Re: Syncrepl out-of-sync detection.

["Jose Ildefonso Camargo Tolosa" <ildefonso.camargo@gmail.com>]

Hi!

On 8/22/06, Bernd Schubert <bernd-schubert@gmx.de> wrote:
> Hello!
(.....)
> Another question, I just tested to simply dump everything with ldapsearch then
> to compare the results with diff. Tthe entries are in principle identical,
> only some newer entries are at different positions within the resulting ldif
> files. Is this common? Of course, that way my rude approach to compare the
> databases will fail. The next attempt would be to compare entry by entry, but
> maybe your script already does this?

Yes, that's common.  I would "split" the file in "entries", sort the
file by "dn" (keeping each object as a whole), and "internally" sort
each object (so that all attributes on each object are in the same
order).  Off course: make sure to keep dn as the first field.  After
sorting the files, you can use diff on them safely.

I was about to take that approach.  The problem is that it would mean
transfering the "whole" ldif from the slave to the master from
time-to-time.  Not too bad, if you only do it, say, once every 48
hours.  Other approach wold be to use rsync; or to send the DNs with
an SHA1 hash of the rest of the entry, and compare.  If something
differs, just issue a warning about it, and then transfer the whole
entry (in order to fill a complete report of the sync status).  But as
you can see, it can be a rather complex system.

I'll give a look at "hobbit" (thanks Buchan), and report about it.

Thanks for all your answers,

Sincerely,

Ildefonso Camargo.