PHP-FPM在單一容器中啟動後如何啟動Nginx（或監聽http埠）？

Question

我在 Cloud Run 上的單一容器中執行 Nginx+PHP-FPM 時遇到問題。我的容器是基於 Alpine，並由 Supervisor 管理 Nginx 和 PHP-FPM 啟動。總的來說，它運作得很好，但是 Nginx 開始監聽 HTTP 連接埠和 PHP-FPM 啟動之間的時間很短。這會導致出現 502 HTTP 錯誤並顯示以下日誌訊息：

6#6: *3 connect() failed (111: Connection refused) while connecting to upstream, client: 169.254.8.129, server: _, request: "POST / HTTP/1.1", upstream: "fastcgi://127.0.0.1:9000"

這裡的問題是，Cloud Run 在開啟 8080 連接埠時確定容器已準備好處理請求。連接埠開啟後，Cloud Run 立即發送請求，但第一次嘗試總是失敗，因為 FPM 尚未準備好。NOTICE: fpm is running, pid 4第一個請求到達但失敗後會顯示日誌訊息。

如何管理Nginx僅在PHP-FPM準備就緒時才開啟其連接埠？

主管配置：

[supervisord]
nodaemon=true
logfile=/dev/null
logfile_maxbytes=0
pidfile=/run/supervisord.pid

[program:php-fpm]
command=php-fpm -F
stdout_logfile=/dev/stdout
stdout_logfile_maxbytes=0
stderr_logfile=/dev/stderr
stderr_logfile_maxbytes=0
autorestart=false
startretries=0

[program:nginx]
command=nginx -g 'daemon off;'
stdout_logfile=/dev/stdout
stdout_logfile_maxbytes=0
stderr_logfile=/dev/stderr
stderr_logfile_maxbytes=0
autorestart=false
startretries=0

Nginx 設定：

# Write temporary files to /tmp so they can be created as a non-privileged user
client_body_temp_path /tmp/client_temp;
proxy_temp_path /tmp/proxy_temp_path;
fastcgi_temp_path /tmp/fastcgi_temp;
uwsgi_temp_path /tmp/uwsgi_temp;
scgi_temp_path /tmp/scgi_temp;

access_log /dev/stdout;
error_log /dev/stderr notice;

server {
        listen 8080 default_server;

        index index.php;

        keepalive_requests    10;
        keepalive_timeout     60 60;

        root /var/www/html/app/public;

        charset utf-8;

        server_name _;

        # Deny hidden files (.htaccess, .htpasswd, .DS_Store).
        location ~ /\. {
            deny            all;
            access_log      off;
            log_not_found   off;
        }
        ########################
        # mappings             #
        ########################

        location ~ \.(js|css|png|jpg|ico) {
            expires 5d;
            try_files $uri $uri/ =404;
            return 404;
        }

        # Allow fpm ping and status from localhost
        location ~ ^/(fpm-status|fpm-ping)$ {
            access_log off;
            allow 127.0.0.1;
            deny all;
            fastcgi_param SCRIPT_FILENAME $document_root$fastcgi_script_name;
            include fastcgi_params;
            fastcgi_pass 127.0.0.1:9000;
        }

        location / {
            client_max_body_size 100m;
            try_files $uri @fpm;
        }

        location @fpm {
            # worker may take long time to finish (max 1 hour)
            fastcgi_read_timeout 3600s;
            fastcgi_split_path_info ^(.+\.php)(/.*)$;
            include fastcgi_params;
            fastcgi_param HTTP_PROXY "";
            fastcgi_pass 127.0.0.1:9000;
            fastcgi_index index.php;
            fastcgi_param SCRIPT_FILENAME /var/www/html/app/public/index.php;
            fastcgi_param SCRIPT_NAME index.php;
        }
}

