[Mailman-Users] Memory problems and stuck list

[Mark T.B. Carroll]

I am using mailman 2.1.3 with GNU licb6 2.3.2 and Python 2.3.2.

We have a couple of low-traffic lists with few subscribers. Things used to
work fine until recently. (I'm not sure what changed, unfortunately.) We
were using mailman version 2.0.11, but upgrading it to see if that made
the problem go away hasn't improved things.

First, incoming mail is now languishing in /var/lib/mailman/qfiles/in/.
I've tried to get it moved with things like "qrunner -v -o -r All" but to
no avail. (Maybe that's no surprise; I don't quite know what I'm doing.
(-:) If I could just get mail periodically shifted by hand, that'd be a
good start. /var/log/mailman/qrunner says it did do a qrunner iteration
for each Runner.

When I start mailman with "mailmanctl -s start" then a number of Python
processes (qrunners) get fairly busy and are around 4Mb apiece. The
machine slows to a crawl and the kernel starts killing them due to out of
memory errors. "killall -9 python" is how to make the machine happy again.
It only has around 32Mb RAM plus a little more than that in swap, which
isn't a lot but should be plenty for running small mailing lists.

We use exim but that's been fine and happy. It's been recording the piping
of incoming mail into "/var/lib/mailman/mail/mailman post foo..." which is
presumably how things end up in qfiles/in/ in the first place. The posts
appear in /var/log/mailman/post. None of the other mailman logfiles show
anything interesting.

Last night I added 128Mb of swap space and tried restarting mailman. This
morning the machine seemed to be pretty much dead, but with no clues in
the logs as to what went wrong. The other (mostly idle) services on the
machine seem to be running much as they should, still, after a reboot;
AFAICT it's only mailman that's unhappy.

Does any of the above sound familiar? Is there other diagnostic stuff I
should do? Unfortunately I don't know any Python.

-- Mark