openldap (or is it nss_ldap?) not answering

[Stefano Maffulli <smaffulli@gmail.com>]

Hello all,

I have installed a mail server with real users, stored in openldap. 
The dovecot server sometimes hangs waiting for an answer from ldap, an
answer that doesn't come.  The openldap server does not crash, it just
sits there.  A restart of the slapd server makes everything work
again.

The system is a dual Xeon, fast raid disks and serves only ~70 users. 
It is running debian-unstable, openldap 2.1 with ldbm backend.  All
default values from debian install, I didn't tweak cache nor threads
in slapd.conf.

As a temporary solution I put a cron job that restarts slapd, but I
would like to understand what is going wrong and what parameters I can
change.  Would it help to change the backend to bdb?  What would be
the procedure to do this safely?  What else can I do?  I have no idea
on how to come out of this situation.  The error is not replicable, it
happens randomly, probably it is related to the amount of people  in
the office.

The only error I see in the log is the following:
Nov 11 18:36:09 csv-mail-01 top: nss_ldap: reconnecting to LDAP server...
Nov 11 18:36:09 csv-mail-01 top: nss_ldap: reconnected to LDAP server
after 1 attempt(s)

and sometimes also apache2 says the same (reconnecting).  Is this
normal?  Finally, is there a tool to benchmark openldap to see how it
performs when changing parameters in slapd.conf?

thanks for any hint
stef