OpenLDAP : anomalies with search and deletions

[Vincent MATHIEU <Vincent.Mathieu@univ-nancy2.fr>]

Hello,


I am encoutering problems with OpenLDAP V1.2.3 on Solaris 2.6,  with
the DB Berkeley 2.7.5 or the GDBM 1.7.3.


The database contains about  20 000 entries.
I cannot delete some entries. When I try to do so (using 
ldapdelete, or    perldap in perl), the server returns a #
1 error message  (LDAP OPERATIONS ERROR).

Here are two examples : 

entry 17751, for which the deletion is successful :
17751
dn: uid=gacoi001,ou=IUTA,ou=etudiants,dc=univ-nancy2,dc=fr
cn: Aurore GACOIN
sn: GACOIN
userpassword: {crypt}*******
uid: gacoi001
mail: Aurore.Gacoin@etudiant.univ-nancy2.fr 
givenname: Aurore
objectclass: top
objectclass: organizationalPerson
objectclass: inetOrgPerson

and  entry 17943 whose deletion fails :
17943
dn: uid=lomba002,ou=IUTA,ou=etudiants,dc=univ-nancy2,dc=fr
cn: Romain LOMBARD
sn: LOMBARD
userpassword: {crypt}lr24796e
uid: lomba002
mail: Romain.Lombard@etudiant.univ-nancy2.fr 
givenname: Romain
objectclass: top
objectclass: organizationalPerson
objectclass: inetOrgPerson

Both entries are strictly identical.

The following are the execution outputs of the query and deletion for
these two accounts :

conn=0 op=23 SRCH
base="OU=IUTA,OU=ETUDIANTS,DC=UNIV-NANCY2,DC=FR" scope=2
filter="(&(objectclass=INETORGPERSON)(uid=GACOI001))"
conn=0 op=23 RESULT err=0 tag=101 nentries=1
DEL
dn="UID=GACOI001,OU=IUTA,OU=ETUDIANTS,DC=UNIV-NANCY2,DC=FR"
conn=0 op=24 RESULT err=0 tag=107 nentries=0


conn=0 op=39 SRCH
base="OU=IUTA,OU=ETUDIANTS,DC=UNIV-NANCY2,DC=FR" scope=2
filter="(&(objectclass=INETORGPERSON)(uid=LOMBA002))"
conn=0 op=39 RESULT err=0 tag=101 nentries=1
DEL
dn="UID=LOMBA002,OU=IUTA,OU=ETUDIANTS,DC=UNIV-NANCY2,DC=FR"
conn=0 op=40 RESULT err=1 tag=107 nentries=0

I have noticed that when doing a query on the OU field with a scope
of  'one' (1) instead of 'subtree' (2), the query is successful for
gacoi001, and fails for lomba002 :

conn=1 op=2 SRCH
base="OU=IUTA,OU=ETUDIANTS,DC=UNIV-NANCY2,DC=FR" scope=1
filter="(&(objectclass=INETORGPERSON)(uid=GACOI001))"
conn=1 op=2 RESULT err=0 tag=101 nentries=1


conn=1 op=3 SRCH
base="OU=IUTA,OU=ETUDIANTS,DC=UNIV-NANCY2,DC=FR" scope=1
filter="(&(objectclass=INETORGPERSON)(uid=LOMBA002))"
conn=1 op=3 RESULT err=0 tag=101 nentries=0


I dumped the base in LDIF, and reloaded it.  This had the
following result : I managed to delete part of the entries that could not
be deleted beforehand but not all of them. On the overall 20,000 entries,
1,000 cause problems (when comparing the number of entries with one
instead of subtree ).

One last detail : when loading the base, 7 error messages occurred 
but I could not decypher their meaning :

id 2064 already in next block
id 2064 already in next block
id 2067 already in next block
id 2068 already in next block
id 2064 already in next block
id 2064 already in next block
id 2064 already in next block

I would gratefully welcome all suggestions, because this server has been
already open and is used as an engine for a student account management
application that we run daily. The problem inhibits its proper
execution.

Thanks in advance,

--

Vincent MATHIEU                 

CRI - Universite NANCY 2             | Email : Vincent.Mathieu@univ-nancy2.fr

Pole Lorrain de Gestion              | Tel   : (33) 03.83.39.63.76

13, Rue Michel Ney - C.O. 75         | Fax   : (33) 03.83.39.64.43

54013 NANCY CEDEX