[Mailman-Users] Strange errors

[Mark Sapiro]

Dan Szkola wrote:
>
>I am running a Solaris 10 box, with mailman-2.1.6rc4 and sendmail
>version 8.13.3. Python version is 2.4.1.
>
>I run sendmail in the following ways:
>
>
>A normal sendmail daemon listening on port 25:
>
>/usr/lib/sendmail -bd -q15m
>
>A persistent queue runner:
>
>/usr/lib/sendmail -qp1m -OPidFile=/var/run/sendmail-qrun.pid
>
>A sendmail with submit.cf config:
>
>/usr/lib/sendmail -Ac -q5m
>
>A sendmail for smtp-auth listening on port 587:
>
>/usr/lib/sendmail -bd -C/etc/mail/sendmail-auth.cf -q15m
>
>
>Anyway, the problem we are having is this:
>
>  After running for several days with no errors, we suddenly start
>  seeing this error:
>
>>>> Mail Delivery Subsystem <MAILER-DAEMON at xxxxxxx.xxx.xxx.xxx>
>10/17/2005 9:18:52 AM >>>
>The original message was received at Mon, 17 Oct 2005 09:18:21 -0500
>(CDT)
>from xxxxxxxx.xxx.xxx.xxx [131.156.xxx.xxx]
>
>   ----- The following addresses had permanent fatal errors -----
>"|/usr/local/mailman/mail/mailman post testlist"
>    (reason: 1)
>    (expanded from: <testlist at xxxxx.xxx.xxx>)
>
>   ----- Transcript of session follows -----
>Traceback (most recent call last):
>  File "/usr/local/mailman/scripts/post", line 35, in ?
>    from Mailman.Queue.sbcache import get_switchboard
>  File "/usr/local/mailman/Mailman/Queue/sbcache.py", line 19, in ?
>    from Mailman.Queue.Switchboard import Switchboard
>  File "/usr/local/mailman/Mailman/Queue/Switchboard.py", line 47, in
>?
>    from Mailman.Logging.Syslog import syslog
>  File "/usr/local/mailman/Mailman/Logging/Syslog.py", line 22, in ?
>    from Mailman.Logging.StampedLogger import StampedLogger
>  File "/usr/local/mailman/Mailman/Logging/StampedLogger.py", line 20,
>in ?
>    from Mailman.Logging.Logger import Logger
>  File "/usr/local/mailman/Mailman/Logging/Logger.py", line 25, in ?
>    from Mailman.Logging.Utils import _logexc
>  File "/usr/local/mailman/Mailman/Logging/Utils.py", line 18, in ?
>    import traceback
>ImportError: No module named traceback
>554 5.3.0 unknown mailer error 1
>
>Has anyone seen this? Should I file a bug on this or send it along
>to the developers list?
>
>Restarting sendmail seems to get rid of the problem for a day or two.
>All sendmail configs have the mailman alias file included and we run
>newaliases to update it on every new list creation.


This is curious indeed for at least two reasons.

It is not a sendmail alias problem, nor does it seem on the face to be
a sendmail problem at all. The post is received by sendmail and piped
to the wrapper with the appropriate arguments. The wrapper invokes the
post script as it should.

The post script then does some imports one of which is
    from Mailman.Queue.sbcache import get_switchboard

sbcache does
    from Mailman.Queue.Switchboard import Switchboard

and so on until Mailman.Logging.Utils does
    import traceback

which results in ImportError: No module named traceback

Now this is really strange because traceback is a Python library module
and this chain of imports leading to 'import traceback' occurs with
every post, so why does it fail now and why does restarting sendmail
fix it?

What happens if you restart Mailman (bin/mailmanctl restart) without
restarting sendmail? Does that fix it? Or do all the qrunners die with
the same ImportError?

I have no reason other than superstition for the following suggestion,
but try

SMTP_MAX_SESSIONS_PER_CONNECTION = 1

in mm_cfg.py and see if that helps. This will cause Mailman to close
the SMTP connection to sendmail after each transaction. This might
avoid the problem, but since I have no idea what the problem is, I
have no idea if this will help.

-- 
Mark Sapiro <msapiro at value.net>       The highway is for gamblers,
San Francisco Bay Area, California    better use your sense - B. Dylan