[Mailman-Users] Re: [EXIM] Probs with Exim/Mailman combo

[Harald Meland]

[Jerry Alexandratos]

> I'm having some problems when some of my users send to some mailing
> lists.  I'm at my wit's end and can't seem to do anything to fix the
> problem.  I'll include the error messages I'm getting.  I can't figure
> out if it's exim tripping up mailman or vice versa or what.
> 
> My setup is as follows:  FreeBSD -current, Exim 2.92, Mailman 1.0rc1.
> 
> >From exim, I get the following log entries:
> 
> 1999-06-09 16:44:55 10rp8d-000I0T-00 SMTP data timeout (message abandoned) on co
> nnection from perspectives.net [63.66.225.70]
> 1999-06-09 16:44:55 10rp8d-000I0S-00 SMTP data timeout (message abandoned) on co
> nnection from perspectives.net [63.66.225.70]
> 1999-06-09 16:44:55 10rp8c-000I0C-00 => |/opt/mailman/mail/wrapper post everyone
>  <Everyone at quansoo.com> D=mailman_aliases T=address_pipe
> 1999-06-09 16:44:55 10rp8c-000I0C-00 Completed
> 
> >From mailman, I get the following log entries:
> 
> Jun 09 16:47:00 1999 TrySMTPDelivery: To ['jerry at quansoo.com']:
> Jun 09 16:47:00 1999 TrySMTPDelivery:
> Mailman.pythonlib.smtplib.SMTPServer
> Disconnected  / Connection unexpectedly closed  (dequeued)

Mailman dequeues (i.e. does not try to re-deliver) messages that seems
to be the cause of "unexpected problems" with the MTA.  Currently,
"unexpected problems" are pretty much any problem at all short of not
being able to connect to the SMTP service.  I'll have a look at the
TrySMTPDelivery Mailman code to try making it be a bit more verbose as
to exactly what went wrong...

However, list archiving is done before the SMTP delivery -- so if your
list has archiving turned on, you will have a copy of the "offending"
message at the end of
/opt/mailman/archives/private/everyone.mbox/everyone.mbox.  You could
try injecting this into Exim by hand to see if that particular message
contains whatever is the cause of your problems.

BTW, I'm using Exim 2.12 on the Solaris 2.6 box that's running my
Mailman installation (although I'm now using a post-1.0rc1 CVS
version), and I haven't seen any problems like this one.
-- 
Harald