[Exim] unnecessary invocations of dial on demand

Okay. I have some more data for this:

exim causes DoD even by running 'exim -bp' on an empty queue with the
standard settings. When I insert a line

primary_hostname = MyMachine

this behavior stops and any local activity is fine. However, now
remote messages are not delivered any more. I have done some debugging
and found this calling 'exim -d9 -M MESSAGEID':

---SNIP------
MyMachine in local_domains? yes (matched MyMachine)
--<SNIP>------
routelist_item = * smtprelay.PROVIDER1.com:post.PROVIDER2.de bydns_a
after handling route_lists items, matched = 1
original hostlist=smtprelay.PROVIDER1.com:post.PROVIDER2.de options=bydns_a
expanded hostlist="smtprelay.PROVIDER1.com:post.PROVIDER2.de" options=bydns_a
DNS lookup of smtprelay.PROVIDER1.com (A6) gave TRY_AGAIN
smtprelay.PROVIDER1.com in dns_again_means_nonexist? no (end of list)
returning DNS_AGAIN
DNS lookup of smtprelay.PROVIDER1.com (AAAA) gave TRY_AGAIN
smtprelay.PROVIDER1.com in dns_again_means_nonexist? no (end of list)
returning DNS_AGAIN
DNS lookup of smtprelay.PROVIDER1.com (A) gave TRY_AGAIN
smtprelay.PROVIDER1.com in dns_again_means_nonexist? no (end of list)
returning DNS_AGAIN
smarthost router deferred gmx.de
message: host lookup for smtprelay.PROVIDER1.com did not complete (DNS timeout?)
added retry item for R:gmx.de: errno=-37 0 flags=0
---SNIP-------

It does not matter, if I am online or not. The DNS lookup fails. Okay, I
could accept that, if it would be always like that. However, without the
'primary_hostname' parameter in the exim config file this DNS lookup
works:

---SNIP-------
localhost in local_domains? yes (matched localhost)
--<SNIP>------
original hostlist=smtprelay.PROVIDER1.com:post.PROVIDER2.de options=bydns_a
expanded hostlist="smtprelay.PROVIDER1.com:post.PROVIDER2.de" options=bydns_a
DNS lookup of smtprelay.PROVIDER1.com (A6) succeeded
DNS lookup of fmrlallmx.PROVIDER1.com (A6) gave NO_DATA
returning DNS_NOMATCH
DNS lookup of smtprelay.PROVIDER1.com (AAAA) succeeded
DNS lookup of fmrlallmx.PROVIDER1.com (AAAA) gave NO_DATA
returning DNS_NOMATCH
DNS lookup of smtprelay.PROVIDER1.com (A) succeeded
---SNIP-------

Note the first line I have cut out. By setting
'primary_hostname=MyMachine' the name 'MyMachine' is used instead of
'localhost'. However without this parameter the documentation of exim says
the primary_hostname is set by calling 'uname()' and uname returns
'MyMachine' as well.

This is really confusing ...

What is causing all this? Exim? libc6 resolver?

The bugreport for 'host' is supposed to be closed. Should I open up
another bug for exim to collect all this debugging stuff? I am not
familiar with the Debian debugging process (not yet :).

Comments, hints, fixes?

Bye, Steffen

On Wed, Nov 07, 2001 at 09:29, Frank Simon wrote:
> > > After some more testing I have found out that this occures whenever
> > > sending a mail, no matter whereto (localhost, MyMachineName or
> > > somewhere else). My dial on demand is always invoked by DNS querying
> > > for MyMachine. I will do some more reading. Hopefully I can
> > > stop that by
> > > changing my exim.conf file somehow. Maybe a new exim version has
> > > introduced a default behavior that causes this.

> > I am experiencing similar problems - my Dial On Demand triggers often
> > when some program wants to contact localhost. Among those is leafnode;
> > I am running a local news-spool (leafnode), and whenever I post a new
> > news message, or even if I access a newsgroup, the ISDN connection is
> > started. Another is ssh....
> >
> > To me it seems like there is something wrong with the configuration of
> > the "lo" interface - in fact it doesn't even appear in the "route"
> > list.

> after installation of my woody, there was a similar (same ?) problem.
> the loopback was loaded but got no ip address (no "iface lo inet
> loopback"-entry in /etc/network/interfaces )and no /etc/hosts-entry.
> I don't know, weather it caused lookups, because of being allways online
> with dsl-flat.