[Mailman-Users] help with rather time-sensitive dilemma??

Barry A. Warsaw barry at digicool.com
Wed Jul 11 07:19:18 CEST 2001


>>>>> "A" == Amanda  <arandall at auntminnie.com> writes:

    A> I am this > < close to giving up on qmail as a viable MTA for
    A> Mailman...

Word of warning: I've never used qmail.
    
    A> That's it. Nothing, nada. There are no errors showing refused
    A> connections, timeouts, etc in the qmail logs (and these show up
    A> very nicely when mail doesn't send, at least under other
    A> [non-mailman] circumstances). The actual message never arrives
    A> in any subscriber's box. In fact, from what I can see of the
    A> above, the message is never being sent OUT to the list. Here's
    A> another clue - Mailman bounce logs show the following (at
    A> present, this email address is the admin!):

    | 15:17:01 (6301) Testlist: arandall at auntminnie.com - 2 more allowed
    | over -521998 secs
    | 15:37:03 (6420) Testlist: arandall at auntminnie.com - 2 more allowed
    | over -523201 secs
    | 15:37:03 (6420) Testlist: arandall at auntminnie.com - 1 more allowed
    | over -523201 secs
    | 15:37:04 (6420) Testlist: arandall at auntminnie.com - 0 more allowed
    | over -523201 secs

    A> Now, qmail can send mail to arandall at auntminnie.com without
    A> issue. I do it all the time... So why does Mailman think it's
    A> bouncing? What does the (6420) designate, and why are there
    A> three of them over a two-second period?

The 6420 is the pid of the process writing the log entry

    | Here's the post log:
    | 15:17:01 (6301) post to testlist from testlist-admin at tux.[foo.bar],
    | size=362, 1 failures
    | 15:37:03 (6420) post to testlist from testlist-admin at tux.[foo.bar],
    | size=296, 1 failures
    | 15:37:03 (6420) post to testlist from testlist-admin at tux.[foo.bar],
    | size=194, 1 failures
    | 15:37:04 (6420) post to testlist from testlist-admin at tux.[foo.bar],
    | size=204, 1 failures

    | Obviously these directly correlate to the bounce log, but no
    | indication of what's really going on here...
    | Here's the smtp log:
    | 15:17:01 (6301) smtp for 1 recips, completed in 0.071 secs
    | 15:37:03 (6420) smtp for 1 recips, completed in 0.487 secs
    | 15:37:03 (6420) smtp for 1 recips, completed in 0.050 secs
    | 15:37:04 (6420) smtp for 1 recips, completed in 0.050 secs

What these log entries are telling me is that SMTPDirect.py is getting
errors during the connection to your smtpd.  Since you don't see "All
recipients refused" in logs/smtp, I know you're not getting a
socket.error or an smtplib.SMTPException.  So either the .sendmail()
call is failing internally and returning some refused recipients, or
we're getting an SMTPRecipientsRefused exception. (see deliver() in
SMTPDirect.py and Python's smtplib module documentation for details).

The logs/smtp entries aren't indicating success, just completion of
the smtpd delivery attempt.  It's quite clear that qmail is returning
failure codes back to the client (in this case Mailman via Python's
smtplib).  What is strange is that what's coming back for some reason
includes arandall at auntminnie.com three times, because it's the same
process that's logging all three immediate failures (via the for loop
in process(), looping over the refused items).  That's odd.

JC gave some good advice for trying to figure out why qmail is
refusing messages from Mailman.  I'd add too other things:

- Try to use Python's smtplib module directly to mimic the same calls
  Mailman is making to see what kind of failure or return value you're
  getting.

- Try adding a syslog() call to SMTPDirect.py's process() function,
  inside the "if code >= 500 and code <> 552" clause to log exactly
  which error codes you're receiving.  Something like:

  syslog('debug', 'Received error code %d for recipient %s' % (code, recip))

  Then tail the logs/debug file to see which error codes your
  getting.  Then try to figure out why qmail is returning these error
  codes. ;)

Hope that helps,
-Barry




More information about the Mailman-Users mailing list