[Go] How do you face-lnmp high concurrency when 502

From:http://www.topthink.com/topic/5683.html

Before PHP-FPM configuration:

Single PHP-FPM instance, using socket mode, memory 8G static mode, start PHP-FPM process number 300, the specific parameters are as follows

Listen=/tmp/Php-Cgi.Sock#listen = 127.0.0.1:9000Listen.Backlog=2048Listen.Allowed_clients=127.0.0.1Pm=StaticPmmax_children = 300 pm.= 50pm min_spare_servers = 30pm.= 250= 0 request_slowlog_timeout = 2  

Because of the schema, code, and so on, a single hundreds of concurrent 502 error occurred.

Preliminary solution: Various related optimizations

Increase Pm.max_children to 400

Nginx and fpm Added Listen.backlog = 2048

Maximum open file handle 65535

/etc/sysctl.conf have been fine-tuned, high concurrency when Nginx initiated the number of connections, far more than the number of PHP-FPM can handle, causing the port (or socket) is frequently locked, causing congestion. 502 Errors still occur

The ultimate Solution:

Enable two PHP-FPM instances, the php-fpm divided into two parts, each listening to a port or socket, which reduces lock, still maintain 400 php-fpm process, 200 per instance, with Nginx upstream load balancing, Poll each socket to process the request.

Specific operation:

CP PHP-Fpm.conf PHP-Fpm2.Confvi PHP-Fpm2.ConfMake the appropriate changes[Global]Pid=/usr/Local/Php/Var/Run/Php-Fpm2.Piderror_log=/usr/Local/Php/Var/Log/Php-Fpm2.Loglog_level=Notice[Www]Listen=/tmp/Php-Cgi2.Sock#listen = 127.0.0.1:9000Listen.Backlog=2048Listen.Allowed_clients=127.0.0.1Pm= StaticPm.Max_children=200Pm.Start_servers=50Pm.Min_spare_servers=30Pm.Max_spare_servers=250Request_terminate_timeout=0Request_slowlog_timeout=2Slowlog=Var/Log/Slow.Logcp/etc/Init.D/Php-Fpm/etc/Init.D/Php-FPM2 VI/etc/Init.D/Php-Fpm2ModifyPrefix=/usr/Local/Phpexec_prefix=${Prefix}Php_fpm_bin=${exec_prefix}/sbin< Span class= "pun" >/php-fpmphp_fpm_conf=< Span class= "PLN" >${prefix}/etc /php-fpm2.=${prefix }/var/run/ php-fpm2. PID                  

Start PHP-FPM2 and then

Configure Nginx

Edit the nginx.conf master configuration file, if the backend uses a virtual host, as I do,

Add to

Upstream Backend{              server  unix:/tmp/php-. Sock;               server unix :/tmp/php-cgi2sock;}vi vhost/testconf               

The

Modify here  fastcgi_pass  backend;  call fastcgi is a way to use load balancing.

location ~  [^/]\.php (/| $  { Try_files $uri =< Span class= "lit" >404; Fastcgi_pass backend; # fastcgi_pass 127.0.0.1:9000;  Fastcgi_index index.; include Fastcgi.; # include pathinfo.conf;  }            /span>                

Restart Nginx.

Wait for verification, the 502 error will be greatly reduced, the website snapped, the consumer very happy.

Summarize:

TCP ports are relatively stable in high concurrency, but using ports is a bit less efficient than the socket. LNMP environment, in the face of high concurrency, in addition to a reasonable architecture, and reasonable tuning, the developer's code logic and efficient code is also an important factor affecting high concurrency. How many times a request is called PHP-FPM, and how much time each PHP-FPM handles is a point that developers need to consider.

[Go] How do you face-lnmp high concurrency when 502