011-11-17T23:02:51Z")
2011/11/17 Jrme L. [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid): > can you test by setting pm = static and pm.max\_children to something a > little bit hight than 10 (12 or 15). > Just to ensure the problem does not come from the dynamic PM. Sure. I just tested it with 20 children: Time taken for tests: 24.580 seconds Complete requests: 1000 Failed requests: 859 (Connect: 0, Receive: 0, Length: 859, Exceptions: 0) The same result ☹ This sould be far better with 20 php daemons, it must be another problem, like a timeout or something like that? Thanks Dieter ---
**Author:** ![komi](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/k/4bbf92/32.png) [@komi](https://www.ruby-forum.com/u/komi)\ **Post date:** [November 17, 2011, 10:59pm UTC](https://www.ruby-forum.com/t/bad-perfomance-with-nginx-and-php-fpm/213972/5 "2011-11-17T22:59:47Z")
2011/11/17 Joe [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid): > How about the logs file? I tried a new benchmark with the old settings: Time taken for tests: 24.939 seconds Complete requests: 1000 Failed requests: 861 (Connect: 0, Receive: 0, Length: 861, Exceptions: 0) In the logfile from php5-fpm i see 1000 requests (everyone with status 200): […] - - 17/Nov/2011:21:55:40 +0000GET /index.php200 /www/index.php 174.837 20224 102.96% - - 17/Nov/2011:21:55:40 +0000GET /index.php200 /www/index.php 175.963 20224 96.61% - - 17/Nov/2011:21:55:40 +0000GET /index.php200 /www/index.php 191.407 20224 83.59% - - 17/Nov/2011:21:55:40 +0000GET /index.php200 /www/index.php 212.139 20224 75.42% […] In the nginx-access.log the same: […] 46.252.24.80 - - [17/Nov/2011:22:55:37 +0100] “GET / HTTP/1.0” 200 5935 “-” “ApacheBench/2.3” 46.252.24.80 - - [17/Nov/2011:22:55:37 +0100] “GET / HTTP/1.0” 200 5931 “-” “ApacheBench/2.3” 46.252.24.80 - - [17/Nov/2011:22:55:37 +0100] “GET / HTTP/1.0” 200 5933 “-” “ApacheBench/2.3” 46.252.24.80 - - [17/Nov/2011:22:55:37 +0100] “GET / HTTP/1.0” 200 5931 “-” “ApacheBench/2.3” […] Thanks Dieter ---
**Author:** ![komi](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/k/4bbf92/32.png) [@komi](https://www.ruby-forum.com/u/komi)\ **Post date:** [November 18, 2011, 4:14am UTC](https://www.ruby-forum.com/t/bad-perfomance-with-nginx-and-php-fpm/213972/6 "2011-11-18T04:14:29Z")
On Fri, Nov 18, 2011 at 5:02 AM, Dieter K. [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid) wrote: > The same result ☹ > > This sould be far better with 20 php daemons, it must be another > problem, like a timeout or something like that? Here’s my result on my own site: [root@eustia ~]# ab -c 10 -n 1000 [http://animebsd.net/](http://animebsd.net/) This is ApacheBench, Version 2.3 \<$Revision: 655654 $\> Copyright 1996 Adam Twiss, Zeus Technology Ltd, [http://www.zeustech.net/](http://www.zeustech.net/) Licensed to The Apache Software Foundation, [http://www.apache.org/](http://www.apache.org/) Benchmarking [animebsd.net](http://animebsd.net) (be patient) Completed 100 requests Completed 200 requests Completed 300 requests Completed 400 requests Completed 500 requests Completed 600 requests Completed 700 requests Completed 800 requests Completed 900 requests Completed 1000 requests Finished 1000 requests Server Software: nginx/1.1.8 Server Hostname: [animebsd.net](http://animebsd.net) Server Port: 80 Document Path: / Document Length: 54635 bytes Concurrency Level: 10 Time taken for tests: 261.211 seconds Complete requests: 1000 Failed requests: 0 Write errors: 0 Total transferred: 54893000 bytes HTML transferred: 54635000 bytes Requests per second: 