I'm getting the ol'

553 host config error: mail loops back to me (MX problem?)
554 username [at] domain.com... Local configuration error

I've looked at the Sendmail FAQ and I don't believe the solution fits
my situation.

The way we have our e-mail environment is as follows:
- An Exchange server and UNIX server are sharing an SMTP namespace
during the mirgration phase
- The UNIX server is authoritative for the domain while Exchange is
not
- MX record points to the Exchange server
- Any external mail that is sent to the Exchange server that is
unresolvable is forwarded to the UNIX server to either deliver it
locally or generate an NDR

Due to the way we are migrating users, we add entries into /etc/
aliases so that users who are still on the UNIX server can send mail
to users who are on the Exchange server.

The problem we are having is that the UNIX server does not send mail
to the Exchange server.

When doing a "sendmail -v username [at] host.domain.com" this is the output
given (host.domain.com being the Exchange server):
---
username [at] host.domain.com... Connecting to host via ether...
220 domain.com ***********************************************
>>> HELO domain.com
250 domain.com Hello [ip.address.of.unix.server]
>>> QUIT
221 2.0.0 domain.com Service closing transmission channel
postmaster... aliased to root
username,root... Connecting to local...
username,root... Sent
Postmaster... aliased to root
root... Connecting to local...
root... Sent
---

It seems from the above output, the UNIX server makes a connection to
Exchange, Exchange acknowledges, then the UNIX server immediately
responds with a QUIT killing the connection. No data is sent for
delivery.

This setup was working fine up until last week. I can use sendmail -v
on our other UNIX machines to the same e-mail address connecting to
Exchange with no problem. Also, when doing sendmail -v to other
external servers, there are no problems.

I am not the UNIX sysadmin, but I am trying to convince him that it is
not our Exchange server that is broke. He believes his server is not
misconfigured. I'd think there has to be something done to
sendmail.cf or whatever to say, "It's okay to send to this server even
though it accepts mail for the same domain I'm responsible for."

Any ideas?

In article <1189542356.838784.53300 [at] v23g2000prn.googlegroups.com>,
walter.wade.jr [at] gmail.com wrote:

> I'm getting the ol'
>
> 553 host config error: mail loops back to me (MX problem?)
> 554 username [at] domain.com... Local configuration error
>
> I've looked at the Sendmail FAQ and I don't believe the solution fits
> my situation.
>
> The way we have our e-mail environment is as follows:
> - An Exchange server and UNIX server are sharing an SMTP namespace
> during the mirgration phase
> - The UNIX server is authoritative for the domain while Exchange is
> not
> - MX record points to the Exchange server
> - Any external mail that is sent to the Exchange server that is
> unresolvable is forwarded to the UNIX server to either deliver it
> locally or generate an NDR

That's bad on two counts:

1. Exposing Exchange to the Internet is a bit like playing footsie in a
public bathroom.

2. Putting an extra SMTP hop between your final delivery system and the
Internet means that unless you are taking special steps to avoid it, you
will be generating NDR's for forged spam and spewing them back at
innocent forgery victims who really have no need to get them.


> Due to the way we are migrating users, we add entries into /etc/
> aliases so that users who are still on the UNIX server can send mail
> to users who are on the Exchange server.
>
> The problem we are having is that the UNIX server does not send mail
> to the Exchange server.
>
> When doing a "sendmail -v username [at] host.domain.com" this is the output
> given (host.domain.com being the Exchange server):
> ---
> username [at] host.domain.com... Connecting to host via ether...
> 220 domain.com ***********************************************
> >>> HELO domain.com
> 250 domain.com Hello [ip.address.of.unix.server]
> >>> QUIT
> 221 2.0.0 domain.com Service closing transmission channel
> postmaster... aliased to root
> username,root... Connecting to local...
> username,root... Sent
> Postmaster... aliased to root
> root... Connecting to local...
> root... Sent
> ---
>
> It seems from the above output, the UNIX server makes a connection to
> Exchange, Exchange acknowledges, then the UNIX server immediately
> responds with a QUIT killing the connection. No data is sent for
> delivery.
>
> This setup was working fine up until last week.

That's surprising. What changed?

> I can use sendmail -v
> on our other UNIX machines to the same e-mail address connecting to
> Exchange with no problem. Also, when doing sendmail -v to other
> external servers, there are no problems.
>
> I am not the UNIX sysadmin, but I am trying to convince him that it is
> not our Exchange server that is broke. He believes his server is not
> misconfigured. I'd think there has to be something done to
> sendmail.cf or whatever to say, "It's okay to send to this server even
> though it accepts mail for the same domain I'm responsible for."
>
> Any ideas?

Don't make Sendmail try to send mail to another machine that calls
itself by the same name. One of the protections from mail loops in
Sendmail is that when it relays mail it looks at the greeting banner and
recognizes when the receiving machine is calling itself by the name
Sendmail thinks of as its own.

