indexes corrupt under heavy load

["Kevin Hicks" <Kevin.Hicks@drintuit.dr.avaya.com>]

I seem to have corrupt indexes if I push my system hard.

I'm running OpenLdap 2.0.18 on Redhat linux 6.2.

I have several equality and presence indexes defined in my slapd.conf like so:

index objectClass pres,eq
index Extension pres,eq
index NodeId pres,eq
index TtName pres,eq
index UserName pres,eq
index SecondExt pres,eq
index netid pres,eq
index NetName pres,eq
index FaxDsStr pres,eq

When I push my system hard,  (several clients reading from ldap, 1 client doing a bunch of reads and writes),
several of my indexes seem to get corrupt.

For instance, after running heavy load I did the following searches:

(Extension=*)        - this should use the index
(Extension>=*)     - this should not use the index

I'm getting different counts from the 2 searches above.  So, depending on if my search is using the index or not, I may or may not find an entry even though it's really there.

If I shut down ldap and do a slapindex to regenerate the indexes, it does seem to properly repair the indexes.  But I really need to avoid the problem altogether.

Any ideas or suggestions?

Thanks,

Kevin Hicks