php-fpm: very high server load

Question

Since today my webserver (nginx + php-fpm + mysql on a VPS) is very slow.

htop says:

 1 [|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||100.0%] Tasks: 63 total, 13 running 2 [|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||100.0%] Load average: 11.67 10.95 6.95 3 [|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||100.0%] Uptime: 00:18:40 4 [|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||100.0%] 5 [|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||100.0%] 6 [|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||100.0%] Mem[||||||||||| 137/1280MB] Swp[ 0/0MB] PID USER PRI NI VIRT RES SHR S CPU% MEM% TIME+ Command 6802 www 20 0 76232 12320 5716 R 27.0 0.9 0:06.48 php-fpm: pool www 7048 www 20 0 75200 12136 5700 R 52.0 0.9 0:03.64 php-fpm: pool www 6699 www 20 0 74176 11124 5700 R 27.0 0.8 0:07.36 php-fpm: pool www 7029 www 20 0 73668 10380 5676 R 42.0 0.8 0:03.52 php-fpm: pool www 6995 www 20 0 76228 12456 5644 R 42.0 1.0 0:03.98 php-fpm: pool www 6858 www 20 0 74172 10684 5620 R 35.0 0.8 0:05.52 php-fpm: pool www 6998 www 20 0 75200 12072 5620 R 37.0 0.9 0:03.95 php-fpm: pool www 7098 www 20 0 75200 12052 5616 R 42.0 0.9 0:02.33 php-fpm: pool www 7093 www 20 0 76228 12496 5612 R 37.0 1.0 0:03.02 php-fpm: pool www 7226 www 20 0 74692 11080 5588 R 32.0 0.8 0:00.66 php-fpm: pool www 

CPU: (cat /proc/cpuinfo | grep model)

model : 44 model name : Intel(R) Xeon(R) CPU E5620 @ 2.40GHz model : 44 model name : Intel(R) Xeon(R) CPU E5620 @ 2.40GHz model : 44 model name : Intel(R) Xeon(R) CPU E5620 @ 2.40GHz model : 44 model name : Intel(R) Xeon(R) CPU E5620 @ 2.40GHz model : 44 model name : Intel(R) Xeon(R) CPU E5620 @ 2.40GHz model : 44 model name : Intel(R) Xeon(R) CPU E5620 @ 2.40GHz 

I think a load of 10 and 100% cpu usage is not normal... How can i fix that?

Edit: I think it could be because of mysql. Here is the PhpMyAdmin Process List:

Show Full Queries ID User Host Database Command Time Status SQL query Kill 3567 bs localhost bs Sleep 2 --- --- Kill 3575 bs localhost bs Sleep 1 --- --- Kill 3576 bs localhost bs Sleep 1 --- --- Kill 3578 bs localhost bs Sleep 0 --- --- Kill 3579 bs localhost bs Sleep 0 --- --- Kill 3580 bs localhost bs Query 0 Writing to net SELECT count( id ) FROM messages WHERE uid_to =345 AND unread =1 Kill 3581 bs localhost bs Sleep 0 --- --- Kill 3582 unauthenticated user connecting host None Connect login --- Kill 3583 phpmyadmin localhost None Sleep 0 --- --- Kill 3584 root localhost mysql Query 0 --- SHOW PROCESSLIST 

As you see, there is only one thread used, but Google Analytics says:

Right now 429 active visitors on site 

I connect to Mysql per PDO. How can I get more threads?

Answer 1

Here is a link to profile your code: StackOverflow: Simplest way to profile a PHP script

Answer 2

The load should not be higher than the amount of cores. Currently you have 6, a load higher than this will produce slow responses.

It looks like you PHP scripts are consuming all CPU resources. Profiling them would be a good idea.

You may also increase PHP processing with some caching. I would recommend give a try to eaccelerator which should be easy to enable, and also take a look on nginx caching options.

Another good profiler we used was xhprof : https://github.com/facebook/xhprof but requires a bit more reading to use.

good luck