Where you want to fix this for your environment is dependent on local
details, but given what you've written I THINK the right choice would be
to give your Exchange host its own proper hostname under your domain
rather than let it think of itself (and greet the world) as just your
domain name (a bad practice for any domain that has multiple hosts...)

--
Now where did I hide that website...

walter.wade.jr [at] gmail.com writes in comp.mail.sendmail:

> I'm getting the ol'
>
> 553 host config error: mail loops back to me (MX problem?)
> 554 username [at] domain.com... Local configuration error
>
> I've looked at the Sendmail FAQ and I don't believe the solution fits
> my situation.

> The problem we are having is that the UNIX server does not send mail
> to the Exchange server.
>
> When doing a "sendmail -v username [at] host.domain.com" this is the output
> given (host.domain.com being the Exchange server):
> ---
> username [at] host.domain.com... Connecting to host via ether...
> 220 domain.com ***********************************************
^^^^^^^^^^ This should be hostname of exchange box
> >>> HELO domain.com
^^^^^^^^^^ This should be hostname of sendmail box

> 250 domain.com Hello [ip.address.of.unix.server]
^^^^^^^^^^ This should be hostname of exchange box

> >>> QUIT
> 221 2.0.0 domain.com Service closing transmission channel
> postmaster... aliased to root
> username,root... Connecting to local...
> username,root... Sent
> Postmaster... aliased to root
> root... Connecting to local...
> root... Sent
> ---

> Any ideas?

Be sure that both exchange and and Sendmail box
is not called as 'domain.com'.

As far sendmail is considered, it's name is 'domain.com'.
And to where it is connected hostname is also 'domain.com'.

Do not use 'domain.com' as hostname at all.

IMHO both end are misconfigured.


/ Kari Hurtta

On Sep 12, 12:10 am, Kari hurtta <hurtta
+comp.mail.sendm... [at] siilo.fmi.fi> wrote:
> walter.wade... [at] gmail.com writes in comp.mail.sendmail:
>
> > I'm getting the ol'
>
> > 553 host config error: mail loops back to me (MX problem?)
> > 554 usern... [at] domain.com... Local configuration error
>
> > I've looked at the Sendmail FAQ and I don't believe the solution fits
> > my situation.
> > The problem we are having is that the UNIX server does not send mail
> > to the Exchange server.
>
> > When doing a "sendmail -v usern... [at] host.domain.com" this is the output
> > given (host.domain.com being the Exchange server):
> > ---
> > usern... [at] host.domain.com... Connecting to host via ether...
> > 220 domain.com ***********************************************
>
> ^^^^^^^^^^ This should be hostname of exchange box> >>> HELO domain.com
>
> ^^^^^^^^^^ This should be hostname of sendmail box
>
> > 250 domain.com Hello [ip.address.of.unix.server]
>
> ^^^^^^^^^^ This should be hostname of exchange box
>
> > >>> QUIT
> > 221 2.0.0 domain.com Service closing transmission channel
> > postmaster... aliased to root
> > username,root... Connecting to local...
> > username,root... Sent
> > Postmaster... aliased to root
> > root... Connecting to local...
> > root... Sent
> > ---
> > Any ideas?
>
> Be sure that both exchange and and Sendmail box
> is not called as 'domain.com'.
>
> As far sendmail is considered, it's name is 'domain.com'.
> And to where it is connected hostname is also 'domain.com'.
>
> Do not use 'domain.com' as hostname at all.
>
> IMHO both end are misconfigured.
>
> / Kari Hurtta

Thank you for your response.

The Exchange server does have host.domain.com specified and not just
domain.com. Sorry about that.

However, when I telnet into the UNIX server over port 25, in the SMTP
banner, I only get domain.com and not host.domain.com.

How does one change that in sendmail? I'd like to know so that I can
tell the UNIX sysadmin to make it as easy as possible on him.

Thanks again,
wwj

walter.wade.jr [at] gmail.com writes in comp.mail.sendmail:

> Thank you for your response.
>
> The Exchange server does have host.domain.com specified and not just
> domain.com. Sorry about that.

Sendmail "mail loops back to me" if sendmail ends to server which have
same name than which is it's hostname when it does SMTP connection.

So something on given data is not real.


> However, when I telnet into the UNIX server over port 25, in the SMTP
> banner, I only get domain.com and not host.domain.com.
>
> How does one change that in sendmail? I'd like to know so that I can
> tell the UNIX sysadmin to make it as easy as possible on him.
>
> Thanks again,
> wwj

Normally sendmail host this from system.

First hostname() system call, and then result is canonifed
by gethostbyname(). Usually that consults /etc/hosts and
DNS.

Probably there is domain.com on /etc/hosts (on same line
than is name returned by hostname()).

If sendmail think that it's name is domain.com, then when
this is changed, domain.com may be needed to be added
to class $=w (on some configurations this is /etc/mail/local-host-names
but that depends actual sendmail.cf).

Class $=w is names which sendmail consideres local (equivalent of
it's hostname).

/ Kari Hurtta