3.83 [#/sec] (mean) Time per request: 2612.112 [ms] (mean) Time per request: 261.211 [ms] (mean, across all concurrent requests) Transfer rate: 205.22 [Kbytes/sec] received Connection Times (ms) min mean[+/-sd] median max Connect: 54 61 134.0 54 3053 Processing: 1260 2545 304.0 2515 3884 Waiting: 456 1534 219.6 1510 2673 Total: 1315 2606 328.8 2570 5409 Percentage of the requests served within a certain time (ms) 50% 2570 66% 2605 75% 2652 80% 2690 90% 2978 95% 3187 98% 3270 99% 3902 100% 5409 (longest request) Not exactly fast (cpu constrained - kvm instance of 1 core/thread e3-1270 cpu, 512MB ram, debian linux 6) but certainly completed all requests without fail. 15 php5-cgi instances ran through supervisord in tcp mode. There are quite a bit plugins installed, too (but no wp-cache of some kind). No tweaking at nginx side, some timeout and limit increase on php side. – O\< ascii ribbon campaign - stop html mail - [www.asciiribbon.org](http://www.asciiribbon.org) ---
**Author:** ![komi](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/k/4bbf92/32.png) [@komi](https://www.ruby-forum.com/u/komi)\ **Post date:** [November 18, 2011, 8:33am UTC](https://www.ruby-forum.com/t/bad-perfomance-with-nginx-and-php-fpm/213972/7 "2011-11-18T08:33:50Z")
2011/11/17 Dieter K. [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid): > The same result ☹ > > This sould be far better with 20 php daemons, it must be another > problem, like a timeout or something like that? what page are you requesting ? Start by requesting a very simple PHP page (\). It’s a good start to figure out what’s going on. Here is what I have with a very short and simple page (20 static children on FPM PHP 5.3 trunk and nginx 1.1.4) fat@dev:~/web$ ab -c 10 -n 1000 [http://xxxxxxxxxx/pwd.php](http://xxxxxxxxxx/pwd.php) Server Software: nginx/1.1.4 Server Hostname: xxxxxx Server Port: 80 Document Path: /pwd.php Document Length: 22 bytes Concurrency Level: 10 Time taken for tests: 0.649 seconds Complete requests: 1000 Failed requests: 0 Write errors: 0 Total transferred: 175000 bytes HTML transferred: 22000 bytes Requests per second: 1539.96 [#/sec] (mean) Time per request: 6.494 [ms] (mean) Time per request: 0.649 [ms] (mean, across all concurrent requests) Transfer rate: 263.18 [Kbytes/sec] received Connection Times (ms) min mean[+/-sd] median max Connect: 0 0 0.4 0 5 Processing: 0 6 2.6 6 37 Waiting: 0 6 2.6 6 37 Total: 1 6 2.9 6 42 Percentage of the requests served within a certain time (ms) 50% 6 66% 6 75% 7 80% 7 90% 7 95% 7 98% 12 99% 19 100% 42 (longest request) Then increase the size the PHP page returns (ex: 64k → \) fat@dev:~/web$ ab -c 10 -n 1000 [http://xxxxxxx/size.php](http://xxxxxxx/size.php) Server Software: nginx/1.1.4 Server Hostname: xxxxxxx Server Port: 80 Document Path: /size.php Document Length: 65536 bytes Concurrency Level: 10 Time taken for tests: 1.242 seconds Complete requests: 1000 Failed requests: 0 Write errors: 0 Total transferred: 65689000 bytes HTML transferred: 65536000 bytes Requests per second: 805.17 [#/sec] (mean) Time per request: 12.420 [ms] (mean) Time per request: 1.242 [ms] (mean, across all concurrent requests) Transfer rate: 51651.05 [Kbytes/sec] received Connection Times (ms) min mean[+/-sd] median max Connect: 0 0 0.4 0 5 Processing: 1 12 2.9 12 34 Waiting: 1 12 2.9 12 32 Total: 1 12 3.0 12 37 Percentage of the requests served within a certain time (ms) 50% 12 66% 13 75% 14 80% 14 90% 16 95% 17 98% 18 99% 20 100% 37 (longest request) Then increase the complexity (size, CPU, memory, fork, database requests, …)