Description of problem: we found leaked storage after some time of server runtime. this was found as possible cause. Version-Release number of selected component (if applicable): [root@fort-pub23 rhel-apachectl-fix]# uname -a Linux fort-pub23 2.6.18-164.6.1.el5 #1 SMP Tue Oct 27 11:28:30 EDT 2009 x86_64 x86_64 x86_64 GNU/Linux [root@fort-pub23 rhel-apachectl-fix]# rpm -qa | grep httpd httpd-2.2.3-31.el5_4.2 [root@fort-pub23 rhel-apachectl-fix]# rpm -qf /usr/sbin/apachectl httpd-2.2.3-31.el5_4.2 [root@fort-pub23 rhel-apachectl-fix]# cat /etc/redhat-release Red Hat Enterprise Linux Server release 5.4 (Tikanga) found on 5.5 too How reproducible: wrong Steps to Reproduce: 1. leave server running with apache running. 2. disk space is leaking. 3. after reboot, space is again available Actual results: Expected results: Additional info: it looks like typo in redirection. here is the fix: # diff -Nru apachectl.orig apachectl.fixed --- apachectl.orig 2009-11-10 15:05:43.000000000 +0100 +++ apachectl.fixed 2010-04-26 17:26:06.000000000 +0200 @@ -101,7 +101,7 @@ case $ARGV in restart|graceful) - if $HTTPD $OPTIONS -t >&/dev/null; then + if $HTTPD $OPTIONS -t &>/dev/null; then $HTTPD $OPTIONS -k $ARGV ERROR=$? else
Those are equivalent in bash; what led do you make that change and what difference does it make? If you are regularly doing a reload/graceful restart of httpd, various modules can leak memory.
thanks for fast response. this redirection was the first thing that was suspicious for us. script hash /bin/sh (not bash) shebang and this type of redirection is not common (>& is used mostly with output redirection by id) i checked that newer apache2 package in debian is using the more common style "2>/dev/null" what is leaking is not ram memory, but disk space. we will follow this issue more and report any informations.
Do you have further feedback on the issue?
no i have nothing to add. i vote for closing this bug as notabug
OK, thanks for getting back to us.