On 2012-10-01 at 17:54 +0200, Davide D'Amico wrote:
> route_data =
> ${lookup{$domain}lsearch{/usr/local/etc/exim/remote-servers}}
> no_more
> In /usr/local/etc/exim/remote-servers I have:
>
> domain1.tld: internal_smtp1::1026:internal_smtp2::1026
> domain2.tld: internal_smtp1::1027:internal_smtp2::1027
> domain3.tld: internal_smtp1::1028:internal_smtp2::1028
>
>
> Everything just is running fine but sometimes (3/4 times every day) a
> mail to any.user@??? is routed to internal_smtp1:1028 or
> internal_smtp2:1028.
>
> Now I'm testing route_list instead of route_data (to avoid the lookup
> operation).
>
> Is there anything I could check?
Exim's optimised for one port per remote host, per the normal SMTP setup
on the Internet, and the retry database keys do not include the port
number.
I'm coming down with a headcold and can't remember the option to set the
maximum number of messages to shove down one connection, but that on the
SMTP transport would help.
Alternatively, use multiple hostnames so that each port has a distinct
hostname to use with it, and set no_retry_include_ip_address on the SMTP
transport. That way, keying will be by just the hostname.
"internal_smtp1_1026" as a hostname, for instance, repeating the port
number, and set up as a CNAME.
If there's a cleaner way in Exim, it's not occurring to me right now,
but see above re my current limited capacity.
-Phil
