Issue Reloading PHP-FPM fails while updating Plesk settings

[ThomasR]

We are using Plesk 12.5 on 15 servers for 6 months now. From time to time, we get a PHP-FPM error while changing settings that require a reload of the webservice. This occurs on both API calls and while using the web interface (as admin and client user).

Error (on screen or API call):

Unable to update hosting preferences. phpinimng failed: Job for plesk-php56-fpm.service failed because a fatal signal was delivered to the control process. See "systemctl status plesk-php56-fpm.service" and "journalctl -xe" for details.Failed to reload plesk-php56-fpm service
​

Log (/var/log/messages):

Sep 27 08:25:18 web03 systemd: Reloaded The PHP 5.6.25 FastCGI Process Manager.
Sep 27 08:25:19 web03 systemd: plesk-php56-fpm.service: main process exited, code=killed, status=12/USR2
Sep 27 08:25:19 web03 systemd: PID 32546 read from file /run/plesk-php56-fpm.pid does not exist or is a zombie.
Sep 27 08:25:19 web03 systemd: Reload failed for The PHP 5.6.25 FastCGI Process Manager.
Sep 27 08:25:19 web03 systemd: Unit plesk-php56-fpm.service entered failed state.
Sep 27 08:25:19 web03 systemd: plesk-php56-fpm.service failed.
Sep 27 08:25:19 web03 systemd: Starting The PHP 5.6.25 FastCGI Process Manager...
Sep 27 08:25:21 web03 systemd: Started The PHP 5.6.25 FastCGI Process Manager.​

We have no idea what is exactly triggering a failed reload (looks like it's a kind of a time-out issue).

Technical information:

• OS: CentOS Linux 7.2.1511 (Core)
• Plesk version: v12.5.30_build1205150826.19 os_CentOS 7
• Using Nginx + PHP-FPM served by Apache:
  • httpd: 2.4.6-40.el7.centos.4
  • nginx: 1.11.1-centos7.16080113
  • php: 5.4.16-36.3.el7_2
  • php-fpm: 5.4.16-36.3.el7_2
  • php54: 5.4.45-centos7.16032517
  • php56: 5.6.26-centos7.16092110
  • php70: 7.0.11-centos7.16092110
• Apache restart interval: 60 seconds

Any idea how we can prevent this behaviour?