[Mailman-Users] Odd smtp-error

Mark Sapiro msapiro at value.net
Sun Oct 24 04:47:05 CEST 2004

Rich West wrote:

>Even more info.. the problem is definitely unpredictable..


>Here's some tests I performed with a newly created test mailing list.
>Test 1
>o Added one person, me, to the list.
>o Sent a message to the list.
>o Received the posting.  No errors or anything were reported in 
>sendmail's logs or mailman's logs.
>o Added one other person to the list, again me, but another local 
>address and its own mailbox.
>o Sent a message to the list.
>o In the sendmail logs, I get (just do the normal replacements for the 
>words "at" and "dot"):
>Oct 23 21:26:08 cranium sm-mta[13718]: i9O1Q8EF013718: <rich.west at 
>wesmo dot com>... User unknown
>Oct 23 21:26:08 cranium sm-mta[13718]: i9O1Q8EF013718: 
>from=<test-bounces+rich dot west=wesmo dot com at lists dot wesmo dot com>, 
>size=1421, class=0, nrcpts=0, proto=ESMTP, daemon=MTA, relay=localhost 

The above envelope from is VERP like.

>Oct 23 21:26:08 cranium sm-mta[13720]: i9O1Q8EE013718: to=<admin at 
>wesmo dot com>, delay=00:00:00, xdelay=00:00:00, mailer=local, 
>pri=85562, dsn=2.0.0, stat=Sent
>o After analyzing, it's important to note the order of the accounts.. 
>the SECOND email address subscribed was the one that succeeded.  The 
>first one, which worked in the first test, was the one that failed.  It 
>made me think that it was acting as if there is no comma in the "to" line...

To: line is irrelevant as we're looking at the SMTP session.

>Wait a sec....  I remember seeing something in the config for Mailman 
>that said something about limits on inbound SMTP connections.
>After changing that value to a "1", compiling the Defaults.py file with 
>the command :
>python -O -c "import compileall; 
>compileall.compile_dir('/var/mailman/Mailman', ddir='/var/mailman/Mailman')"
>I sent another message under the "test2" configuration as above.  That 
>worked just fine.  Interesting.

Now you're on to something.

>I went a did another test by adding about 10 addresses (all of which 
>were generating errors previously), and submitting another test email.  
>Quite suddenly that seemed to work.  In fact, I've been testing it for 
>about 45 minutes (yes, this email took a while to write, but I figured I 
>would send it in for completeness of the original request for help) and 
>I have not had any problems..
>Weird.. I never had to alter that value before...  Will that have a 
>negative impact on things in any way?!

Yes, but not too much. The VERP like address above is used for enhanced
bounce detection and is controlled by VERP_DELIVERY_INTERVAL and other

If every message is sent from a VERP like address, each recipient must
be sent in a separate transaction because each has a different
envelope from. The SMTP session looks like (replies ommitted for

MAIL FROM: envelope from including VERP like recipient 1 address
RCPT TO: recipient 1
 message text
MAIL FROM: envelope from including VERP like recipient 2 address
RCPT TO: recipient 2
 message text
repeat for each additional recipient

This is the big performance hit with VERP like envelope from because
the message text must be repeated for each recipient as opposed to
non-personalized, non-VERP which may have just one MAIL FROM:, a bunch
of RCPT TO: and one DATA.

In your case, setting SMTP_MAX_SESSIONS_PER_CONNECTION = 1 takes the
above SMTP session with multiple transactions and puts QUIT, get
disconnected, reconnect, EHLO between each message. This adds some
impact, but not alot compared to the impact of doing the VERP like
addressing at all.

Note that I am using "session" and "transaction" per RFC 2821. in that
terminology, the option would be more properly called

It seems like there is some problem with your MTA accepting multiple
transactions per session or possibly with the way Mailman sends them.

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

More information about the Mailman-Users mailing list