High Availability Email
lee_hoffman at brown.edu
Sat Aug 30 16:16:23 EDT 2003
We use multiple two-box mailstore clusters running cyrus, drbd, and
linux-ha to store the actual mail. On top of this we have loadbalancers
running a set of ldap boxes for authentication, and perdition to
loadbalance the frontend mail connections.
DRBD + Heartbeat (linux-ha) for the backend mail store boxes has worked
for over a year for us, but If you have the cash I recommend using
two-box redundantly shared fibrechannel SANs instead ... a lot more
expensive, but less wonky.
On Saturday, August 30, 2003, at 09:26 AM, Gary C. New wrote:
> I am gearing up to migrate our systems to a high availability email
> topology and was wondering what the current solutions are to provide
> such an architecture?
> I need a solution that synchronizes/mirrors/replicates user mail
> stores across several physical servers for redundency.
> Some of my research has pointed me to Cyrus Murder and the MUPDATE
> protocol, but it sounds like even in this type of configuration the
> back-end server would still be a single point of failure.
More information about the Info-cyrus