I have an issue with mod_fcgid spawning new php processes and after some
time it looses control over php-cgi. I see many php-cgi processes running
for long time. When I check parent process ID I see it is init owning them.
They aren’t zombies though. It means httpd lost control over these
processes. For time being number of process grows to the point my server
goes out of memory. Stopping and starting httpd doesn’t kill them. Only way
to stop php-cgi processes is to send the kill signal directly to the
process.
Here are my wrapper settings:
#!/bin/sh
# Do not change this file
PHPRC=/etc/
export PHPRC
export PHP_FCGI_MAX_REQUESTS=5000
export PHP_FCGI_CHILDREN=0
exec /usr/bin/php-cgi
Here is our mod_fcgid config settings:
PHP_Fix_Pathinfo_Enable 1
FcgidProcessTableFile /tmp/fcgidshm_aux7
FcgidIPCDir /tmp/.fcgidsock_aux7
FcgidMinProcessesPerClass 0
FcgidMaxProcessesPerClass 70
FcgidMaxProcesses 8000
FcgidMaxRequestLen 1073741824
FcgidMaxRequestInMem 20971520
FcgidIdleTimeout 30
FcgidIdleScanInterval 60
FcgidBusyTimeout 300
FcgidBusyScanInterval 120
FcgidErrorScanInterval 20
FcgidZombieScanInterval 3
FcgidProcessLifeTime 60
FcgidIOTimeout 300
I see this problem on mod_fcgid 2.3.6 with Apache 2.2.3 distributed by
Cloudlinux repository for CentoOS 5 as well as for most recent version of
mod_fcgid 2.3.7 with Apache httpd 2.2.15 for Cent OS 6.2.
Only solution I have so far is script like one shown here:
http://www.sonassi.com/knowledge-base/abandoned-php-fcgi-process-clean-up/
Screenshot showing example of a php-cgi processes that have lost parent
control.
http://quick-markup.com/p/500f1d0461f16
**I have cronjob doing httpd graceful restart each hour. The issue here may
be related to that cronjob. Although I haven’t observed it is caused by
graceful restart. Usually it takes couple of days when I see php-cgi
processes being old.
Is there anything else I should check in my system to eliminate this issue?
--
