SUMMARY: sendmail configuration for cluster

From: Jan-Willem de Bruijn (J.W.deBruijn@CTG.TUDelft.NL)
Date: Tue Oct 04 1994 - 19:21:31 CDT


Let me first repeat the problem as originally stated:

> We have eight machines, one of which functions as mailserver (it serves
> the /var/mail directory through NFS) and mailhost (mentioned in the NIS
> hosts map). (It is also the NIS master.)
> The NIS domain (from which figures out its own domain) is
>, the DNS domain is
> All machines have the file as, even the
> mailhost, because what we want is that all mail not for a local user
> (i.e. everything that has a Fully Qualified Domain Name as address) is
> sent to a mail relay host (not one of our Suns), that takes care of the
> mail domain (for reply address rewriting and such stuff).
> (The subsidiary machines have as relay host the mailhost, the mailhost
> itself uses an external machine as relay.)
> What happens is that the following lines are consistently wrong in
> outgoing mail:
> >From user@mailhost date
> Return-path: <user@mailhost>
> From: <user@mailhost> (fullname)
> Nowhere is the mail domain being added to the mailhost
> name (as it should, in my opinion).
> I've looked in the file, and in each case it is the $g
> macro that is involved. (Not that I know what to do with it :-).

After having a look at the suggested sendmail.8.6.9 distribution, which
also looked to be a daunting task, I decided to jump in at the deep end
and have a closer look at
I think I have it configured properly now. The problem of the missing
domain names in some of the header lines, was overcome by using the line


in the on the mailhost (instead of DMether as I previously
The reason for using the subsidiary file on the mailhost as well, was
to have it send all mail addressed to the symbolic mailaddress , to the mail-relay (an external machine where
the address translation is being performed), without having to tweak
the rewriting rules. But as everyone pointed out, using the is
better (only that file defines the ddn mailer), but the necessary thing
was adapting ruleset 6 to read:

R$*<@$-.$=m>$* $1<@$2.LOCAL>$4

so that only @hostname.$m is recognized as being LOCAL, but not @$m .

Thanks go to the following people for their suggestions and helpful

Clark Rudder <>
Claus Assman <>
Sam C. Nicholson <>

and especially

Stefan Rauthenberg <rauthenberg@HHI.DE>

Thank you, all of you.


Jan-Willem de Bruijn
Section Applied Geophysics
Dept. of Mining and Petroleum Engineering
Delft University of Technology

This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:09:11 CDT