Re: partial and fractional replication

[Steve Hodges <steve.hodges@oit.gatech.edu>]

Thanks Howard.  Of course, slapd writes the log and slurpd propagates
it...

More specifically, I would like to replicate not so much a branch of the
tree but rather to replicate only certain entries, based on some
filter.  For example, if a boolean attribute called "calendarUser" were
set to true, I would then replicate that entry from a master ldap server
to a calendar-specific child ldap server.  And, I would only want to
replicate a few select attributes over to the calendar-specific ldap
server, not all of them; that's what I mean by fractional replication,
which I think you indicated is possible in the current openldap.

But then I also want to be able to extend the schema on the
calendar-specific ldap server, and add on attributes particular to the
calendar application that are not part of the master ldap server.  So,
in this way, the calendar-specific ldap server would be a master (since
it would not refer me back to a master when I try an update) but in
another sense, it would function as a slave since it would receive some
replication from the master ldap.

I suppose I could maybe get around this issue by performing my
operations on the "child" ldap by binding as the replicateDN, and then I
would not get a referral back to the master when I add an attribute
specific to that child ldap.

Maybe I am making things to hard by trying to do something too far
afield from the way slapd/slurpd replication was designed to work.  Has
anyone else wanted to set up replication in this fashion and had
success, or I am trying to do things in too weird a way as far as slapd
and slurpd are concerned?

Thanks!

-steve

Howard Chu wrote:
> 
> Code for replicating subtrees of a DIT has been in place for a long time.
> (Since I don't remember which is which of partial vs fractional at this
> moment, I'm going to avoid using those terms.) The development code also
> supports filtering so that only certain attributes will be replicated. This
> would seem to answer your question, although you're not specific about how
> you want to choose which entries are to be replicated.
> 
> By the way, it's slapd that determines what gets replicated, and slapd just
> spits out the relevant info to the replog. slurpd takes whatever is in the
> replog and sends it on its way. There's no issue of "is slurpd there yet"...
> 
>   -- Howard Chu
>   Chief Architect, Symas Corp.       Director, Highland Sun
>   http://www.symas.com               http://highlandsun.com/hyc
>   Symas: Premier OpenSource Development and Support
> 
> > -----Original Message-----
> > From: owner-openldap-software@OpenLDAP.org
> > [mailto:owner-openldap-software@OpenLDAP.org]On Behalf Of Steve Hodges
> > Sent: Thursday, April 11, 2002 1:55 PM
> > To: openldap-software@OpenLDAP.org
> > Subject: partial and fractional replication
> >
> >
> > I have read the postings on the development list about partial and
> > fractional replication, but I am not clear if anyone has a working (or
> > even somewhat working) implementation of a slurpd that will handle these
> > tasks.
> >
> > In brief, I would like to implement an architecture where a large
> > central ldap server replicates only certain entries and only certain
> > attributes of those entries to several smaller, application-specific
> > ldap servers.
> >
> > If slurpd is not quite there yet, I could always just write some scripts
> > to do periodic comparisons and add/modify/deletes, but I wanted to check
> > if anyone has made progress on this yet.
> >
> > Or, are there other ldap products on the market that do this kind of
> > replication?
> >
> > Thanks!
> >
> > -steve hodges
> > Georgia Tech