[wplug] SMTP delay

Wise, Jeremey jeremey.wise at agilysys.com
Tue Aug 24 09:41:13 EDT 2004


I would second this. I have a Pentium 133mhz with 40MB RAM running ten
users SAMBA, Mail, Web and it runs fine and has been for about four
years now.

On Mon, 2004-08-23 at 20:25 -0400, Bill Moran wrote:
> duncanhutty at comcast.net wrote:
> 
> > I am testing a mail server and have been told that even an old k6II 500
> > with 320MB should be able to cope with so many zillion messages a month
> > with no problem even with LDAP and spamassassin.
> > When I (handwavingly) test this setup, sending a single, tiny message,
> > it takes about 15 seconds to 'finish' transmitting the message to the
> > mail server (on the LAN). I am calculating that time from monitoring
> > Outlook's send and receive dialogue box.
> > When I test from the same machine to my ISP's mail server (obviously
> > NOT on the LAN), the same message takes <1s to transmit.
> > Trying to eliminate error and check my testing, I made sure that there
> > was no other traffic/activity on the server. I monitored
> > /var/log/mail.log and saw 16s between the client connecting and
> > disconnecting.
> > Seems to me like 16s is a long time for a 1k message even on a vintage
> > PC like this. Are my expectations out of whack?
> > Or should I be expecting better performance and investigating why this
> > is so slow?
> > If I move the LDAP server to another machine, would I expect better
> > perfomance? I would think that network vs. localhost lookups would be
> > slower, but maybe removing the cpu cost to another cpu might help.
> 
> The potentialtech.com mailserver is only an 800mhz and it's also our
> HTTP and DNS server.  I see no such delays on it, so I can't imagine
> that it would take a 500mhz 15 seconds to transmit 1 email.
> 
> It sounds suspiciously like a timeout.  Are you sure the DNS is working
> as the mailserver would expect?  I've seen more network problems boil
> down to a program trying to do a reverse DNS lookup and timing out.
> What MTA are you using?  Can you increase logging to determine exactly
> where the 15 seconds is occurring?  I mean, is it between the HELO and
> the DATA?  Or is it after the last line of the DATA has been transmitted?
> This might give a clue as to what the holdup is.
> 
> I can't imagine any legitimate reason why that machine would take so long.
> 
-- 
Call if you have any questions,

Jeremey Wise
Jeremey.Wise at Agilysys.com
Office: (440)-519-6006
Cell: (216)-647-1121
MCSE,CNE,CSE



More information about the wplug mailing list