Message Doubles

[Stephen Swaney]

> -----Original Message-----
> From: mailscanner-bounces at lists.mailscanner.info [mailto:mailscanner-
> bounces at lists.mailscanner.info] On Behalf Of Matt Kettler
> Sent: Monday, March 27, 2006 10:01 PM
> To: MailScanner discussion
> Subject: Re: Message Doubles
> 
> spart cus wrote:
> > Hi guys,
> > I know i've read this problem before.Though im not using the MS yet. But
> > know im currently experiencing this with some of my clients. How can i
> > check this out.
> 
> Start off by comparing Message-ID: headers.. If the message is a dupe
> occurring
> at the MTA layer, they should be the same. If they're different, some
> piece of
> software that thinks it's at the client level re-generated the message
> (ie: a
> auto-forwarder rule in a client).
> 
> Follow up with comparing the path in the Received: headers. Do they differ
> at
> all? If so, one of the servers involved in the difference is responsible.
> --

I believe this may be happening at two of our sites. One runs:

	SuSE 8.0
	MailScanner version 4.51.6
	sendmail-8.12.6-210
	Lock Type = flock

The other runs:

	CentOS release 3.6
	MailScanner version 4.51.6
	sendmail-8.13.1-3.RHEL4.3
	Lock Type = posix

Log messages indicate that sendmail attempts to deliver the message over and
over but apparently never receives and acknowledgement of delivery from the
client. A typical log entry:

Mar 27 12:03:08 smzz sendmail[20642]: k2OL5BPX014779: to=<rzzz at zzzzzzz.com>,
delay=2+19:57:57, xdelay=00:10:23, mailer=smtp, pri=23836807,
relay=[66.1.1.5] [66.1.1.5], dsn=4.0.0, stat=Deferred: 451 Timeout waiting
for client input

This message keeps repeating until the message is manually deleted from the
outbound mail queue. The recipient gets a copy of the message every time
delivery is attempted.

I have copies of the qf and df files for two of these messages. A quick scan
of the messages does not reveal anything obviously amiss to me. 

I'm not yet 100% convinced it's a MailScanner problem. Other software /
hardware could be intercepting the receiver's acknowledgement. At the SuSE
site the problem hasn't occurred since we turned off the PIX "sendmail
helper" function but since the problem occurs rarely, it just might be too
early to tell. 

At the CentOS site, turning off the PIX  at the receiving end (if in fact
this was done correctly) did not solve the problem. Still it's interesting
that PIX was involved in both receiving sites.

Still this problem did start about the same time the TNEF code was added to
MailScanner. We've been trouble shooting it on our own as we did not
originally think it was caused by MailScanner. Because of this new list
thread and the differences in sendmail versions and OS versions at the
troubled sites I'm beginning to wonder.

If anyone would like copies of these "repeating" messages please email me
off list.

Steve

Stephen Swaney
Fort Systems Ltd.
stephen.swaney at fsl.com
www.fsl.com