On 2013-08-22 at 20:02 +0530, dhanushka ranasinghe wrote: > im getting following error in our exim log ,
>
> SMTP data timeout (message abandoned) on connection from
>
> is there any reason for that, how can is incease the smtp data time out
> value
The reason is almost certainly some kind of network problem between you
and the sender. I suspect that you have a tunnel somewhere, with
endpoints in RFC1918 address space, or with bad packet filters on a
misconfigured firewall, so that Path MTU Discovery has been broken and
fragmented packets are being blocked too.
The consequences of broken Path MTU Discovery are that when you start
getting full sized packets, the end-points can't figure out the largest
size to send without getting fragments. The consequences of dropping
fragments in this case are that you get broken TCP connections that stop
making progress, as soon as one side starts sending full-sized packets.
I think I've given enough terms here to help you use a search engine to
find out more.
You can temporarily work around the issues locally to the mail-server by
using whatever your interface management tool is (ifconfig(8), ip(8)) to
lower the device MTU. That will hack around the broken PMTUD at the
cost of lowering the throughput capacity of the interface, maintenance
overhead and other issues from leaving a broken network in place. If
you need help making the MTU change "permanent" on the mail-server,
you'll need to ask folks familiar with your choice of OS/Distribution:
it's not a problem specific to mail-servers.
Note that if you mostly run web-servers where the mail-server is, then
those will be mostly sending full-sized packets _outbound_ and only
suffering from this problem for full-sized _inbound_ packets if you have
sites with lots of cookies, so it might be that the mail-server is
simply the first place that you're noticing a broader problem.
You don't want to increase the timeout, but if you decide to ignore my
advice and try it anyway, then you're looking for the
`smtp_receive_timeout` option.
