[Date Prev][Date Next]
(ITS#6006) slapcat data invalid when using glued dbs
Full_Name: Quanah Gibson-Mount
OS: Linux 2.6
Submission from: (NULL) (18.104.22.168)
In looking at moving to 2.4 from 2.3, I was examining how glue entries are
handled (particularly since my databases are rooted at "", and thus always
glued). In particular, I'm looking at using cn=config, and of course, setting
up replication. In 2.3 I used slapd.conf, so enabling replication was always an
offline process. For 2.4 with cn=config, my goal is to specifically limit how
often people have to restart the server. With that in mind...
(a) In 2.4, if you enable replication via cn=config, the glue entry does not
exist in the slapcat output until after the server has been stopped/started.
While this doesn't break anything (the data you get from slapcat can be used on
a replica), it also isn't particularly consistent.
(b) In 2.4 (after restarting), the glue entry data from slapcat simply looks
while in 2.3, it looks like:
I.e., in 2.3 it is a real entry, while in 2.4, it does not appear to be a real
entry at all.
The major problem with this, of course, is that this is invalid LDIF, and cannot
be loaded by slapadd:
[zimbra@freelancer tmp]$ /opt/zimbra/openldap/sbin/slapadd -F
/opt/zimbra/data/ldap/config -b "" -l ldap.bak.2
slapadd: dn="" (line=1): no objectClass attribute
_ 0.09% eta none elapsed none spd 203.7 k/s
Which of course is a major problem, since I should be able to restore my DB as
exported by slapcat.