[Mimedefang] Performance concerns

[Phil Eschallier]

Folks;

I got some help from this list a bit back (one of many to need to be told to
use tmpfs on Solaris for /var/spool/MIMEDefang).  Thank you!!

My larger MD + SA deployment has been running well.  I recently enabled the
filter_sender / filter_recipient / filter_relay processing, and performance
has really been hit hard -- even if I just have filter functions returning
continue.

I replicated the production environment on a local server.  I now have this
as the filter_sender() function on the test server:

# Filter by sender.
sub filter_sender {
    my($sender, $hostip, $hostname, $helo) = @_;
    my $snd = $sender;

    $snd =~ s/^<(.*)>$/$1/;
    $snd = lc($snd);

    if ($snd =~/\@buxtech\.com$/
                                                ) {
        md_syslog('info', "Sender: '$sender' ($snd): No more filtering");
        return('ACCEPT_AND_NO_MORE_FILTERING', 'OK');
    } else {
        md_syslog('info', "Sender: '$sender' ($snd): Filtering continues");
        return('CONTINUE', 'OK');
    }
}

just for testing / logging purposes.  Oh ... and my mimedefang-filter is
pretty much the generic one (from install).

What I've found is that for each message I send, filter_sender() is being
called twice (about 10 seconds apart).  Sendmail is configured to queue all
mail to mimic the larger environment.

I looked thru some of the archives and searched on Google; I can't find this
discussed.  It seems to me that my performance concerns could be helped by
only calling filter functions once.

Am I missing something obvious?  Is this because sendmail is queuing then
processing messages?  The log file states that the relay address is the same
for each duplicate run (my filter_relay() function also runs twice per
message, and simply logs info via md_syslog()) otherwise I suppose I could
skip the second check if the second relay address was local.

Thoughts?  Pointers?

... Phil