Graceful restart php-fpm

Php-Fpm allows for a graceful restart of child, usually with the reload keyword instead of restart on the init script.

Setup

All you have to do is to set process_control_timeout to something other than 0 for this to work. innodb_lock_wait_timeout defaults to 50 seconds, it’s a good practice to set process_control_timeout with the default of innodb_lock_wait_timeout.

process_control_timeout mixed Time limit for child processes to wait for a reaction on signals from master. Available units: s(econds), m(inutes), h(ours), or d(ays) Default Unit: seconds. Default value: 0.

To see if the php-fpm process is running on Centos you can use sudo systemctl status php-fpm

● php-fpm.service - The PHP FastCGI Process Manager
   Loaded: loaded (/usr/lib/systemd/system/php-fpm.service; enabled; vendor preset: disabled)
   Active: active (running) since Thu 2018-01-01 11:57:37 EEST; 1 months 0 days ago
  Process: 9294 ExecReload=/bin/kill -USR2 $MAINPID (code=exited, status=0/SUCCESS)
 Main PID: 32688 (php-fpm)
   Status: "Processes active: 0, idle: 2, Requests: 246, slow: 0, Traffic: 0.6req/sec"
    Tasks: 3
   Memory: 73.6M
   CGroup: /system.slice/php-fpm.service
           ├─ 9301 php-fpm: pool www
           ├─ 9302 php-fpm: pool www
           └─32688 php-fpm: master process (/etc/php-fpm.conf)

You can reload the php-fpm process just by running sudo /bin/systemctl reload php-fpm. You can find out what systemd reload php-fpm will do by looking at the ExecReload= option in the [Service] section in the php-fpm.service unit file.

[Unit]
Description=The PHP FastCGI Process Manager
After=syslog.target network.target

[Service]
Type=notify
EnvironmentFile=/etc/sysconfig/php-fpm
ExecStart=/usr/sbin/php-fpm --nodaemonize
ExecReload=/bin/kill -USR2 $MAINPID
PrivateTmp=true
RuntimeDirectory=php-fpm
RuntimeDirectoryMode=0755

[Install]
WantedBy=multi-user.target

Or by running

sudo systemctl show php-fpm --property=ExecReload

Output

ExecReload={ path=/bin/kill ; argv[]=/bin/kill -USR2 $MAINPID ; ignore_errors=no ; start_time=[Fri 2021-12-10 21:56:37 EET] ; stop_time=[Fri 2021-12-10 21:56:37 EET] ; pid=9294 ; code=exited ; status=0 }

So the -s HUP means configuration reload, start the new worker processes with a new configuration. Gracefully shutdown the old worker processes.

If you liked this post, you can share it on Twitter. Also you can follow me on Github or endorse me on LinkedIn.