Re: Openldap 2.3.39 Solaris 9 (maybe memory leaks).

[Quanah Gibson-Mount <quanah@zimbra.com>]

--On December 17, 2008 1:31:43 PM +0100 Andrea Cirulli <acirulli@gmail.com> 
wrote:

> Hi all,
>
> we are experiencing a lot of problems with openldap 2.3.39 on Solaris 9.
>
> We have the following set:
> 1- One producer Solaris 9 openldap 2.3.39
> 2- 19 Consumer, synchronized with the producer through syncrepl, Solaris
> 9 openldap 2.3.39
> 3- We manage the authentication on 2000 systems , almost equally
> distributed on the 20 ldap server
>
> The producer is often going down, each hour and less during the day.
>
> We noticed that every time the Producer is going down the virtual memory
> used is bigger than 3,8 Gb and the RSS (resident set size) is bigger than
> 2,1 Gb.
> This is a snapshot of prstat when openldap is about of crashing:
>
> PID USERNAME  SIZE   RSS STATE  PRI NICE      TIME  CPU PROCESS/NLWP
> 1965 root     3873M 2179M sleep   60    0   0:41:40 0.5% slapd/27.
>
> The openldap is compiled in this way:
> ./configure --enable-bdb --enable-ppolicy --with-tls
>
> If these  informations are not enough, I could provide other informations.
>
> It seems that there is a memory leak.
>
> P.S: We are using these systems since 2,5 years, the number of the
> managment systems is always growing up. We have this problem since 2
> weeks.

What are the settings in your DB_CONFIG file?  What are your cache settings 
in slapd.conf?  Have you changed either of these recently?  I'll also note 
that OpenLDAP 2.3.39 is quite old.  2.3.43 was the final release in the 2.3 
series, and 2.4.13 is the current release.  You might want to look at the 
changes file for 2.3 and see if anything related to a memory leak was fixed 
between 2.3.39 and 2.3.43.

--Quanah

--

Quanah Gibson-Mount
Principal Software Engineer
Zimbra, Inc
--------------------
Zimbra ::  the leader in open source messaging and collaboration