[Mailman-Users] Mailman getting stuck while trying to send a message

--On Sunday, April 24, 2005 7:28 PM +0100 Remi Delon wrote:

BAsically with *any* mailing list software it needs to be configured to
just take the mail from the mailing list app, and ask questions later.
Same rule goes for LSoft ListServ, majordomo, etc. The problem you're
running into could be almost anything but i'd bet it has something to do
with bad addresses in the mailing list rather than your mailman version.
Find out what the address is, go to the list admin interface and using your
site pass logging and unsubscribe the junk address.

Thanks for the insight Michael... If a junk address is indeed all it
takes to cause Mailman to hang then it's quite disapointing ...
How would I go about finding that junk address (there are more than 200
in that list) ? Isn't it true that all addresses must have been valid at
some point ? (otherwise people wouldn't have been able to confirm their
subscription). That makes it hard to see which ones are junk ...

Thanks,

Remi.

Remi> Thanks for the insight Michael... If a junk address is
Remi> indeed all it takes to cause Mailman to hang then it's quite
Remi> disapointing ...

It's not. I've never seen Mailman hang like that in more than 4
years, with versions 2.0.13 and 2.1.5.

Remi> How would I go about finding that junk address (there are
Remi> more than 200 in that list) ?

It's probably the one listed in smtp-failure, see below.

Remi> Isn't it true that all addresses must have been valid at
Remi> some point ? (otherwise people wouldn't have been able to
Remi> confirm their subscription).

True. But that's not the kind of "junk" that causes problems for
Mailman; Mailman will disable delivery or unsubscribe addresses that
become invalid. The "junk" that causes Mailman to blow up are
syntactically invalid (eg, contain illegal characters) and must be
introduced by the list admin using "mass subscribe" or a command line
tool. If you've never done that, ie, all your users subscribed
themselves, then you will have no problematic junk addresses, just
undeliverable ones that will be automatically weeded out.

In your earlier message you provided this log:

smtp-failure:Apr 24 07:39:11 2005 (26501) delivery to
<removed-email-address> failed with code -1: please run connect() first

While it's possible that there's a connect() function in Mailman that
Mailman is failing to run, that seems very unlikely ... we would have
heard about it by now. There is a connect() function in many OSes,
for the "socket" interface layered over raw Internet connections. It
seems quite possible to me that your OS or network is hosed, resulting
socket connections appearing to be internally broken, and Mailman
being unable to connect to a remote host for long periods of time. In
that case, it might very well hang assuming that it should keep trying
until the network admin fixes whatever the problem is.

As far as I can tell, Mac OS X sometimes wedges itself in this way.
An SSH connection will simply go away ... until I connect to the same
host in a separate connection. Then the shell will come back up, the
X tunnel will start working again, etc, etc.

Remi> I'm currently running version 2.1.4. Could this be fixed by
Remi> upgrading to the latest version ?

Waiting a bit and upgrading to 2.1.6 (due out real soon) might (or
might not) be a good idea.

However, if my guess that the smtp-failure message is coming from the
OS is correct, upgrading Mailman is not very likely to fix it.

Forwarding to mailman-users, I can't do anything about this....
Remi> Actually, you have [heard about it] ...
Remi> http://mail.python.org/pipermail/mailman-users/2004-October/040344.html

Remi> Looks like I'm not the only one with that problem ...

I said "maybe it's a system issue?" and Remi said:

Remi> Hmm ... Seems a bit unlikely to me ... This is a production
Remi> machine running linux RedHat with sendmail as the SMTP
Remi> server and there are about 100 people using the machine and
Remi> sendmail on it all the time ...

So is the other guy. That's not enough to apply the law of large
numbers, but I wouldn't ignore it, either.

Are you using a RedHat-supplied Python? Did you build Mailman yourself?