[Mailman-Users] Stuck outgoing message, MX, DNS

[Richard Martin]

I am posting this solution back to the list so the title will be searchable.

In addition to changing the sendmail.mc file, one can hand edit the sendmail.cf
file and change the Timeout.rcpt setting.  In many default configurations, this
line is merely commented out.

Glen Foster wrote:

> It is not just bad MX records, many DNS irregularities are responsible
> for this behaviour.  It is because sendmail considers unavailable
> domain resolutions potential SPAM.
>
> The sendmail configuration parameter
>
> define(`confTO_RCPT',`60m')dnl
>
> governs how long sendmail will try to resolve a RCPT.  It can be
> significantly shortened without too much difficulty.  However, setting
> it too short will result in non-delivery to legitimate addresses that
> are behind slow, busy, or intermittent links.  See the sendmail docs.
>
> Alternatively, you can set sendmail to queue incoming messages rather
> than trying immediate delivery by putting this in your .mc file:
>
> define(`confDELIVERY_MODE',`defer')dnl
>
> Then sendmail will accept and try to deliver these, and bounce them
> back if undelivered after the usual delay.  The bounces are then
> processed by Mailman to clean the list.
>
> Note that circumventing the anti-SPAM measures may have undesireable
> effects on publically-available MTAs.  However, in practice, I haven't
> run into serious problems doing it.
>
> If you do this, there are other operational tunings that you may want
> to investigate including limiting the number of daemons sendmail is
> running simultaneously.
>
> Of course, after customizing your .mc file you have to re-generate
> your sendmail.cf and restart sendmail.
>
> If you don't want to change your sendmail configuration, the only
> alternative is to manually delete the hanging messages from
> ~mailman/qfiles.  Each message is represented by two files, the
> message itself (*.msg) and a control file (*.db).  The
> ~mailman/bin/dumpdb program will show the contents of the control
> file.  You should turn off qrunner while you are manipulating the
> queue files.
>
> Finally, if you do nothing, eventually the qfiles will time out and be
> deleted.  There is a parameter in ~mailman/Mailman/Defaults.py that
> controls how long this is.  But this will make delivery logy for
> legitimate recipients and bounce processing is not performed.  Other
> parameters can be adjusted (e.g. shortening the time qrunner has to
> deliver) as well.
>
> I've tried all these, you get the most mileage out of configuring
> sendmail to defer delivery.
>
> Good luck,
> Glen Foster <gfoster at gfoster.com>
>
> Richard Martin writes:
>  > I am refining the definition of the problem 'Can not check MX records'
>  >
>  > Apparently there is a bad address in the users list, bad in the respect
>  > that the destination server has a broken MX record. consequently, the
>  > posted message is patiently waiting in queue, hitting a brick wall every
>  > minute.
>  >
>  > How can I flush out or delete messages queued by Mailman?  I have looked
>  > thorough the directories and archives, but I can't seem to locate much
>  > on this one.
>  >
>  > --
>  > Richard Martin
>  > OriGen, inc.
>  > Austin, TX 78703

--
Richard Martin
OriGen, inc.
Austin, TX 78703