I have just come across a slighty odd instance where a local server is legitimately trying to send out large quantities of mail (12K at a time) and my internal MX is unimpressed, dropping the connection with "too many messages in one connection" in the rejectlog after around 6K of messages.
Now the bit that confuses me here is that I have left:
smtp_accept_queue_per_connection as the default (this is not in the config file at all), so by my reckoning and according to Philip Hazel's book "a single SMTP reception process can receive any number of messages". I cannot see anything in the online documentation about why the connection is being dropped or what any default value now is.
Is there some internal default limit that is being triggered here to cause the connection to drop?
I have to admit that the internal sending server appears to have changed its behaviour in that previously it would send the mails one at a time over individual connections over a relatively long period of time where now it seems to be trying one connection and sending the whole lot. I guess the theory is that 'they' can say 'the messages have been sent in 5 mins' rather than 'yes the messages are being sent now and may take a while'. Either way makes no difference to me as they are still subject to AV checking.
In addition we have updated Exim in the last few months from something so old I cannot recall the version to a more recent 4.72 (Debian package) - however the smpt_accept_* parameters are set the same and I would expect behaviour to be the same.
We have set:
smtp_accept_max = 150
smtp_accept_reserve = 50
smtp_reserve_hosts = +relay_from_hosts
relay_from_hosts are trusted server networks etc (where the originating server is based).
I can set smtp_accept_queue_per_connection to a suitably large number - but that leaves the question: should I need to?
Thanks
Paul
Paul Osborne
Systems Analyst
Infrastructure Group
Computing Services
Canterbury Christ Church University
Tel: +44 1227 782751
