Bug #77934 [NEW]: php-fpm kill -USR2 not working

From: marc+bugsphp at cdmon dot com Operating system: Debian 9 PHP version: 7.3.4 Package: FPM related Bug Type: Bug Bug description:php-fpm kill -USR2 not working Description: ------------ When you try to restart graceful php-fpm it holds untill process_control_timeout triggers and sends TERM to each child. This happens from version 7.1.20 and 7.2.8, the fix that brokes this was http://bugs.php.net/73342 the update was at 19 Jul 2018. I was debugging the logs and found that the master process sends the SIGQUIT to the childs, and then no reply from childs to master happens since process_control_timeout stops: >>> [24-Apr-2019 10:14:26.740642] DEBUG: pid 23148, fpm_pctl_kill_all(), line 157: [pool marctestqa.com] sending signal 3 SIGQUIT to child 23165 The logs stucks on fpm_event_loop for process_control_timeout time, then makes another fpm_pctl_kill_all but with SIGTERM: >>> [24-Apr-2019 10:14:26.740697] DEBUG: pid 23148, fpm_event_loop(), line 417: event module triggered 1 events >>> [24-Apr-2019 10:14:46.741631] DEBUG: pid 23148, fpm_pctl_kill_all(), line 157: [pool marctestqa.com] sending signal 15 SIGTERM to child 23165 Ok, fpm_event_loop is a timer that waits to the fpm_event_queue_fd: >>> ret = module->wait(fpm_event_queue_fd, timeout); (line 409, file fpm_events.c, php 7.3.4) I will suppose that on debian the module system is epoll, then module = fpm_event_epoll_module();, and then we have .wait = fpm_event_epoll_wait, ret = fpm_event_epoll_module->fpm_event_epoll_wait(fd,t) And I can't follow this because I don't have more Linux internals knowledge. Test script: --------------- edit php-fpm.conf and put some time into process_control_timeout: [global] process_control_timeout = 20s Then open 2 terminals, in one call: watch -n0,1 "systemctl status php-fpm.service". On the other term call: kill -USR2 PHPFPMPID*. then see how it waits the 20 seconds and force the kill (SIGTERM) onto all the childrens. If you try this into 7.0 or prior to 7.2.8 or 7.1.20 it works. *: PHPFPMPID: The current process ID of the Master. Expected result: ---------------- A gradual restart of all the pools while they are freeing the sockets. Actual result: -------------- Wait till process_control_timeout, then SIGTERM all the pools. -- Edit bug report at https://bugs.php.net/bug.php?id=77934&edit=1 -- Try a snapshot (PHP 5.4): https://bugs.php.net/fix.php?id=77934&r=trysnapshot54 Try a snapshot (PHP 5.5): https://bugs.php.net/fix.php?id=77934&r=trysnapshot55 Try a snapshot (trunk): https://bugs.php.net/fix.php?id=77934&r=trysnapshottrunk Fixed in SVN: https://bugs.php.net/fix.php?id=77934&r=fixed Fixed in release: https://bugs.php.net/fix.php?id=77934&r=alreadyfixed Need backtrace: https://bugs.php.net/fix.php?id=77934&r=needtrace Need Reproduce Script: https://bugs.php.net/fix.php?id=77934&r=needscript Try newer version: https://bugs.php.net/fix.php?id=77934&r=oldversion Not developer issue: https://bugs.php.net/fix.php?id=77934&r=support Expected behavior: https://bugs.php.net/fix.php?id=77934&r=notwrong Not enough info: https://bugs.php.net/fix.php?id=77934&r=notenoughinfo Submitted twice: https://bugs.php.net/fix.php?id=77934&r=submittedtwice register_globals: https://bugs.php.net/fix.php?id=77934&r=globals PHP 4 support discontinued: https://bugs.php.net/fix.php?id=77934&r=php4 Daylight Savings: https://bugs.php.net/fix.php?id=77934&r=dst IIS Stability: https://bugs.php.net/fix.php?id=77934&r=isapi Install GNU Sed: https://bugs.php.net/fix.php?id=77934&r=gnused Floating point limitations: https://bugs.php.net/fix.php?id=77934&r=float No Zend Extensions: https://bugs.php.net/fix.php?id=77934&r=nozend MySQL Configuration Error: https://bugs.php.net/fix.php?id=77934&r=mysqlcfg