[comp.protocols.tcp-ip] Sendmail/SMTP/TCP mystery?

jaynes@med.umich.edu (William Jaynes) (02/25/91)

Someone felt that this problem may be a more general TCP protocol discrepancy
causing packets to be lost or ignored, so here is what I have posted elsewhere..

----------
Last week I noticed that messages to a particular host were being queued on
my mailhub.  The sendmail message  is

  (Deferred: Connection timed out during result wait with uv1.im.med.umich.edu)

These messages never get delivered.  All attempts fail until they time out and
are bounced.

The mailhub is a SunSLC.  The host in question is a Vax running some
Wollengong TCP-IP/SMTP etc. software.  Up until last week there was not a problem.
Both of us sys admins swear that we haven't changed anything. (of course)

Here are some facts:
    - A short message (less then about 1100 bytes) will be delivered.
      A longer message will not.
    - I forced my Suns (4.1 and 4.1.1) to skip the mailhub and mail direct
      and got the same results: longer messages were queued with the same
      deferred message.
    - I forced my Decs (Ultrix 4.0) to skip the mailhub and mail direct
      and did NOT get the same results.  Mail worked fine.
    - From my Suns, if I telnet to port 25 of the Vax and fake mail, I can
      send as big a message as I want.
    - From Suns that are not on my net (141.214) and which mail direct to the
      Vax, I can send as big a message as I want. (One is 4.0.3c)
    - This one Vax is the ONLY host with which there is a problem.
      Other Vaxen on the same ethernet don't have a problem. (I don't know
      if the other Vaxen use the same SMTP software.)
    - The data files are properly terminated with a 0x0a.

Using a Sniffer, I looked at packets during delivery attempts that failed.
The Suns simply aren't sending a <CR><LF>.<CR><LF> after the data is
sent.  Instead they send the data again and again until the connection
times out.  But they do send <CR><LF>.<CR><LF> on the short messages.

Why this one Vax?  Why the Suns and not the Decs?  What's going on?
-------

I have since seen that the SUNs and DECs are conifgured NOTRAILERS.
Doing an ifconfig on the VAX doesn't give a reference to trailers.

PIRARD%vm1.ulg.ac.be@CUNYVM.CUNY.EDU (Andr'e PIRARD) (02/27/91)

On Mon, 25 Feb 91 15:33:52 GMT William Jaynes said:
>(Deferred: Connection timed out during result wait with uv1.im.med.umich.edu)
>These messages never get delivered.  All attempts fail until they time out and
>are bounced.
>The mailhub is a SunSLC.  The host in question is a Vax running some
>Wollengong TCP-IP/SMTP etc. software.  Up until last week there was not a

A problem we've had seems just faintly related, but might help.
A Ultrix host was repeatedly trying to send to a Sun what appears to be the
end part of a mail item, until the 3-days limit expired.
[ub4b.cs.kuleuven.ac.be to georges.montefiore.ulg.ac.be]
19 lines from a dest queue file w/o control starting (probably spilled here):
           returns the network address and name of all the subscribers. If you
 do not wish
          your name to be available to others  in this fashion, just issue a
 "SET MACMAIL
-- etc..., end of quote --
On each and every attempt, the Sun's sendmail was producing a core dump.
I am in charge of none of those hosts, just routing in between. The problem
may have be started by a longer than usual outage of routers in test.
Yet.

Andr'e PIRARD             SEGI, Univ. de Li`ege
B26 - Sart Tilman         B-4000 Li`ege 1 (Belgium)
pirard@vm1.ulg.ac.be  or  PIRARD%BLIULG11.BITNET@CUNYVM.CUNY.EDU