On Sat, Mar 27, 2004 at 11:02:31AM +0100, Leonardo Boselli wrote:
> lookuphost:
> driver = lookuphost
> transport = remote_smtp
> self = fail
> pass_on_timeout = true
>
> The problem is number 3: althopught reading in the exim documentation
> looks as if pass_on_timeout works either on any connection problem i
> keep having these lines in my log:
>
> 2004-03-27 10:11:38 1B75Cb-00085x-00 == aqhcqstmc@???
> T=remote_smtp defer (101): Network is unreachable
pass_on_timeout handles timeouts in *routing*, not in the transport
final delivery. Examples would be DNS timeouts (i.e. remote DNS
server is not responding, DNS connectivity is down, etc.).
To help you save some time, here is a well-tested example of dealing
with many (if not all) of your issues at once:
remote_domains:
driver = dnslookup
ignore_target_hosts = 127.0.0.0/8
self = fail
pass_on_timeout
transport = remote_smtp
fallback_hosts = fallbackhost.yourdomain.com
no_more
send_to_fallback:
driver = manualroute
route_data = fallbackhost.yourdomain.com
transport = remote_smtp
Benefits of this configuration:
1. The first router will handle any remote deliveries. If routing
succeeds, it hands it off to remote_smtp for direct remote delivery.
If the transport cannot connect to the remote server for any reason,
the "fallback_hosts" options tells the transport to send it on
to your fallback host of "fallbackhost.yourdomain.com".
2. If the routing determines that the MX records point to you, the "self"
option will catch it and fail the message (since it was supposed to
be a remote delivery)
3. If the DNS lookups timeout (i.e. DNS nameservers unreachable for
the remote site), the "pass_on_timeout" option will cause it to
go on to the next router, which manualroutes the message specifically
to the fallback host. This is the only way to get DNS timeouts to
immediately send to your fallback rather than sit on the queue. The
fallback_hosts option on the router will NOT catch DNS timeouts, which
is why two routers are required here.
4. The "no_more" option on the first router is required to keep
routing failures from falling through to the send_to_fallback router.
Hope this helps.
--
Dean Brooks
dean@???
