Re: (ITS#4626) "glue" objectclass in syncrepl consumer

[hyc@symas.com]

ahasenack@terra.com.br wrote:
> On Fri, Sep 22, 2006 at 07:48:33PM +0000, hyc@symas.com wrote:
>> Yes, this makes sense.
>>
>> Since all of the CSNs under ou=global are greater than any CSN under 
>> dc=example,dc=com, none of those entries are sent. In server B your 
>> consumer should be configured on dc=example,dc=com, with a searchbase of 
>> ou=global,dc=example,dc=com. That will allow the server B provider to 
>> stay in sync with its consumer.
> 
> With this change, the replication is better, but the B server cannot be
> written to anymore: moving the syncrepl directive to the
> dc=example,dc=com database makes it read-only because it is then
> considered a consumer.

Hm, yes, that's unfortunate. Perhaps we should take this discussion over 
to the -devel mailing list. The whole concept of replication contexts vs 
database contexts seems to need re-thinking.

-- 
   -- Howard Chu
   Chief Architect, Symas Corp.  http://www.symas.com
   Director, Highland Sun        http://highlandsun.com/hyc
   OpenLDAP Core Team            http://www.openldap.org/project/