RE: Replication between 2.2 and 2.0

["Chris Robertson" <crobertson@gci.com>]

>> -----Original Message-----
>> From: Pierangelo Masarati [mailto:ando@sys-net.it]
>> Sent: Saturday, August 13, 2005 12:23 AM
>> To: Chris Robertson
>> Cc: openldap-software@openldap.org
>> Subject: Re: Replication between 2.2 and 2.0
>> 
>> 
>> > Sorry in advance.  This message is going to be a bit long.
>> >
>> > I have about 140 sites, spread across about 20 divisions, 
>> underneath a
>> > single company.  We are staring an upgrade of servers, but 
>> said upgrade is
>> > disruptive and is going to take some time (possibly years).  The old
>> > servers are running OpenLDAP 2.0 (2.0.23 to be exact), and 
>> the new servers
>> > are (going to be) running 2.2(.13).
>> 
>> You should go at least with the latest stable; no reason, thou, to use
>> anything less than 2.2.28 (out this weekend, hopefully).
>> 
>> >
>> > Blithely unaware, I set up a new server as a test, and 
>> tried to get it to
>> > replicate data from the central LDAP server (2.0.27).  This 
>> did not work.
>> > So I did a bit of research and found the schism between 
>> OpenLDAP 2.0 and
>> > 2.2, with a hopeful note that replication from a 2.2 master 
>> to a 2.0 slave
>> > should be possible
>> > 
>> (http://www.openldap.org/lists/openldap-software/200505/msg003
> 40.html).
>>
>> I set up a test using 2.2 as the master and 2.0 as the slave, with a
>> replica directive including
>> "attr!=structuralObjectClass,entryUUID,entryCSN".  Replication failed, and
>> the reject file showed "ERROR: Constraint violation: creatorsName: no user
>> modification allowed".
> 
> This message is likely telling you that the slave is not seeing the
> modification as occurring with the "updatedn" identity; that's the reason
> "creatorsName", "createTimestamp", "modifiersName", "modifyTimestamp" do
> not get allowed.  They are at least supposed to be allowed, if not
> required, in OpenLDAP 1.X (and since UMich's 3.3 as far as I can tell).
> 
>> Fair enough.  I expanded the attr line to include
>> "creatorsName,createTimestamp,modifiersName,modifyTimestamp".
> 
> You should check instead if the "replica" directive and the slave
> configuration match.
> 

Ugh.  That was it.  Thanks.

> p.
> 
> -- 
> Pierangelo Masarati
> mailto:pierangelo.masarati@sys-net.it
> 
> 
>     SysNet - via Dossi,8 27100 Pavia Tel: +390382573859 Fax: +390382476497
> 

Chris