--On Thursday, August 17, 2006 1:09 PM -0400 Amith Varghese <[EMAIL PROTECTED]> wrote:
I'm running FC-3 with OpenLDAP RPMS (v. 2.2.29). Recently, my backend was corrupted which caused me to rebuild the database. After doing some research, I learned the BDB backend was the "preferred" backend and that support for ldbm is on the way out. So I converted to the BDB format using slapcat/slapadd and a new configuration file. Everything converted fine, but I'm now experiencing worse performance that I did with the ldbm backend. I've read the online docs about how to tune the BDB backend and added the following parameters to my slapd.conf file directory /var/lib/ldap cachesize 100000 checkpoint 1024 5 In my /var/lib/ldap/DB_CONFIG file I have the following set: set_cachesize 0 104857600 1 set_lg_regionmax 1048576 set_lg_max 10485760 set_lg_bsize 2097152 However, when I use slapd_db_stat to look at the BDB backend statistics, I see the following: 264KB 48B Total cache size 1 Number of caches 272KB Pool individual cache size 212781 Requested pages found in the cache (88%) It looks like the backend is using some kind of default (264KB) instead of the 100MB I specified. Does anyone know why this may be happening? Hopefully someone can point out an obvious mistake I'm making.
Did you stop slapd and run db_recover with the correct version of db_recover after creating the DB_CONFIG file? It is only read one time, at the creation of the DB environment.
I'll also note that 2.2.29 is very old, and I believe RH FC3 links it against BDB 4.3, which has seemed rather unstable in comparison to BDB 4.2 or BDB 4.4. You probably want to investigate using Buchan Milne's packages, or a pre-packaged distribution like is available from Symas Corporation (<http://www.symas.com>).
--Quanah -- Quanah Gibson-Mount Principal Software Developer ITS/Shared Application Services Stanford University GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html
