Re: Replication with database-backend

[Yary Hluchan <yary@apicom.com>]

>When using the database backend (slapd-sql), how does one incooperate
>replication for data that gets edited directly in the database and not
>through the master server? Is there a way to accomplish this with
>OpenLDAP?

You have a trigger that catches the updates, inserts, and deletes.
Log them to a file in LDIF format, and send it off to slurpd.

>One possible solution would be to use caching LDAP servers instead of
>slaves. These could forward the requests to the master or, if the
>master is not available, answer from their own cache. Do you think the
>effort would be justifiable trying to modify OpenLDAP to function as
>such a cache or is there simply a better solution?

A slave always answers search requests itself. It only forwards write &
update requests back to the master.  What you're describing as a "caching"
server is really a slave server that you only connect to when you can't
connect to the master.

If you want backup for when the master goes down- search through the
"multimaster" thread on this list- best answer from that was to have
one of the slaves become the new master when it can't find the "real"
master.

-y

~~~~~
The Moon is Waxing Crescent (3% of Full)