Corrupt DB recovery?

[John Dalbec <jpdalbec@cc.ysu.edu>]

I got the message 'I/O error' from my LDAP client while adding new records to
the directory; the LDAP server died in short order.  Nothing in the logs.  I
could restart the server, but as soon as I tried to resume updating it would
die.  I figure my GDBM files must have become corrupted somehow.

I ran slapcat to a file, deleted /var/lib/ldap/*.gdbm, started the server, and
ran ldapadd, but I got error messages about unwritable attributes (creator,
createTimestamp, etc.).  I ran low on patience and stopped the server and ran
slapadd instead, plus slapindex for good measure.  Then I restarted the server
and it couldn't find the entries I created.
1) ldapadd seems to be preferred.  Are my results from using slapadd typical?
2) What is the complete list of attributes that I have to strip from the
'slapcat' output to get ldapadd to work?
3) How many times will I need to run 'ldapadd -c' before everything is back in
the database?

This is actually for future reference now;  I wound up deleting the slapcat
output file and rebuilding the directory with the migration scripts. 
(Fortunately, the system is not yet in production.)
Thanks,
John