On Thu, Oct 10, 2024 at 10:45:08PM +0100, Andrew C Aitchison via Exim-users
wrote:
>
> > I posted this problem as an exim bug, but it was immediately dismissed
> > because the MTA rejection response indicates a temporary problem, so
> > apparently retrying 4 more times inside a second to different IP addresses
> > for the same domain is deemed acceptable behavior rather than waiting a
> > normal retry delay. If it was a connection failure rather than a message
> > rejection, I'd understand,
>
> I could possibly (not saying thst I will) do something about this
> particular bug, but I don't yet know what.
>
> Doing it right is not too hard; doing something that works when the
> other side is behaving in ways that suit themselves is harder, and
> clearly explaining what you are doing in the documentation is a lot of
> effort.
For it is worth, there's precedent for having a maximum delivery attempt
count (before deferring the message for a later queue run) that is lower
than the maximum connection attempt count.
The MTA I happen to use, by default tries to connect to at most 5 IP
addresses, but attempts at most 2 deliveries (IIRC a 4XX greeting or
EHLO response counts as a connection failure, deliveries begin once
"MAIL FROM:..." is sent).
This choice is at least defensible, and also configurable. If someone
with the right skills and expertise has dev cycles, something similar
could work for Exim.
If one really wants to maximise the odds of success, one might also want
to make sure that the choice of IP addresses to try within the maximum
count does not starve out one either the IPv4 or IPv6 address families
when both types of address are associated with (equal preference) MX
hosts. The idea is to include some of each in the otherwise randomly
ordered IP selection.
--
Viktor.
--
## subscription configuration (requires account):
## https://lists.exim.org/mailman3/postorius/lists/exim-users.lists.exim.org/
## unsubscribe (doesn't require an account):
## exim-users-unsubscr...@lists.exim.org
## Exim details at http://www.exim.org/
## Please use the Wiki with this list - http://wiki.exim.org/
