Cyrus IMAP and SASL on replicated machines

[Jeroen van Meeuwen (Kolab Systems)]

J. Pilfold-Bagwell wrote:
> Hi All,
> 
> I have a Cyrus box that I set up about 3 years ago that's been running
> flawlessly.  Recently though, as we're becoming increasingly reliant on
> email, it was decided that we're going to set up a DRBD replicated system.
> 

While I'm not trying to negate that decision having been made, as I too, 
amongst many other people, enjoy the existence and purpose of DRBD, I have to 
ask whether or not the contradiction / distinction between storage-level and 
application-level replication has been taken into account.

Please note that what I'm about to say, hopefully generating some feedback 
from other people as well, would mean more baggage for me to put into all 
sorts of Cyrus IMAP Deployment Guides and such.

An example scenario is where the 'master' (SQL, IMAP) server is the active 
server; it's DRBD replicated storage segment cannot just be live (locking, 
(a)synchronous filesystem operations, etc.); This type of scenario says 'warm 
failover' at best, I recon.

With application-level replication however, noted that master-master (round-
about) or multi-master replication has to be a replication scenario the 
application is capable of dealing with, both systems could be active (doesn't 
matter which one you hit).

I suppose the point is DRBD is ideal for Disaster Recovery, and insert a 
remark on synchronizing new (large) volumes over little bandwidth if you like, 
whereas application-level replication may just bring you high-availability, 
load-balancing and disaster-recovery.

Just a few of my thoughts, I hope you appreciate and if you feel like it, 
don't hesitate to question! ;-)

Kind regards,

Jeroen van Meeuwen

-- 
Senior Engineer, Kolab Systems AG

e: vanmeeuwen at kolabsys.com
t: +44 144 340 9500
m: +44 74 2516 3817
w: http://www.kolabsys.com

pgp: 9342 BF08