我已啟用 FPM ping/狀態頁面。我可以使用它們來觸發 Nginx 連接埠開啟嗎？

更新1：

我嘗試調整supervisord優先順序並啟動秒：

...
[program:php-fpm]
...
priority=100
startsecs=3

[program:nginx]
...
priority=200

但沒有成功：

[18-Dec-2020 00:31:04] NOTICE: ready to handle connections
[18-Dec-2020 00:31:04] NOTICE: fpm is running, pid 3
2020-12-18 00:30:30,689 INFO success: php-fpm entered RUNNING state, process has stayed up for > than 3 seconds (startsecs)
2020-12-18 00:30:28,388 INFO success: nginx entered RUNNING state, process has stayed up for > than 1 seconds (startsecs)
Error POST 502 549 B 3.286 s Google-Cloud-Scheduler https://***.run.app/
169.254.8.129 - - [18/Dec/2020:00:30:27 +0000] "POST / HTTP/1.1" 502 150 "-" "Google-Cloud-Scheduler"
169.254.8.129 - - [18/Dec/2020:00:30:27 +0000] "POST / HTTP/1.1" 502 150 "-" "Google-Cloud-Scheduler" "35.187.131.214"
2020/12/18 00:30:27 [error] 6#6: *3 connect() failed (111: Connection refused) while connecting to upstream, client: 169.254.8.129, server: _, request: "POST / HTTP/1.1", upstream: "fastcgi://127.0.0.1:9000", host: "***.run.app"
2020-12-18 00:30:26,937 INFO spawned: 'nginx' with pid 4
2020-12-18 00:30:26,829 INFO spawned: 'php-fpm' with pid 3
2020-12-18 00:30:25,730 INFO supervisord started with pid 1
2020-12-18 00:30:25,704 CRIT Supervisor is running as root. Privileges were not dropped because no user

兩個應用程式仍然由 Supervisord 同時啟動，並且 nginx 首先初始化。RUNNINGSupervisord 應用於應用程式的狀態對於 Cloud Run 來說毫無意義。

Answer 1

現在，我最終得到了以下入口點腳本，確保 PHP-FPM 在啟動 nginx 之前運行：

#!/usr/bin/env sh
set -e

# Start PHP-FPM as a daemon in the background
php-fpm -D

# Wait until PHP-FPM is up and accepts connections. Fail if not started in 10 secs.
for run in $(seq 20)
do
  if [ "$run" -gt "1" ]; then
    echo "Retrying..."
  fi
  RESPONSE=$(
    SCRIPT_NAME=/fpm-ping \
    SCRIPT_FILENAME=/fpm-ping \
    REQUEST_METHOD=GET \
    cgi-fcgi -bind -connect 127.0.0.1:9000 || true)
  case $RESPONSE in
    *"pong"*)
      echo "FPM is running and ready. Starting nginx."
      # Run nginx without exiting to keep the container running
      nginx -g 'daemon off;'
      exit 0
      ;;
  esac
  sleep .5
done
echo "FPM has failed to start on-time, exiting"
exit 1

該apk add fcgi命令是必需的（對於 Alpine linux）。

我還有一個假設，該php-fpm -D命令總是在 FPM 準備好後退出，因此不需要循環，只需一個接一個地運行命令即可。但我沒有測試過。

Answer 2

Supervisor 確實可以使用一個startdelay選項，但目前沒有。要延遲 nginx 的啟動，請將指定的命令更改為 Supervisor (相關主題）：

1. 運行一個shell進入睡眠狀態，然後直接啟動nginx：

   [program:nginx]
   command=bash -c "sleep 5 && exec nginx -g 'daemon off;'"
   ...
   

2. 執行一個在啟動 nginx 之前休眠的腳本：

   [program:nginx]
   command=delayed-nginx.sh
   ...
   

   delayed-nginx.sh：

   #!/bin/bash
   sleep 5
   exec nginx -g 'daemon off;'
   

原答案

我沒有專門使用Supervisor，但是這個答案可能適合你。這將導致 Supervisor 僅php-fpm在運行時啟動 nginx 。

[program:php-fpm]
command=php-fpm -F
...
priority=100

[program:nginx]
command=nginx -g 'daemon off;'
...
priority=200

您可能還需要設定startsecs為更高的值（比預設的 1 秒），以便 Supervisor 僅php-fpm在運行了較長時間（例如 5 秒）後才考慮啟動：

[program:php-fpm]
command=php-fpm -F
...
priority=100
startsecs=5

請參閱主管文件以了解更多資訊。