Hello Group,

A bit of background: I use a server -let's call this one SERVER1-
running Scalix 11.3 to manage all my domains and mailboxes which uses
sendmail to relay outbound mails to another server of mine -let's call
this one SERVER2- as a smarthost. SERVER2 runs sendmail as well,
resides in the same network and uses an external server (my ISP's one)
as a smarthost for all outbound mails.

On SERVER2 amavis in the dual-sendmail configuration is running as
well for anti-spam/anti-virus purposes.
So far so good, this was the easy part. SERVER2 is the entry point for
all mails (also for fetchmail, running on a third server, the
gateway), does its checks and if the mail is clean it will be
delivered.

Scenario:
fetchmail@gateway --> sm-daemon1@SERVER2 --> amavis@SERVER2 --> sm-
daemon2@SERVER2 --> mailboxes@SERVER1

In SERVER2's mailertable there are entries for the domains hosted on
SERVER1 in order to deliver the mails to the respective mailboxes on
SERVER1. This is actually working OK for all domains which are
completely hosted on SERVER1.

But -as always- there is an exception: there is a domain "partially"
hosted on SERVER1 -let's call it difficult.domain- because one user in
my network wants to keep this other e-mail address of his for
historical reasons, all other users of this domain are hosted
somewhere else in the outside world.

Now, here is my problem: If I send a mail to a user @difficult.domain
other than the one locally hosted, the mail keeps bouncing back.
Scenario:
1. SERVER1 delivers it via sendmail to SERVER2 (the smarthost) because
the address is not managed locally
2. SERVER2 delivers it back to SERVER1 because its mailertable says
that it has to do so

Now my question to the group: is there any configuration (a
mailertable or some other entry) which would allow SERVER2 to deliver
the mails @difficult.domain only for one user to SERVER1 and for the
rest to the smtp server of my ISP? Is this possible at all?

Any would be very much appreciated since I have tried all I could
but I am still stuck ...


Thanks a lot in advance to you all,
LP