On 09/05/2010 04:59 PM, Wietse Venema wrote:
Jeroen Geilman:
As for your original question, the combined processing time of all your
smtpd_* checks will still be reflected in the delay-"a" value (pre-queue).
Whatever time postfix itself adds for processing will be either static
or insignificant (unless you have lots of expensive maps) .
The bulk of the time in smtpd_mumble_restrictions happens between
these two logfile records:
smtpd[pid]: connect from name[addr] (reverse/forward name lookup completed)
smtpd[pid]: queueid: client=name[addr] (first recipient accepted)
The delay-'a' value measures the time since the incoming queue file
was created, i.e. when the first recipient was accepted.
This is why most rejects are logged with NOQUEUE as the queue ID.
Wietse
Ack, thanks - I often wondered about that one (and the "unknown[unkown]"
client conundrum - but that's probably an artifact of the socket
handling code.)
I guess you could try to calculate something out of the logs between
those two events, but there may be a lot of noise in that (at a minimum,
you'd have to match up the pids).
Having a syslogd that records milliseconds would also help in that; I
know rsyslogd can write ms logs, since I use that myself.
J.
