[Date Prev][Date Next]
Re: slapd feature
Tony Earnshaw wrote:
> Subject: Re: slapd feature
>>> Is it possible to organise the "synchronisation" of schema files on slave
>>> servers ? I mean that each time one changes schema file on master schema
>>> files on slave is updated automaticaly.
> I take it that people don't have so many slaves/consumers that
> individual configuration would become impossible. Although Novell has a
> good solution with eDirectory (had already with NDS 4.1), but that's
> another ball game ;)
I can offer some ideas. These are catered not so much to OpenLDAP as
they are to the general problem of having to manage configuration
files for lots of machines.
First, refactor your configuration files as much as possible. The
1) keep as much stuff in a common configuration file as possible.
2) use separate files, and include directives to bring in sections
that vary on a per-machine basis.
For example, you might have a slapd.conf (common configuration) with
# per machine directives for replication
Then, your `repository' of configuration files would have
replication.conf.HOST-A, replication.conf.HOST-B, etc.-- each has
replication directives for a particular HOST.
Of course, you need to take steps to ensure that
replication.conf.HOST-A is installed as replication.conf on HOST-A,
and replication.conf.HOST-B is installed as replication.conf on
HOST-B. This brings us to part two ...
Use a file distribution tool. On unix, rdist
(http://www.magnicomp.com/rdist/) is pretty effective at doing this.
In your case, you'd want to distribute common configurations,
per-machine configurations, and schema files.
A variant on Part I is to use perl/make/ant/m4/etc. to generate
configuration files for distribution.
I recently heard about a tool named pica (http://pica.sourceforge.net/).
It looks pretty interesting - particularly in it's ability to
`pre-process' files before distributing them. Haven't had a chance to
try it, though.