php-fpm - server reached pm.max_children setting, consider raising it

[zSprawl]

I recently spun up a new machine on Oracle Linux 8 with php 7. The first time I fired up Zoneminder, the logs were complaining about not finding www.sock, so a new package called php-fpm was installed, which resolved the issue.

After about a day, I could no longer bring up the web interface. I checked /var/log/php-fpm.log and it had:

Code: Select all

[14-Aug-2022 09:24:23] WARNING: [pool www] server reached pm.max_children setting (50), consider raising it

So I did as it recommended, and increased pm.max_children to 100 in the /etc/php-fpm.d/www.conf file.

The next day, the same issue returned:

Code: Select all

[14-Aug-2022 09:24:23] WARNING: [pool www] server reached pm.max_children setting (100), consider raising it

I raised it again, but as you can imagine, this is merely putting off the inevitable. You can see over time the number of forked processes for php-fpm grow:

Code: Select all

[root@zm8 ~]# ps -ylC php-fpm --sort:rss
S   UID     PID    PPID  C PRI  NI   RSS    SZ WCHAN  TTY          TIME CMD
S     0  408575       1  0  80   0 24788 75908 -      ?        00:00:07 php-fpm
S    48 1190904  408575  0  80   0 27260 81134 -      ?        00:00:17 php-fpm
S    48 1190901  408575  0  80   0 27304 81134 -      ?        00:00:44 php-fpm
S    48 1190905  408575  0  80   0 27308 81134 -      ?        00:00:21 php-fpm
S    48 1210446  408575  0  80   0 27312 81134 -      ?        00:00:19 php-fpm
S    48 1197151  408575  0  80   0 27316 81134 -      ?        00:02:45 php-fpm
....
....
...
S    48 1380619  408575  0  80   0 27388 81134 -      ?        00:01:35 php-fpm
S    48 1445589  408575  1  80   0 27412 81134 -      ?        00:08:30 php-fpm
S    48 1507951  408575  1  80   0 27412 81134 -      ?        00:03:11 php-fpm
S    48 1405821  408575  0  80   0 27416 81134 -      ?        00:01:10 php-fpm
S    48 1407333  408575  0  80   0 27416 81134 -      ?        00:02:29 php-fpm
S    48 1414085  408575  0  80   0 27420 81134 -      ?        00:01:20 php-fpm
S    48 1239524  408575  0  80   0 27428 81134 -      ?        00:03:57 php-fpm
S    48 1306148  408575  0  80   0 27432 81134 -      ?        00:07:56 php-fpm
S    48 1241461  408575  0  80   0 27460 81134 -      ?        00:01:16 php-fpm
S    48 1267270  408575  0  80   0 27472 81134 -      ?        00:08:19 php-fpm
S    48 1402922  408575  0  80   0 27480 81134 -      ?        00:01:20 php-fpm

The issue is easily resolved with a systemctl reload php-fpm, so for now, I put a crontab daily task to do this to cover up the issue, but I'm wondering if anyone has run into this or has a better idea? In fact, I wouldn't be shocked if it was related to the memory leak threads, but I don't know. I have noticed the memory used for the zoneminder process keeps increasing over time, and swap is eventually used.

If you read this far, thanks!