[exim] Fwd: Delay in Exim smtp acknowledgment

Sorry, for reposting this, but can anyone help?


---------- Forwarded message ----------
From: Steve Karlsen <eximaccount@???>
Date: Thu, 11 Nov 2004 07:47:30 +0000
Subject: Delay in Exim smtp acknowledgment
To: exim-users@???


Hi everyone,

I have a problem which I would like to ask for some help with, I have
a few ideas around why the problem occurs, but maybe someone else has
experinced this problem.

Currently using Exim 4.30, within a private network, and hence it only
talks to trusted boxes within this network.

The box is a SUN v440 4x processors, 8Gb Ram
some configs are as follows : Pls note the commented out ones as well.

receive_timeout = 2m
smtp_receive_timeout = 2m

delay_warning = 0h
smtp_accept_max = 70
smtp_accept_max_per_connection = 20
smtp_accept_max_per_host = 70
# smtp_accept_queue_per_connection =
# smtp_accept_reserve =
smtp_connect_backlog = 20
# smtp_rate_limit_hosts =
# smtp_rate_limit_mail = 2,0.5,2,4m
# smtp_rate_limit_rcpt = 2,0.5,2,4m
# remote_max_parallel =
# smtp_load_reserve =
# deliver_load_max =
# queue_only_load =
# queue_run_max =

#log_refused_recipients = true
# log_smtp_confirmation = true
# log_ip_options = true

My problem is, that I have a client sending mail to this box, which
experiences a very quick smtp conversation (less than 0.5secs)
throughout, until the final dot to end the conversation is given by
the client, from then on it takes exim around 20 to 30 secs to
acknowledge with the message ID.

Now the exim box isn't busy, cpu idle is between 80 and 95%, we have
seperate spool disks and a seperate disk for the operating system, and
iostat shows that none of these are busy.
Yet one can clearly see from snoops in Ethereal, that the the last
stage from client issuing the dot to Exim acknowledging, takes 20 to
30 secs, while the whole smtp conversation before the dot, took less
than 0.5 secs??

The client that is sending does seem to have queues of messages
waiting to go out and does seem to have around 50 or so connections
established to wards the exim box. The exim box also seems to be
refusing connections constantly from the client, because it is
exceeding smtp_accept_max = 70. When I tried manually opening up
connections towards the exim box, I did have to try a few times as my
connection request was refused. When I did finally connect and enter
data, it all went smoothly, but I had a delay in response to the dot
of about 12 secs.
Could the smtp_connect_backlog = 20 be accepting connections in backlog?

I cannot think what else is causing the delay in the exim response, as
the box is most certainly not busy. Would increasing smtp_accept_max
help? there is no delay in the client responses.

many thanks

Steve