Philip Hazel wrote ("Re: [Exim] Drivers and DNS lookup failures."):
> On Fri, 17 Dec 1999, Theo Honohan wrote:
>
> > If a director passes an address to the SMTP transport, and the
> > transport encounters a failed DNS lookup, the type of the item added
> > to the retry list is "R", (not "D").
>
> Do you mean there is a setting of "hosts" on the SMTP transport?
Yes.
> Failure there is deliberately remembered as "R" because it is a
> routing failure. "D" failures correspond to actual delivery
> attempts, and their retry database keys include the IP address,
> which of course is not available if the DNS lookup has failed.
Hmmm. The symptom that caused me to notice this was that a temporary
DNS failure caused a message being handled by this director:
labs_internal:
driver = smartuser
transport = internal_delivery
and passed to this transport:
internal_delivery:
driver = smtp
hosts = zebra
hosts_override = true
to be bounced immediately to the sender, with the following log messages:
1999-12-17 11:46:30 11yvq9-0000vE-00 <= \
user@??? H=example.demon.co.uk (packet.office.internal) \
[158.152.24.165] U=root P=smtp S=1530 id=385A2210.742FFBDB@???
1999-12-17 11:47:15 11yvq9-0000vE-00 == \
spqr@??? <spqr@???> T=internal_delivery defer (-33): \
failed to lookup IP address for zebra
1999-12-17 11:47:15 11yvq9-0000vE-00 ** \
spqr@??? <spqr@???>: retry timeout exceeded
1999-12-17 11:47:16 11yvq9-0000vE-00 Error message sent to user@???
1999-12-17 11:47:16 11yvq9-0000vE-00 Completed
The retry configuration is
* * F,1h,15m; G,16h,2h,2; F,4d,12h
This wasn't the behaviour I was expecting; Am I missing something obvious?
