Re: [Mailman-Users] Analysis of reasons for bounce responses

On Thu, 2003-07-31 at 06:08, Richard Barrett wrote:

This is not my problem but one soemone approached me about off list.
I'll summarise their situation and I would appreciate any thoughts,
comments, sanity checking on my analysis.

Situation:

1. List admin's domain is hosted by a server run by a hosting company.

2. The server is (probably) running Linux

3. Mailman 2.1.2 is provided on the server with, I believe, CPanel

4. The server's MTA appears to be Exim 4.20 handling both incoming and
   outgoing messages to/from Mailman

5. Mailman list's basically work OK much of the time

6. Periodically messages out from Mailman are being bounced en masse
   leading to the subscriber accounts being disabled

7. The bounce response attached to the bounce action notice produced by
   Mailman when subscribers are disabled in these circumstances is from
   the server local MTA. The MTA lists what appears to be all the
   addresses to which the outgoing message was sent with a note against
   each saying "unrouteable mail domain maildomain.tld". The bounce
   response asserts that this is a permanent failure. The messages are
   each being sent to a number of different mail domains not just a single
   one

8. The list admin with the problem has no access to either the Mailman
   or MTA logs

9. The server administrator is supposedly claiming that these bounces
   are a result of "a bug in Mailman for which no patch is available"

My analysis is:

1. The problem is not obviously due to any bug in Mailman

2. The problem is more likely due to an IP routing problem which
   periodically afflicts the server and its ability to reach the MTAs for
   the mail domains addressed by the outgoing messages

3. The MTA is accepting the outgoing messages from Mailman and then
   finding it has problems of its own in delivering mail which are
   unrelated to Mailman as the origin of the messages

But I have to concede I know nothing about Exim 4.20. I know even less
about CPanel.

Any thoughts or anatomically feasible suggestions welcome.

I have seen this behavior a few times and in each case it was resource problem on the server. They need to check on the memory of the box and how much is being used during the bounces. I'm afraid it is only something that the sysadmin can really do.

I have also seen this problem when the /var directory was made too small. Could this be a problem with the users services being chrooted and them running out of quota space while processing large messages?

Good Luck - Jon Carnes