[Mailman-Users] Outgoing mail failure

[Ed Wilts]

I'm experiencing a severe problem with an outgoing list.  After a fair
bit of testing on a small number of recipients, we gave it an address
list of 44322 addresses.  When we posted to it overnight, a few
unexpected things started happening, which I'll attempt to document
below.
1.  smtp file reports a lot of failures:
[root at mioc2 log]# grep 'Jul 16' smtp | grep -v 'Connection refused'
Jul 16 01:15:35 2003 (21028) smtp for 44322 recips, completed in 252.830 seconds
Jul 16 01:26:02 2003 (22215) smtp for 43485 recips, completed in 583.220 seconds
Jul 16 01:40:04 2003 (27387) smtp for 42745 recips, completed in 524.006 seconds
Jul 16 01:41:43 2003 (27387) smtp for 3 recips, completed in 3.295 seconds
Jul 16 01:43:03 2003 (27387) smtp for 3 recips, completed in 1.122 seconds
Jul 16 01:53:26 2003 (32732) smtp for 42245 recips, completed in 366.774 seconds
Jul 16 02:02:22 2003 (4857) smtp for 41485 recips, completed in 108.675 seconds
Jul 16 02:03:58 2003 (4857) smtp for 3 recips, completed in 1.694 seconds

The first attempt appeared to try and deliver to all the recipients, but
then it looks like it retried and retried, over and over again.  Ther
are over 47,000 'Connection refused' messages in the smtp file.

I see a corresponding number of entries in the post log.

2.  The smtp-failure file has a log of (ignore) messages in it, like so:
Jul 16 01:15:36 2003 (21028) -1 ewilts at ewilts.org (ignore)
There are currently 936796 ignore entires in smtp-failure

3.  When qrunner runs, the load average goes through the roof - 40-60
and higher, until it's killed.  It forks python processes like crazy.

We're running mailman 2.0.13 on Red Hat Linux 7.0 with a 2.2.24 kernel.
It's a dual P III/500 system with 256MB memory.
We're using sendmail-8.11.6-25.70

Since the smaller test cases (~400 users) worked fine, I'm assuming that
sendmail and mailman are properly configured and talking to each other
as they should.  I don't know what the problem is with this larger
batch.

The message is small - sendmail reports it as size=1581.
The sendmail queue is almost empty - it hasn't topped 200 since the
posting went out.  At 1am it was 1 request, and at 2am it was only 74
(mostly bounces I think).

Any ideas would be greatly appreciated.

-- 
Ed Wilts, Mounds View, MN, USA
mailto:ewilts at ewilts.org