Re: Slow response to smtp clients

On Tue, 6 Apr 2004 Victor.Duchovni@morganstanley.com wrote:

> On Tue, 6 Apr 2004, Mipam wrote:
>
> >
> > I did follow the document, but i still didnt figure out what's wrong.
> > Especially the messages like i posted in the original mail:
> >
> > Apr 6 15:46:09 rbns01 postfix/smtpd[879]: maps_find:
> > recipient_canonical_maps: virusalert@domain.com: not found
> > Apr 6 15:46:09 rbns01 postfix/smtpd[879]: maps_find:
> > recipient_canonical_maps: virusalert: not found
> > Apr 6 15:46:09 rbns01 postfix/smtpd[879]: maps_find:
> > recipient_canonical_maps: @domain.com: not found
> > Etc.
> >
>
> This is completely normal, uninteresting and does show any delay. Please
> show two consecutive log lines (for a single process) where the timestamps
> differ by many seconds...

Thanks for the answer. I took the liberty the study some more traffic.
A connection started at 09:55:58.

09:55:58.926404 x.x.x.x.3824 > y.y.y.y.25: S [tcp sum ok] 935
413832:935413832(0) win 32120 <mss 1460,sackOK,timestamp 1087970642
0,nop,wscale0> (DF) (ttl 58, id 30162, len 60)

Then a lookup about who x.x.x.x is etc, then after the RCPT TO: command
the blackhole lists are being checked:

09:55:59.140576 y.y.y.y.49152 > 209.208.0.2.53: [bad udp cksum b8
9f!] 40767 [1au] A? x.x.x.x.dynablock.njabl.org. ar: . OPT
UDPsize=4096 (60) (ttl 64, id 57781, len 88, bad cksum 0!)

So here i guess dynablock.njabl.org is asked wheter x.x.x.x is listed.
At last

09:56:11.033200 y.y.y.y.49152 > 204.108.129.10.53: [bad udp cksum
eb1c!] 5868 [1au] A? x.x.x.x.bl.spamcop.net. ar: . OPT UDPsize=4096
(55) (ttl 64, id 57874, len 83, bad cksum 0!)

Then the last rbl check is done about 12 seconds later.
At last

09:56:11.124201 y.y.y.y.49152 > 195.149.107.22.53: [bad udp cksum
e61!] 31178 [1au] A? domain.com.dsn.rfc-ignorant.org. ar: . OPT
UDPsize=4096 (57) (ttl 64, id 57877, len 85, bad cksum 0!)

And after this postfix send an OK to the client.
In other words, the delay is prolly caused by the many rbl's i'm
consulting. So following
http://www.porcupine.org/postfix-mir...E.html#sniffer

came up with the answer.
In other words, the delay is my own bad. On yes, my server has broadcom
5704C NICs with have tcp offloading stuff, and it should be supported, but
somehow i keep seeing bad tcp/udp checksums, i've gotto figure out
what's going on with that.
Bye,

Mipam.