RE: architecture: active active frontends sharing common backend

[Steve Gardell]

I suspect all wrong. Cyrus' use of memory mapped files
just doesn't seem to lend itself to this. If the goal
is strictly HA, I guess that I would suggest a traditional
dual ported direct attached RAID, SAN, or the use of 
the new replication mechanisms in 2.3 

 -Original Message-
 From: [EMAIL PROTECTED] 
 [mailto:[EMAIL PROTECTED] On Behalf Of 
 Chad A. Prey
 Sent: Wednesday, April 05, 2006 3:56 PM
 To: info-cyrus@lists.andrew.cmu.edu
 Subject: architecture: active active frontends sharing common backend
 
 We're about to embark on a High Availability email project 
 and I was wondering as to the feasibility of having multiple 
 cyrus frontend machines access a single backend machine 
 (Veritas, Netapp filer, whatever it takes etc.) The n+ 
 front-end machines would all share a /var/lib/imap and a 
 /var/spool/imap on the single backend machine. DB locking and 
 other contingency issues would be left to the backend device 
 whatever that may be. A load balancer would be placed in 
 front of the cyrus boxes to handle failover etc.
 
 Does this work? Is anyone doing it now? Am I going about this 
 all wrong?
 
 Please don't tell me about Murder, it's not for HA as much as 
 it distributes load.
 
 Chad Prey
 Sr. Systems Administrator
 Salk Institute for Biological Studies
  
 -- 
 
 
 Cyrus Home Page: http://asg.web.cmu.edu/cyrus Cyrus Wiki/FAQ: 
 http://cyruswiki.andrew.cmu.edu List Archives/Info: 
 http://asg.web.cmu.edu/cyrus/mailing-list.html
 
 

Cyrus Home Page: http://asg.web.cmu.edu/cyrus
Cyrus Wiki/FAQ: http://cyruswiki.andrew.cmu.edu
List Archives/Info: http://asg.web.cmu.edu/cyrus/mailing-list.html

architecture: active active frontends sharing common backend

[Chad A. Prey]

We're about to embark on a High Availability email project and I was
wondering as to the feasibility of having multiple cyrus frontend
machines access a single backend machine (Veritas, Netapp filer,
whatever it takes etc.) The n+ front-end machines would all share
a /var/lib/imap and a /var/spool/imap on the single backend machine. DB
locking and other contingency issues would be left to the backend device
whatever that may be. A load balancer would be placed in front of the
cyrus boxes to handle failover etc.

Does this work? Is anyone doing it now? Am I going about this all wrong?

Please don't tell me about Murder, it's not for HA as much as it
distributes load.

Chad Prey
Sr. Systems Administrator
Salk Institute for Biological Studies
 
-- 


Cyrus Home Page: http://asg.web.cmu.edu/cyrus
Cyrus Wiki/FAQ: http://cyruswiki.andrew.cmu.edu
List Archives/Info: http://asg.web.cmu.edu/cyrus/mailing-list.html