Re: [Rlib-users] Memory leaks in RLib 1.3.5 / RLib CVS Version(s) libxml2 calls?! HELP!!!

[Devonium D. <dev...@de...>]

Bob,

Thanks for such a quick response.  I have downloaded RLib 1.3.6 and
installed it on our server (server software described below).  In our
Apache logs we still get child processes dying whenever a report is run
using RLib.

Apache log file:
[Tue Nov 01 15:07:36 2005] [info] Shared memory session cache
initialised
[Tue Nov 01 15:07:36 2005] [info] Init: Initializing (virtual) servers
for SSL
[Tue Nov 01 15:07:36 2005] [info] Server: Apache/2.0.46, Interface:
mod_ssl/2.0.46, Library: OpenSSL/0.9.7a
[Tue Nov 01 15:07:36 2005] [notice] Apache/2.0.46 (Red Hat) configured
-- resuming normal operations
[Tue Nov 01 15:07:36 2005] [info] Server built: Sep  1 2005 09:10:34
[Tue Nov 01 15:07:36 2005] [debug] prefork.c(1051): AcceptMutex: sysvsem
(default: sysvsem)
[Tue Nov 01 15:19:52 2005] [warn] child process 845 still did not exit,
sending a SIGTERM
[Tue Nov 01 15:19:52 2005] [warn] child process 846 still did not exit,
sending a SIGTERM
[Tue Nov 01 15:19:52 2005] [warn] child process 847 still did not exit,
sending a SIGTERM
[Tue Nov 01 15:19:52 2005] [warn] child process 848 still did not exit,
sending a SIGTERM
[Tue Nov 01 15:19:52 2005] [warn] child process 849 still did not exit,
sending a SIGTERM
[Tue Nov 01 15:19:52 2005] [warn] child process 850 still did not exit,
sending a SIGTERM
[Tue Nov 01 15:19:52 2005] [warn] child process 851 still did not exit,
sending a SIGTERM
[Tue Nov 01 15:19:52 2005] [warn] child process 852 still did not exit,
sending a SIGTERM
[Tue Nov 01 15:19:52 2005] [warn] child process 934 still did not exit,
sending a SIGTERM
[Tue Nov 01 15:19:52 2005] [warn] child process 1199 still did not exit,
sending a SIGTERM
[Tue Nov 01 15:19:52 2005] [warn] child process 1202 still did not exit,
sending a SIGTERM
[Tue Nov 01 15:19:52 2005] [info] removed PID
file /etc/httpd/run/httpd.pid (pid=768)
[Tue Nov 01 15:19:52 2005] [notice] caught SIGTERM, shutting down
[Tue Nov 01 15:19:53 2005] [info] mod_unique_id: using ip addr
72.3.139.140

I am looking into the use of valgrind and will report as soon as I
figure out how to use it :)

Thanks!
Chris


On Tue, 2005-11-01 at 11:02 -0500, Bob Doan wrote:
> The easiest thing to do is valgrind it.  That usually will tell you
> where the problem is.  If you can run it from the command line you can
> probably get it to core dump, and then you can inspect that w/ gdb.
> 
> Also try RLIB 1.3.6 to see if it fixes it.
> 
> http://www.sicom.com/~bdoan/rlib-1.3.6.tar.gz (I uploaded a fresh copy
> today w/ some more fixes - one was a core dump when using "eval")
> 
> - Bob
> 
> 
> 
> On Mon, 2005-10-31 at 22:59 -0500, Devonium Developer wrote:
> > We running PHP 5.0.4 with SOAP on Apache 2 with RHEL3.
> > 
> > We are running the most recent versions of:
> > - libxml2
> > - MySQL 4.x (And libraries)
> > - gd (gd-progs)
> > 
> > The Issue:
> > Whenever I run a report with RLib 1.3.5 or CVS (HTML or PDF) our soap
> > components stops responding.  We've been running these SOAP components
> > for 4 or 5 months with no issues until we rolled RLib out last week.
> > Now whenever we hit one of our RLib reports SOAP stops responding
> > intermittently.  The only thing I can find in common between the 2
> > modules is libxml2 of which we are running the current version.
> > 
> > >From what I can figure out some of the child precesses in apache hang
> > and whenever a SOAP component tries to reuse the child the child id not
> > found/available to be used so apache dies for that call and kills the
> > child precess.  Refresh the page and it works sometimes and sometimes
> > not.  Remove RLib from app and this stops happening.  Re-add it back in
> > and it starts again.
> > 
> > BTW, we develop on FC4 and do not see this same behavior with the same
> > application.
> > 
> > The customers love the reports but this whole application dying thing is
> > killing it :(  Thanks in advance for any help you may provide.
> >