On Tue, 13 Nov 2001, Wilco Noordermeer wrote:
> I noticed 'empty' messages being delivered to my mailbox, containing a
> Return-path, Envelope-to, Delivery-date, Received, Msg-ID, From: and Date,
> but nothing else; no body or subject.
Did these correlate with the sent messages?
> It turned out that my /var partition had filled up and was at 102%
> (stupid, but true). During that period, I only received messages from
> machines running Postfix, but they were 'empty' as described. The Exim
> 'mainlog' just mentions the delivery of these messages, ans I see nothing
> strange about them.
Did Exim mainlog give a size for these messages?
> The Postfix maillog says this:
>
> delay=7, status=deferred (host
> my.exim-running.host [12.23.34.45] said: 452 space shortage, please try
> later)
>
> To make sure this was not Exim 3.16, I upgraded to 3.33 and tested this
> agein. I'm seeing the same behaviour. Does anyone have an idea why these
> messages are partially delivered, while the sending MTA does note that
> there's not enough space, and it therefore defers the delivery? Is this
> correct behaviour or not?
I cannot understand the behaviour. If Exim says 452 space shortage, it
would not have received the message, and I cannot see how it could go on
to try to deliver it. Indeed the 452 is given in response to the MAIL
FROM command. After that, the Postfix MTA would be unable to send the
rest of the message.
Hmm. Tests show that there is a bug in Exim, but it does not cause this
behaviour. The bug is that, if the sending host ignores the 452 error,
and tries to send the message, Exim does not complain. This I will fix
in Exim 4. However, I don't for one moment think that Postfix would
misbehave like that, so it does not explain anything.
I cannot reproduce the effect you report in my tests. Can you reproduce
it using a telnet session, or is it only from Postfix? Is there a way of
logging the SMTP dialogue on a failing transaction? (From the Exim end
you can run a daemon with -d9 set, but it would be helpful to see what
the other end thought was happening too.)
Philip
--
Philip Hazel University of Cambridge Computing Service,
ph10@??? Cambridge, England. Phone: +44 1223 334714.
