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

Re: slapd fails to add multi-valued attributes from slurpd (ITS#2384)

I note that replicating between heterogeneous versions of
OpenLDAP is not supported.  slurpd(8) assumes both the
master and the slave are of the same version as itself.

At 11:32 AM 3/17/2003, you wrote:
>Full_Name: Andrew Hamlin
>Version: 2.0.27
>OS: Linux 2.2.19
>URL: ftp://ftp.openldap.org/incoming/
>Submission from: (NULL) (
>We've recently upgraded from 1.2 to the 2.0 version of openldap. Now we are
>finding that multi-valued attributes are not being replicated between servers. 
>Test scenario:
>1)  Setup a 1.2 master and a 1.2 replica and 2.0 replica.
>In real life we're using 2 or more 2.0 servers, not a mix of 1.2 and 2.0!
>2)  Add an entry with a multi-valued attribute. 
>We use some custom objectclasses, so I've just paraphrased the entry here. Setup
>a test user that includes a valid multivalued attribute.
>dn: uid=testuser@somewhere.com, dn=somewhere, dn=com
>uid: testuser@somewhere.com
>gn: test
>sn: user
>cn: test user
>multivalueattrib: line1 of a multi valued entry
>multivalueattrib: line2 of a multi valued entry
>multivalueattrib: line3 of a multi valued entry
>objectclass: person
>3) After running ldapadd look at the reject file for the 2.0 replica. It will
>not have replicated to this server with the error: Type or value exists.  
>Search either of the 1.2 servers and the entry will be there.
>This test has narrowed the possible failure points down to the 2.0 version of
>slapd (atleast as the most likely suspect). 
>slapd v.1.2 is adding the entry correctly. 
>slurpd 2.0 is not even running because the master is running openldap 1.2. 
>For some reason the 2.0 version of slapd is rejecting a multivalued attribute
>when the source of the new entry is the slurpd daemon. 
>Oh, for the time being the rootdn is doubling as the updatedn on the replicas.
>I've found no mention of this on any mailing list archives or web sites, other
>than one mail posting in russian which I translated through a babelfish site and
>was able to decipher enough that they may have been seeing the same problem we
>Thanks for you help,