Re: slapd mem usage and lockup

[Quanah Gibson-Mount <quanah@stanford.edu>]

--On Thursday, April 01, 2004 10:11 AM -0600 Digant Kasundra 
<digant@uta.edu> wrote:

> Hello everyone,
>
> I'm running OpenLDAP 2.2.8, compiled against the latest Heimdal,
> Cyrus-SASL, BerkeleyDB and OpenSSL.  Yesterday, I run a script that added
> 45,000 entries and it ran just fine and afterwards, the master and the
> slaves were still responding just fine.  Afterwards, the machines sat
> idle all night.
>
> This morning, the master is running fine and only showing minimal memory
> usage, but the two slaves are unresponsive and the machine seems to have
> only 6K of free memory (and the slapd processes are taking 15% of the
> memory).  Keeping in mind that these are only clunky test boxes with 768M
> of RAM, I'm still curious as to if I should be concernedl  I certainly
> don't want this behaviour on my live machines when we deploy.
>
> Right now, in slapd.conf, I have checkpoint 1025 5, and cachesize 50.
> What else can I toy with to keep memory in check or what else should I
> investigate?

How is your DB_CONFIG configured?  If you are using BDB as your backend you 
really *must* have a DB_CONFIG file.

--Quanah

--
Quanah Gibson-Mount
Principal Software Developer
ITSS/TSS/Computing Systems
ITSS/TSS/Infrastructure Operations
Stanford University
GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html