From owner-freebsd-questions  Thu Jun 27 07:53:20 1996
Return-Path: owner-questions
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.5/8.7.3) id HAA03288
          for questions-outgoing; Thu, 27 Jun 1996 07:53:20 -0700 (PDT)
Received: from croute.com (ishm2.croute.com [199.97.106.1])
          by freefall.freebsd.org (8.7.5/8.7.3) with SMTP id HAA03278
          for <questions@freebsd.org>; Thu, 27 Jun 1996 07:53:15 -0700 (PDT)
Received: from bldg1.croute.com by croute.com (4.1/SMI-4.1)
	id AA15169; Thu, 27 Jun 96 09:53:11 CDT
Received: from COMPUROUTE/SpoolDir by bldg1.croute.com (Mercury 1.13);
    Thu, 27 Jun 96 9:52:50 +600
Received: from SpoolDir by COMPUROUTE (Mercury 1.13); Thu, 27 Jun 96 9:52:22 +600
From: "Larry Dolinar" <LARRYD@bldg1.croute.com>
Organization:  CompuRoute, Inc.
To: questions@freebsd.org
Date:          Thu, 27 Jun 1996 09:52:14 +600 CDT
Mime-Version:  1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7BIT
Subject:       sendmail, routing, and port 25 dying slowly
X-Confirm-Reading-To: "Larry Dolinar" <LARRYD@bldg1.croute.com>
X-Pmrqc:       1
Priority: normal
X-Mailer: Pegasus Mail v3.22
Message-Id: <64A1FCB05AA@bldg1.croute.com>
Sender: owner-questions@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

Forgive me, but this doesn't apply entirely to FreeBSD (however I hope it 
may be the solution)...

I'm using a Sun Sparc 2 (SunOS 4.1.3) for a relay host and local DNS.  We 
also run 2 Novell networks.  Here's the dreaded diagram:

                         smtp
ISP /\/\ Ascend P50 -------------- nw1 ----------- nw2 --------
                     199.97.106.x         private        38.234.83.x/24

The Ascend has static routes to the private net (192.168.2.x) and the 
subnetted class A (38.234.83.x).  'smtp' is on the 199.97.106.x 
net, and likewise has static routes, both through 'nw1' (not its host 
name) for the "private" and "38" networks; the "private" route 
accomodates contact from 'nw2' for mail originating from their network 
(via Mercury/Pegasus).

These static routes became necessary because some VP type just had to run 
Netscape from the 38 network, a second "class C" that was given us by our 
ISP, long after the first class C (199) was acquired.  At least, it 
seemed to fix the problem.

There is a long-standing thing about Novell 3.12 fileservers and their
TCP/IP routing as it comes "out of the box".  If you use RIP, default
routes are pretty much screwed (determined by order of the bind
statements for the network cards).  If you turn it off, a different
approach (using their IPCONFIG) allows specifying static routes in a
GATEWAYS file.  There are other more esoteric approaches with NLSP and
OSPF I've yet to tackle.

This second approach (IPCONFIG) was used on 'nw1' and 'nw2' because the 
RIP approach doesn't really support netmasks (as it is, we had to add to
/etc/networks and /etc/netmasks on the relay host to get Sun's routes to
work).  ping, ftp, Netscape, and (we thought) mail were fine.

(audience shouts of: GET ON WITH IT!   8)

Now to the problem: incoming mail on 'smtp' degrades gradually
(apparently) as long as these static routes are installed on the
smartmailer.  Outgoing mail is fine, but port 25 response just gets more
and more sluggish (verifiable by trying to telnet into port 25 from
another host, even on the 199 network). 

Kill those routes to "private" and "38" and it wakes up again, pretty
quickly.  While all routes are installed, all parties can ping each
other, and the outside, without incident.

Perhaps this is a Sun problem with DNS, SMTP, and static routes to "non-
standard" networks all on the same box.  I'm considering resubscribing to 
sun-managers and posting this problem, though the traffic is pretty high 
and the tone borders on acidic.

Does any of this sound familiar (the problem, that is)?  I'm considering 
moving the SMTP functions to what will shortly be our FTP server (running 
2.1.0-R), and perhaps leaving the DNS where it is.  In reality both our 
primary and secondary name servers are with the ISP; what we're using 
locally seems more of a skeleton DNS.

What I'm hoping is that with this SunOS version being pretty old, FreeBSD 
doesn't have this kind of problem and will make a workable alternative.  
Nonetheless it's tough troubleshooting things "in production", as most of 
you well know.  For historical (and perhaps obvious) reasons, 'nw1' and 
'nw2' aren't going away.

Any insights provided are very welcome.  I left out configuration files 
to keep the bandwidth down; hopefully this doesn't cloud the description 
of the problem too much.  I'll be more than happy to provide them.

thanks for listening,
larry