Re: instant failures???

Page principale
Supprimer ce message
Répondre à ce message
Auteur: Philip Hazel
Date:  
À: Pengar Enterprises Inc. & Shire.Net LLC, Tom Samplonius
CC: exim-users
Sujet: Re: instant failures???
On Mon, 17 Mar 1997, Pengar Enterprises Inc. & Shire.Net LLC wrote:

> These messages came back within a minute of sending an initial message to
> the address. How come it doesn't age the failures anymore and keep on
> trying? It just gives up after the first check?


It ages *host* failures, not message failures. If a host has been down
for 4 days, why keep on trying for another 4? See the last page of
chapter 30 in the (1.61) manual (The paragraph starts "Special processing...".)
Hmm. This stuff is all rather buried; I should give it more prominence
in the next edition.

On Mon, 17 Mar 1997, Tom Samplonius wrote:

> Probably because the retry database indicates that the site has been
> re-tryed many times in the past, and just isn't working. So Exim tries
> one more time, and bounces it.


Indeed, except that the "one more time" is dependent on the
delay_after_cutoff option in the smtp transport. In the default state,
it will try one more time only after the next retry time has expired.
With the default retry rule, this means it will try only once every 8
hours after a site has been dead for 4 days. If you set delay_after_cutoff
to be false, then it will try each IP address that has not been tried
since the message arrived. Given a constant stream of messages, this
setting will try the broken address much more often.

On Mon, 17 Mar 1997, Pengar Enterprises Inc. & Shire.Net LLC wrote:

> It doesn't even seem to try it one more time. It just sees it marked
> "unusable" and doesn't even try. How does one reset something in the retry
> database? A quick scan of the various command line params doesn't show
> anything.


There isn't an easy way to do that, but it can be done by running the
exim_fixdb utility and deleting the record for the host:ip-address that
is failing.

Another thing you can do, if you know that the host has come back
on-line, is to queue a message using -odq and force a delivery with -M.
That will always try to do a delivery.

--
Philip Hazel                   University Computing Service,
ph10@???             New Museums Site, Cambridge CB2 3QG,
P.Hazel@???          England.  Phone: +44 1223 334714