[Mailman-Users] partial (silent) non-delivery

[David Lee]

On Mon, 5 Jun 2006, Mark Sapiro wrote:

> David Lee wrote:
> >
> >System:
> >   Fedora Core 5
> >   Mailman 2.1.8 (RPM: mailman-2.1.8-0.FC5.1)
> >   sendmail 8.13.6 (RPM: sendmail-8.13.6-0.FC5.1)
> >
> >
> >On the test list for our staff (89 members), some of the emails are
> >failing, while others succeed.  On a failure email:
> >1. it only reaches 59 (out of 89) recipients;
>
>
> What is in Mailman's 'smtp', 'smtp-failure' and 'bounce' logs?
> Specifically, are there any bounces recorded in the bounce log, (Is
> bounce processing turned on for the list?) and how many recipients are
> indicated in the smtp log entry?

Thanks for the reply, Mark.  Appreciated.

smtp:  All cases:
     <...> smtp to ourlist for 89 recips, completed in ...

smtp-failure:  Nothing

bounce:  No error indications.

"bounce processing" webpage indicates "Yes" for all the yes/no tickboxes.




> >2. it loses the hidden headers ("Received:" etc.) prior to the Mailman
> >   expansion;
>
>
> Mailman doesn't remove any Received: headers.

On these duff emails, the headers prior to Mailman are absent.  Something
has removed them.  I'm speculating (and it is mere speculation without
evidence!) that a badly formatted incoming (or mismatched content;
8-bit-isms, etc.) email might be exposing some sort of lurking problem in
Mailman.


> >3. Instead of:
> >      From: a.n.other at our.site
> >      Sender: ourlist-bounces at our.site
> >   we see just:
> >      From: ourlist-bounces at our.site
> >   (no sender; "From" replaced by expected "Sender" contents).
>
>
> Are these the actual headers in the raw message or just what an MUA
> shows? If the latter, see
> <http://www.python.org/cgi-bin/faqw-mm.py?req=show&file=faq02.003.htp>.
> If the former, I suspect Exchange.

Although the department is Exchange based (don't ask!) I'm personally
using good old trusty Pine to read this information from the copy of the
email that comes to me (fortunately I'm one of the 59 who receive it).
Yes, Outlook and OWA are (in general) conflating the "Sender:" and "From:"
so I'm painfully aware of that presentation issue.


> >4. The sendmail log shows one of the names to get truncated (from
> >   "a.b.longname at our.site" to "a.b.lon at our.site".
>
>
> If the member address on the list is "a.b.longname at our.site", mailman
> won't change it. (Unless maybe it's the one with the apostrophe, but
> even then, I'd look outside Mailman first.)

Again, this was just an observation.  Something, somewhere is going wrong,
and this is a _set_ of symptoms on such emails.


> >5. The mailman log files (bounce,post,smtp,vette) don't show anything
> >   unusual at all.
>
>
> See item 1 above - what is the number of recipiients in the smtp log
> entry?

89 (everyone).  But the sendmail log says 59 (on the bad emails; 89 on the
good).


> >6. In the body of one of the emails I noticed that what the user had
> >   typed as "===========" (a set of equals signs as a visual underliner)
> >   appeared as "=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D...", as though
> >   some MIME misunderstanding had occured.
>
>
> That is a raw rendering of a quoted-printable encoding. If you are
> looking at a raw mesage and it has a "Content-Transfer-Encoding:
> quoted-printable" header, this is normal. Otherwise, the encoding is
> getting lost somewhere. It's possible, but unlikely that Mailman did
> it. Bugs in this area should be fixed in 2.1.8.
>
>
> >A thought: The emails are emanating from Exchange 2003, Service Pack 2.
> >Might Exchange be generating something non-conformant (its bug) which
> >Mailman could then be tripping over (Mailman problem)?
>
>
> Exchange can screw up anything, but once the message gets to Mailman,
> if it is sent at all, it should be sent to all eligible members. Are
> all 89 members non-digest and delivery enabled?

The list is set up automatically with everyone identical: a typical
entry is ticked for "mods", "nodupes" and "plain".


> Are there or were there ever Topics defined for the list? (I really
> want to know the recipient count from the 'smtp' log.)

"Topics"?  I'm new to Mailman (but with several years email experience).
Guide me!


>
> >Incidentally, one of the mail names on the list contains an apostrophe of
> >the form: Someone.O'Surname at our.site .  We have seen this sort of thing
> >cause trouble very occasionally on other email systems.  But if that
> >really were relevant, then I would expect it to be consistent for all the
> >mails to the list, not just some of them.
>
>
> If this is an issue, it would only affect the one user unless sendmail
> doesn't like it somehow and drops it and either all preceeding or
> following RCPT addresses in that transaction.

The list membership is basically static.  The apostrophe person has been
here many years and her email has been fine.


>
> I just reread the initial 'fails sometimes' part which I sort of
> ignored. This is more puzzling, but most of what I ask above is still
> relevant. Could there be some kind of anti-spam content filtering
> affecting only the 30 missing? Are the failures from different posters
> than the successes? Again, on the failed posts, what is the recipient
> count from the 'smtp' log - what does sendmail's maillog say?

Yes it is that "sometimes" which is baffling.  An email to the list failed
(as described, with that set of symptoms) this morning.  A repost
(admittedly slightly edited)  to appease those who would complain of
double receipt worked (all symptoms gone).

Shall I zip up the logs and email them to you off-list?  (I've had a good
look through them, but you may still spot something I've missed.)

Are there any debug/log/trace things/flags/switches that I could enable?



-- 

:  David Lee                                I.T. Service          :
:  Senior Systems Programmer                Computer Centre       :
:                                           Durham University     :
:  http://www.dur.ac.uk/t.d.lee/            South Road            :
:                                           Durham DH1 3LE        :
:  Phone: +44 191 334 2752                  U.K.                  :