nginx+php-fpm性能參數優化原則

隨著nginx的使用率不斷提高，很多朋友的服務器都在改用nginx，但是nginx處理php程序的方式是php-fpm，與apache的handler有所不同，由于很多人不會優化配置，導致網站打開的時候502、503錯誤的發生率非常高。本文中泰州網站建設跟(gen)大(da)家(jia)分享一下nginx+php-fpm性(xing)能參數的(de)優化。

1.worker_processes越(yue)大(da)越(yue)好(hao)（一定數量后性能增加不明顯）

 2.worker_cpu_affinity所(suo)有cpu平(ping)分worker_processes要(yao)比每個worker_processes都跨cpu分配性(xing)能(neng)要(yao)好;不考慮(lv)php的執(zhi)行，測試結果worker_processes數量是cpu核數的2倍性(xing)能(neng)最優

3.unix domain socket（共享內(nei)存的(de)方式）要(yao)比tcp網絡(luo)端口配(pei)置性能要(yao)好(hao)

不考慮backlog，請求速度有量(liang)級的飛躍，但錯(cuo)誤率(lv)超過50%

加上backlog，性能有10%左右(you)提升

4.調整nginx、php-fpm和內核的backlog（積(ji)壓(ya)），connect() to unix:/tmp/php-fpm.socket failed (11: Resource temporarily unavailable) while connecting to upstream錯誤的返回會(hui)減少

nginx：

配置文件的server塊

listen 80 default backlog=1024;

php-fpm:

配置文件的

listen.backlog = 2048

kernel參數：

/etc/sysctl.conf，不能(neng)低于上面的配置

net.ipv4.tcp_max_syn_backlog = 4096
net.core.netdev_max_backlog = 4096

5.增加單臺服(fu)務器上的php-fpm的master實例，會增加fpm的處理(li)能(neng)(neng)力，也能(neng)(neng)減(jian)少報錯(cuo)返(fan)回的幾率

 多(duo)(duo)實例(li)啟動(dong)方(fang)法，使(shi)用多(duo)(duo)個配置文件：

/usr/local/php/sbin/php-fpm -y /usr/local/php/etc/php-fpm.conf &
/usr/local/php/sbin/php-fpm -y /usr/local/php/etc/php-fpm1.conf &

nginx的fastcgi配(pei)置

   upstream phpbackend {
#    server  127.0.0.1:9000 weight=100 max_fails=10 fail_timeout=30;
#    server  127.0.0.1:9001 weight=100 max_fails=10 fail_timeout=30;
#    server  127.0.0.1:9002 weight=100 max_fails=10 fail_timeout=30;
#    server  127.0.0.1:9003 weight=100 max_fails=10 fail_timeout=30;
    server  unix:/var/www/php-fpm.sock weight=100 max_fails=10 fail_timeout=30;
    server  unix:/var/www/php-fpm1.sock weight=100 max_fails=10 fail_timeout=30;
    server  unix:/var/www/php-fpm2.sock weight=100 max_fails=10 fail_timeout=30;
    server  unix:/var/www/php-fpm3.sock weight=100 max_fails=10 fail_timeout=30;
#    server  unix:/var/www/php-fpm4.sock weight=100 max_fails=10 fail_timeout=30;
#    server  unix:/var/www/php-fpm5.sock weight=100 max_fails=10 fail_timeout=30;
#    server  unix:/var/www/php-fpm6.sock weight=100 max_fails=10 fail_timeout=30;
#    server  unix:/var/www/php-fpm7.sock weight=100 max_fails=10 fail_timeout=30;
   }
 
     location ~ .php* {
       fastcgi_pass  phpbackend;
#      fastcgi_pass  unix:/var/www/php-fpm.sock;
       fastcgi_index index.php;
    ..........
    }

6.測試(shi)環境和結果

內存2G

swap2G

cpu 2核Intel(R) Xeon(R) CPU E5405  @ 2.00GHz

采用ab遠程(cheng)訪(fang)問測試，測試程(cheng)序為php的字符串處理程(cheng)序

1） 在開4個php-fpm實例，nginx 8個worker_processes每(mei)個cpu4個worker_processes，backlog為1024，php的backlog為2048，內核(he)backlog為4096，采用unix domain socket連接的情(qing)況下，其他保持(chi)參數不變

性能(neng)和錯誤率較為平衡，可接(jie)受，超過4個fpm實(shi)例(li)，性能(neng)開始下降(jiang)，錯誤率并沒(mei)有(you)明顯下降(jiang)

結論是fpm實(shi)例數，worker_processes數和cpu保持倍數關系，性能(neng)較(jiao)高

影響性能和報錯的參(can)數為

php-fpm實例(li)，nginx worker_processes數量，fpm的max_request，php的backlog，unix domain socket

 10W請求，500并發無報(bao)(bao)錯，1000并發報(bao)(bao)錯率為0.9%

500并發：

Time taken for tests:  25 seconds avg.

Complete requests:    100000

Failed requests:     0

Write errors:      0

 Requests per second:   4000 [#/sec] (mean) avg.

Time per request:    122.313 [ms] (mean)

 Time per request:    0.245 [ms] (mean, across all concurrent requests)

Transfer rate:      800 [Kbytes/sec] received avg.

1000并發：

Time taken for tests:  25 seconds avg.

Complete requests:    100000

Failed requests:     524

(Connect: 0, Length: 524, Exceptions: 0)

Write errors:      0

Non-2xx responses:    524

Requests per second:   3903.25 [#/sec] (mean)

Time per request:    256.197 [ms] (mean)

Time per request:    0.256 [ms] (mean, across all concurrent requests)

Transfer rate:      772.37 [Kbytes/sec] received

2）在其他參數不變，unix domain socket換為tcp網絡端口連接，結果如(ru)下(xia)

500并發：

Concurrency Level:    500

Time taken for tests:  26.934431 seconds

Complete requests:    100000

Failed requests:     0

Write errors:      0

Requests per second:  ; 3712.72 [#/sec] (mean)

Time per request:    134.672 [ms] (mean)

Time per request:    0.269 [ms] (mean, across all concurrent requests)

Transfer rate:      732.37 [Kbytes/sec] received

1000并發(fa)：

Concurrency Level:    1000

Time taken for tests:  28.385349 seconds

Complete requests:    100000

Failed requests:     0

Write errors:      0

Requests per second:   3522.94 [#/sec] (mean)

Time per request:    283.853 [ms] (mean)

Time per request:    0.284 [ms] (mean, across all concurrent requests)

Transfer rate:      694.94 [Kbytes/sec] received

與(yu)1）比(bi)較，有大約10%的(de)性能下降

7. 5.16調整fpm的max_request參數為(wei)1000，并發1000報錯返(fan)回降到200個以下，

Transfer rate在800左右