[Mimedefang] Busy Mail Server

[Cormack, Ken]

Ray,

My solution to a high queue on my MIMEDefang server was to impliment a
"spill-over" host.

Sendmail makes this incredibly simple to do.  Just configure a second system
with all the same software and configurations of Sendmail, MIMEDefang, and
so on.

Then, on the primary machine, add the following to your sendmail.mc and
rebuild your .cf file...

dnl   THIS NEXT LINE WILL CAUSE SENDMAIL TO TRY ONCE, TO
dnl   DELIVER A MESSAGE TO THE DESTINATION SERVER.  IN THE
dnl   EVENT THAT ATTEMPT FAILS, THE NEXT LINE WILL TELL SENDMAIL
dnl   TO HAND-OFF THE MESSAGE TO THE SPECIFIED "SPILL-OVER"
dnl   HOST.  THAT HOST WILL THE CONTINUE ATTEMPTING TO DELIVER
dnl   THE MESSAGE TO IT'S DESTINATION.  THIS HAS THE EFFECT OF
dnl   OFFLOADING UNDELIVERABLE (OR SLOW TO DELIVER) MESSAGES
dnl   TO A SECOND SYSTEM, KEEPING THE MAIL QUEUE ON THE
dnl   PRIMARY HOST "LEAN AND MEAN".
define(`confFALLBACK_MX', `mail02.yourdomain.com')

>From a DNS perspective, set up the secondary host as an MX record with a
hagher number (lower priority) than the primary.  For instance, if your
primary has an "MX 10" record, assign your secondary an "MX 20" record.

Net result:  The primary will always be tried first.  Having a lean mail
queue, it's response time will be incredibly fast.  And while the primary
host is handling new mail, the secondary server, having minimal incoming
connections except those coming from the primary, can keep chugging away on
the second-and-subsequent-attempt queue congestion.

Our spill-over host has been in place now for a couple months, and the
results have been well worth the effort.

Ken

-----Original Message-----
From: Ray Parish [mailto:rparish at digitalwebzone.com]
Sent: Thursday, November 20, 2003 1:24 AM
To: mimedefang at lists.roaringpenguin.com
Subject: [Mimedefang] Busy Mail Server


I have 2 Servers (Pentium Xeon 2.8ghz servers, 2 gig ram). I have mimedefang
only rewriting the subject line to add "Potential SPAM::"

I have 

MX_SLAVE_DELAY=1
MX_MINIMUM=20
MX_MAXIMUM=40


So that being said, here is the scoop.

I have about 95 domains hitting these 2 boxes, then scanned and then sent
off to my Exchange 2000 Backend Enviroment.

I get about 160 concurrent connections to one box and the load is so high,
it is not able to accept any more connections. Then I start getting
"connection timedout" to my Exchange enviroment. 
This of course causes the mail to start Queueing up on the MIMEDefang
server.

OS is Fedora.

Anyone have some suggestions?

Thanks