[Date Prev][Date Next] [Chronological] [Thread] [Top]

Lockup when moding one particular entry



   I am running OpenLDAP 2.2.26.  I have recently run into a problem:
when I try to add a new member attribute to an entry that has top and
groupOfNames as its only objectClasses, slapd will hang when trying to
run ldap_mod_add.  The ldap log file shows the line stating that it is
performing a mod:

slapd[19807]: conn=78 op=1 MOD attr=member

but it will not finish and write the next line that shows it completed.
Slapd will then hang indefinitely. If I kill and restart slapd, it no
longer runs properly and I must run db_recover to repair the database.
If I try and run slapcat on this corrupted database without recovering
it first, slapcat hangs and the ldif file it writes contains only
entries up to several entries before the group entry that is causing the
problem.


The strange part about this problem is that once I have killed slapd and
recovered the database (using db_recover) I can do the exact same
operation (i.e., ldap_mod_add) and it works fine. But the next morning
if I try again it will hang on the same operation.


I can do other ldap_mod_adds on other entries of different objectClasses
and they do not hang and then do the group mod and it hangs.  Running
slapcat or ldapsearch on the entry in question from a non-corrupted
database shows no difference between this entry and any other
non-problematic entry with the same two object classes and similar
member attributes.

I even retyped the offending group entry and used ldapdelete to remove it and added it back in using ldapadd and the problem still persists.

Any ideas about what is happening and/or causing this condition would be
greatly appreciated.


-- David Whidbey Telecom Internet and Broadband System Programmer/Sys Admin