architecture and DIT change strategy

[Guillaume Rousse]

Hello.

I'm trying to find the best way to conduct a consequent change in our
data model and servers topology, with the fewer service disturbing.

Before the reorganisation, we were a single entity, splitted on three
different sites. As a consequence, we had a single database for all our
users and groups:

dc=new,dc=foo,dc=tld
|-users
| |-site1
| |-site2
| |-site3
|-groups
|-site1
|-site2
|-site3

The master server is hosted on one site, and we have slave servers on
three sites

After the reorganisation, we are three different entities, and I'd like
to break the tree in the three different databases, each site hosting a
server acting as the master for its own base, a slave for the two others:

dc=site1,dc=foo,dc=tld
|-users
|-groups

dc=site2,dc=foo,dc=tld
|-users
|-groups

dc=site3,dc=foo,dc=tld
|-users
|-groups

The change also involves dropping the last part of the original suffix,
which is no longer relevant.

I'm currently investigating the usage slapo-rwm to provide virtual views
of the current database according to the new structure, so as to
progressively migrate applications configuration first, then write an
automated conversion tool, and finally convert the virtual bases to new
ones. But maybe they are better strategies ?

[FRLinux]

On Tue, Feb 9, 2010 at 5:16 PM, Guillaume Rousse

migrate applications configuration first, then write an automated conversion
tool, and finally convert the virtual bases to new ones. But maybe they are
better strategies ?
From my own painful experience, best scenario would be to script the

migration of users by querying them according to sites and preparing a
new ldif for a test server. Trying to get into views might work but is
most likely to break due to some configuration changes (triggered or
not by you).

Just out of curiosity, which version of OpenLDAP are you using?

Cheers,
Steph