LDAP how is Failover done?

[Christopher Browne]

On Fri, Aug 5, 2011 at 8:57 AM, Ivan Avery Frey
<ivan.avery.frey-Re5JQEeQqe8AvxtiuMwx3w at public.gmane.org> wrote:

> At first glance I prefer Model 1. Even for the postgres folk and Chris will
> correct me if I'm wrong, multi-mastering is a "hard" problem.

It's *really* hard for the RDBMS case, basically because foreign keys
+ triggers provide a large amount of "magic" where there may be more
going on behind the scenes when you do an update, and keeping that
consistent across nodes becomes much harder.

Consider the case where you're managing inventory...

 - On node #1, you try taking 5 units of widget X out of inventory to
satisfy a customer order, while, concurrently,

 - On node #2, a different customer is supposed to get 15 units of widget X.

The DBMS might actually contain (e.g. - enforce) two constraints:
  a) Can't have less than 0 units of widget X, and, since there were
only 7 in stock, both orders can't be handled.
  b) Customers can't go over their credit limits, and, actually, the
customer buying 5 units is over their limit.

(In fact, since one customer can't afford to pay, and there's not
enough stock to satisfy the other one, the whole set of would-be sales
have to get tossed out.)

Coordinating the extra "balance bits", and making sure that
constraints get satisfied, is what makes multimaster a problem, in the
general case.

LDAP is a bit of a different story; it certainly doesn't include those
sorts of constraints or triggers, with the attendant consequence that
people can't model that, and so don't have those sorts of challenges
in their systems.

As a DIRECTORY service, (the "D" in LDAP), you don't capture balances
of things - what you're supposed to record are things that other
systems might want to reference.  And that fits reasonably well with
the ability to 'go multimaster.'

Now, the experiences I have had working with LDAP tend to make me wish
that I had instead done something less unpleasant, like poking burning
needles in my eyes.  I don't quite know why this is; I don't think
it's a "relational myopia" or anything such.

But it appears that those that use LDAP for directories have
reasonable success with its multimaster replication model.  A data
model that'll work successfully with LDAP is likely to be not too
unfriendly to multimaster replication.

My warning would surround "burning needles in eyes" and any other
sorts of "despair-inducing" aspects of its general use and deployment.
 Once you've got a deployment you're confident with, I gather MM isn't
that much more scary.
-- 
When confronted by a difficult problem, solve it by reducing it to the
question, "How would the Lone Ranger handle this?"
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists