So, your OpenLDAP server has crashed and possibly your database is 
corrupted? Yeah, I know - its a bitch.

I seriosly hope you're using back-bdb (BerkeleyDB backend) and not
back-ldbm. If you're using back-ldbm, pray you have a ldif-backup
stored somewhere - and possibly as resent as possible. If this is 
the case, jump to step [2].

NOTE!

    Before doing any of these, make a backup of your directories
    containing your database(s).

    This is just some general advices, and I do not take any 
    responsibility for whatever happens to your database.

1. Recovering a corrupted berkeleydb-based OpenLDAP database

    Find out which version of berkeleydb slapd is compiled with.
    This will somewhat have something to do with which utility
    you have to use to recover your data.

    $ ldd /usr/sbin/slapd | grep libdb
        libdb-4.2.so => /usr/lib/libdb-4.2.so (0x4006b000)

    In this example, my slapd was compiled with ver 4.2.x
    and my tool for recovery whould be /usr/bin/db4.2_recover

    $ cd <path/to/my/databasefiles/> (usually /var/lib/ldap)
    $ db4.2_recover

    On some systems, you'll only have db_recover - so you'll check to be
    shure you have the correct version before you try recovery.
    
    If this fails.. add argument -c (catastrophic recover) as 
    described in http://www.sleepycat.com/docs/ref/transapp/recovery.html

    OpenLDAP has also a description in their FAQ on how to do bdb-recovery
    based on snapshots of your db and logfiles. 

2. Recovery using flatfile exported LDIF-files.
    
    To be 100% shure you have an OK database, with all indexes working,
    you should first go to the directory containing your databases, after
    shutting down slapd.
    
    $ cd <path/to/my/databasefiles/> (usually /var/lib/ldap)

    Then delete all files, except DB_CONFIG if it exists
    Now you can start playing with slapadd, OpenLDAPs offline database
    creation tool.

    $ slapadd -v -l <your-backup.ldif>

    ...and possibly add other nice arguments (man 8 slapadd)

    What happens next is your database is created, the -v switch will
    tell you which object is being added - and if you have any indexes
    configured in slapd.conf, slapadd will re-create these as well. So
    the more indexes you have, the longer slapadd will take. 

    In some cases, I've made a slapd-without-indexes.conf I use for slapd. 
    Just to get stuff imported as quickly as possible. Then I start slapd
    in readonly mode while slapindex works on re-creating indexes in 
    the background.