RE: [exim] Using Exim For Everything

トップ ページ
このメッセージを削除
このメッセージに返信
著者: Mark Lowes
日付:  
To: Vahric MUHTARYAN
CC: exim-users, 'Steve Sobol'
題目: RE: [exim] Using Exim For Everything
On Wed, 2004-12-22 at 16:30 +0200, Vahric MUHTARYAN wrote:
> Could you tell me about your disgn ?!
>
> İs Ldap server on seperate box.


LDAP is a cluster of three boxes (it's used for more than just the mail
backend). Final delivery is to a number of machines (hardware raid1),
we tried a storage solution but had some problems which confirmed our
view on the advisability of putting everything into a single point of
failure (your mileage will vary on this).

> How many mailbox do you do you support ?!


Around 60k active with a range of mail patterns from "sod all" through
to "you get how much mail!?", oh and users who think they have to check
mail every 15 seconds just in case something has arrived.

> Can You scale up configuration under hreavy load?!


Yes, generally through putting enough hardware in that it's running very
happily under 'normal' load and adding more when we hit a pre-defined
breakpoint. Keeping the users happy and mail flowing quickly is cheaper
than trying to use the minimum hardware to keep the costs down.

> Which documents I have to read for make design/installation and managment ?!


Actually most of the design is down to keeping the delivery path as
short as possible, so avoid sending mail round in circles (ie to spam
filtering) unless necessary, reject as much up front as possible
(non-existent users, undeliverable sender addresses etc etc). Fast
disks, a good network and understanding what is "normal" for your setup
and monitor everything (delivery rates, queue sizes, collection rates
etc etc). That way you can plan putting in new hardware long before the
users notice that there was going to be a problem.

Also if you provide backup MX for any domains get them onto a separate
box where they can clog their own queue without buggering up the
important stuff.

    Mark


--
Mark Lowes <hamster@???>