Re: general/406: Severe memory leak -- possibly due to SIGSEGV

[Randy Terbush <ra...@zyzzyva.com>]

Lewis,

Please give me a list of all modules that you have compiled into 
your server.

FWIW - I have been running several high volume servers on FreeBSD-2.1.x
and 2.2.x. I too see the eventual full swap area. I managed to 
track down the cause of SEGVs I was seeing as beeing a problem with 
PHP files. I have not had a single SEGV in 12 hours.

However, swap space is still disappearing at the same rate.


> 	The contract type is `' with a response time of 3 business hours.
> 	A first analysis should be sent before: Wed Apr 16 15:00:01 PDT 1997
> 
> 
> >Number:         406
> >Category:       general
> >Synopsis:       Severe memory leak -- possibly due to SIGSEGV
> >Confidential:   no
> >Severity:       critical
> >Priority:       medium
> >Responsible:    apache (Apache HTTP Project)
> >State:          open
> >Class:          sw-bug
> >Submitter-Id:   apache
> >Arrival-Date:   Wed Apr 16 11:40:01 1997
> >Originator:     lewiz@netcom.com
> >Organization:
> apache
> >Release:        Apache/1.2b8
> >Environment:
> FreeBSD 2.2.1-RELEASE #0: Sun Apr 13
> Server version Apache/1.2b8 PHP/FI-2.0b11.
> (256MB RAM, httpd.h:#define HARD_SERVER_LIMIT 500)
> >Description:
>  procs    memory     page                         faults       cpu
>  r b w    avm   fre  flt  re  pi  po  fr  sr w0   in   sy   cs us sy id
>  2 1 0 122464 87344   19   0   0   0  19   0  1  762 1970 1236 27 17 56
> 
>  grep -i SEGV error_log | wc -l
>  632  <-- 55 per hour!
> 
> ** Continual segment violations, to the tune of almost 1 per minute,
> are slowly eating away at server memory resources.  Free memory gets
> depleted until a server reboot is necessary.  Memory not being freed.
> 
> 
> 
> [Mon Apr 14 09:49:28 1997] httpd: caught SIGSEGV,
>      attempting to dump core in /usr/local/etc/httpd
> [Mon Apr 14 09:50:07 1997] httpd: caught SIGSEGV,
>      attempting to dump core in /usr/local/etc/httpd
> >How-To-Repeat:
> Just run Apache/1.2b8, and do "vmstat" over a period of days.
> On a heavily-loaded site, you will soon find that memory is not
> being reclaimed, and the grim reaper soon visits.  I can give
> you access to my site, if need be.
> >Fix:
> A salary for the developers might be a nice start.
> %0
> >Audit-Trail:
> >Unformatted:
> 

Re: general/406: Severe memory leak -- possibly due to SIGSEGV

[Lewis De Payne <le...@netcom.com>]

On Wed, 16 Apr 1997, Randy Terbush wrote:
> 
> Please give me a list of all modules that you have compiled into 
> your server.

Please also look at my PR#389, or search the database using keyword
SEGV -- there are others with that problem who may not be using PHP.
For your convenience, the modules are:
Platform:
  FreeBSD 2.2.1-RELEASE #0: Sun Apr 13
  Server version Apache/1.2b8 PHP/FI-2.0b11.
  (256MB RAM, httpd.h:#define HARD_SERVER_LIMIT 500)
Compiled-in modules:
  http_core.c
  mod_mime.c
  mod_access.c
  mod_dir.c
  mod_cgi.c
  mod_alias.c
  mod_env.c
  mod_log_config.c
  mod_imap.c
  mod_browser.c
  mod_status.c
  mod_php.c    <- latest CVS release

> 
> FWIW - I have been running several high volume servers on FreeBSD-2.1.x
> and 2.2.x. I too see the eventual full swap area. I managed to track
> down the cause of SEGVs I was seeing as beeing a problem with PHP
> files. I have not had a single SEGV in 12 hours.  However, swap space 
> is still disappearing at the same rate.

It would appear that you've done enough testing to possibly pinpoint
it to PHP.  That's not good news.  I have to run PHP -- no way around
it.  I've forwarded a copy of this to Rasmus (PHP devel).  I'll keep
you in the loop on related email I receive.

What are you doing in lieu of PHP on your servers?  Out of curiosity,
what's your load (visitors/day and hits/day, server config(s), etc).

Thanks for your message!