New mail server setup

Michael K. Smith - Adhost mksmith at adhost.com
Wed Sep 16 16:32:16 UTC 2009


> -----Original Message-----
> From: owner-freebsd-questions at freebsd.org [mailto:owner-freebsd-
> questions at freebsd.org] On Behalf Of Steve Bertrand
> Sent: Wednesday, September 16, 2009 7:09 AM
> To: Matthew Seaman
> Cc: questions at freebsd.org
> Subject: Re: New mail server setup
> 
> Matthew Seaman wrote:
> > Steve Bertrand wrote:
> >
> 
> >> If anyone has a setup that has redundancy for their IMAP/POP
> services,
> >> and a method to keep the changing data relatively up-to-date, I'd
> love
> >> to hear about it.
> >
> > Now, that is a different kettle of fish.  This is a job for cyrus
> imap.
> > I suggest googling for 'cyrus murder' -- this is almost, but not
> quite,
> > a fully resilient mail store / IMAP system.  Your mail store is
> divided
> > into frontend IMAP protocol servers which handle user auth etc. and
> back-end
> > mail stores.  The protocol layer servers are fully resilient and you
> can
> > fail over a user session at will, but the mailstores don't quite get
> there:
> > mail is replicated across different stores, but actions modifying
the
> mail
> > store are not transactional across all the mail stores. Or in other
> words,
> > you can lose a small amount of data if one of the mail stores goes
> bang at
> > precisely the wrong moment.  Even so, it will do better at keeping
> multiple
> > copies of a mailstore in synch than any locally scripted rsync
setup.
> 
> This is *EXACTLY* what I was looking for!
> 
> The possibility of loosing an extremely small amount of data far
> outweighs the possibility of a multi-hour outage where 3,000 users are
> receiving "can't reach the POP3 server" errors.
> 
> Besides, our incoming SMTP gateway boxes cache all incoming email for
> 24
> hours, and we can re-deliver any message to the back-end we wish
during
> that window.
> 
> I really try my best to design/implement all the systems I can like
our
> networks... multiple paths and extremely quick convergence. Being able
> to take a box down to test/perform an upgrade, or during a failure
> without client impact is well worth any initial large learning curve
> imho.
> 
> Thanks,
> 
> Steve

Hello Steve:

Another approach would be a cluster of Postfix servers and Dovecot
servers behind PF load balancers.  We have 3 "POP" servers (IMAP/POP), 9
Mail Servers, 2 Defer servers and 5 Filter servers that process over 20
million messages a day without a blip.  We can take individual servers
out of the pool for maintenance, etc.  Everything is fed to a set of
redundant NAS for the data storage and common configuration files.

Regards,

Mike

--
Michael K. Smith - CISSP, GISP
Chief Technical Officer - Adhost Internet LLC mksmith at adhost.com
w: +1 (206) 404-9500 f: +1 (206) 404-9050
PGP: B49A DDF5 8611 27F3  08B9 84BB E61E 38C0 (Key ID: 0x9A96777D)




More information about the freebsd-questions mailing list