Re: slapd memory leak (ITS#249)

[dnehring@telemedia.de]

On Fri, Aug 13, 1999 at 06:57:47AM -0700, Kurt D. Zeilenga wrote:
> At 07:56 AM 8/13/99 GMT, dnehring@telemedia.de wrote:
> >Doesn't work (Sun/SPARC Solaris 2.6 with egcs 1.0.3):
>
> The FAQ has many hints to help you use pthreads under Solaris.
>   http://www.openldap.org/faq/index.cgi?file=11

Yes, recompiled it with

ac_cv_func_pthread_create=no ol_cv_kthread_flag=no ol_cv_pthread_flag=no ol_cv_pthreads_flag=no ol_cv_thread_flag=no CFLAGS="-s -O3 -D_REENTRANT" ./configure --disable-referrals --enable-shared --enable-cldap --with-ldbm-api=db2 --with-threads=posix

> >Hhmh, the machine has 768 MByte physical memory and 1.2 GByte of
> >swap. My slapd.conf values are:
> >
> >cachesize       100000
> >dbcachesize     15000000
> >dbcachenowsync
> >sizelimit       1000000
> >
> >Are this really too high values?
>
> Not if the slapd process can allocate enough memory.  Many
> operating systems have default per-process limits which are
> quite low.

We have no process limits. The recompiled version needs also very much
memory. Here my slapd.conf section:

cachesize       1000
dbcachesize     10000000
dbcachenowsync
sizelimit       1000000

After some updates with ldapmodify, the process needs about 400 MByte of
physical memory! This problems does not happen after thousands of random
reads. I'm really sure that there is definitely a memory problem. Should
I try to recompile slapd without thread-support?

Dirk

--
Dirk Nehring                | Phone: +49 5241 80-1560
Telemedia GmbH & Co. KG     | Fax:   +49 5241 80-9518
Carl-Bertelsmann-Str. 161 I | EMail: dnehring@telemedia.de
33311 Gütersloh             |