为什么要监控php-fpm的状态？ 因为php-fpm多进程模型，线上一般都是static静态进程数配置，很容易因为I/O慢导致进程全部占满，造成拒绝服务。通过Prometheus采集PHP-FPM的性能指标来监控PHP-FPM

开启php-fpm status

php-fpm配置

1
2
3
4
5

#php-fpm.conf

[www]
pm.status_path = /phpstatus
listen = /dev/shm/php-cgi.sock

nginx配置

1
2
3
4
5
6

#nginx.conf

location ~^/phpstatus$ {
    fastcgi_pass   unix:/dev/shm/php-cgi.sock;
    include fastcgi.conf;
}

php-fpm status详解

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16

curl http://localhost/phpstatus

pool:                 www
process manager:      static
start time:           09/Mar/2022:08:19:49 +0000
start since:          18505708
accepted conn:        30247806
listen queue:         0    //请求等待队列，如果这个值不为0，那么要增加FPM的进程数量
max listen queue:     0    //请求等待队列最高的数量
listen queue len:     0    //socket等待队列长度
idle processes:       99    //空闲进程数量
active processes:     1     //活跃进程数量
total processes:      100   //总进程数量
max active processes: 19    //最大的活跃进程数量
max children reached: 0
slow requests:        3859  //慢日志数量

数据采集

php-fpm-exporter采集数据

1
2
3
4
5

#下载exporter
#https://github.com/bakins/php-fpm-exporter/releases

#启动exporter
./php-fpm-exporter.linux.amd64 --addr 0.0.0.0:9190 --endpoint=http://localhost/phpstatus

配置Prometheus

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22

#targets.json

[
    {
          "targets":["10.0.0.100:9190"],
          "labels": {
              "job": "web-php",
              "app": "web-php-test",
              "env": "test",
              "instance": "10.0.0.100:9190"
          }
      },
      {
          "targets":["10.0.0.101:9190"],
          "labels": {
              "job": "web-php",
              "app": "web-php-dev",
              "env": "dev",
              "instance": "10.0.0.101:9190"
          }
      }
]

prometheus

配置Grafana

grafana

模板JSON