Whats your config like..

Fri Apr 26 09:27:49 IST 2002

David

The following is a common technique for dealing with long, slow, queues
in sendmail. I suspect a similar technique can be applied with Exim. I
give some detail for the benefit of those new to sendmail.

Before pruning the sendmail "mqueue" directory in these situations it is
best to create a parallel queue directory called (say) "mqueue_slow" and
move all currently queued mail from "mqueue" into this new queue. You
can then restart the sendmail listener with an empty "mqueue". New mail
should start flowing immediately. 

You can then weed out of "mqueue_slow" the spam and other junk messages
if you wish. The next step is to run the slow queue on a regular basis
until it is empty. In crontab you create a sendmail invocation to run
every two or four hours to process mail in this queue. If the slow queue
directory was named /var/spool/mqueue_slow then, in an appropriate
crontab entry, you would invoke sendmail thus:

  /usr/lib/sendmail -OQueueDirectory=/var/spool/mqueue_slow -q 

You should probably add "-OTimeout.queuereturn=5d" to the above line so
that the time to live in the queue is extended (assumes that you
normally purge a job after 3 days in the queue - increase the above
value as appropriate for your site). Note also that "sendmail" may be
somewhere other than in /usr/lib.  

Quentin
---
PHONE: +44 191 222 8209    Computing Service, University of Newcastle
FAX:   +44 191 222 8765    Newcastle upon Tyne, United Kingdom, NE1 7RU.
------------------------------------------------------------------------
"Any opinion expressed above is mine. The University can get its own."  

> -----Original Message-----
> From: David Lancaster [mailto:dml at unb.ca] 
> Sent: 25 April 2002 18:25
> To: MAILSCANNER at JISCMAIL.AC.UK
> Subject: Re: Whats your config like..
> 
> 
> Just a thought, how good is your IO configuration?
> We had a SunOS mailserver get bunged up by a slew of spam, 
> and the combination of a slow scsi disk for the mailqueue, 
> and the problem that FFS/UFS has with directories with a 
> large number of files caused a massive IO wait problem...
> 
> Faster disk and some liberal pruning of spam mails that were 
> sitting in the queue fixed things right up...  Similar to 
> your situation, moving the bouncing email out of the queue 
> allowed it to keep up with new incoming mail just fine, the 
> disk is a just a future precaution.
> 
> Mind you, this was without sendmail alone, without mailscanner...
> 
> D.
> 
> On Thu, 25 Apr 2002, Kelly Hamlin wrote:
> 
> > heres something strange.. :)
> > i had about 7000 messages in my outgoing queue and most have been 
> > there 2-4 days, i let sendmail run for 30 minutes without 
> acceptning 
> > any new connections, then removed all the outgoing queue, 
> and now its 
> > keeping up.. What else is wierd, my secondary machine, only 
> has about 
> > 1500 in its outgoing queue (mainly undelioverable stuff) 
> and it keeps 
> > running good..
> >
> > Didnt make sense to me, so im posting to let ya know, but i 
> appriciate 
> > your input about your config.
> >
> > ----- Original Message -----
> > From: "Quentin Campbell" <Q.G.Campbell at NEWCASTLE.AC.UK>
> > To: <MAILSCANNER at JISCMAIL.AC.UK>
> > Sent: Thursday, April 25, 2002 12:21 PM
> > Subject: Re: Whats your config like..
> >
> >
> > Kelly
> >
> > Ideally you want to be running two or more boxes of the 
> same type as 
> > MX hosts for your domain(s). If each MX host has the same 
> precedence 
> > value in the MX record then they should implicitly load 
> share via DNS 
> > round-robin selection.
> >
> > Quentin
> > ---
> > PHONE: +44 191 222 8209    Computing Service, University of 
> Newcastle
> > FAX:   +44 191 222 8765    Newcastle upon Tyne, United 
> Kingdom, NE1 7RU.
> > 
> ----------------------------------------------------------------------
> > --
> > "Any opinion expressed above is mine. The University can 
> get its own."
> >
> >
> > > -----Original Message-----
> > > From: Kelly Hamlin [mailto:fizz at bomb.net]
> > > Sent: 25 April 2002 17:12
> > > To: MAILSCANNER at JISCMAIL.AC.UK
> > > Subject: Re: Whats your config like..
> > >
> > >
> > > I do have a second box setup to be a secondary MX with no child 
> > > limit. Ive got a couple p3 800's laying around, im gonna see if i 
> > > cant make use of these.
> > >
> > > ----- Original Message -----
> > > From: "Quentin Campbell" <Q.G.Campbell at NEWCASTLE.AC.UK>
> > > To: <MAILSCANNER at JISCMAIL.AC.UK>
> > > Sent: Thursday, April 25, 2002 10:45 AM
> > > Subject: Re: Whats your config like..
> > >
> > >
> > > > -----Original Message-----
> > > > From: Kelly Hamlin [mailto:fizz at bomb.net]
> > > > Sent: 25 April 2002 13:28
> > > > To: MAILSCANNER at JISCMAIL.AC.UK
> > > > Subject: Whats your config like..
> > > >
> > > >
> > > > Im running a  Dual 266 / 512 Megs Ram, on Slackware 8.0
> > > > I process anywhere between 25 and 35 thousand emails a day.
> > > >
> > > > My question to you is.. What is your setup like and how
> > > many emails do
> > > > you process?
> > > >
> > > > Reason im curious is for the last 4 days ive come in to 
> about 10k 
> > > > messages queued up. Its like it just stops working. If i
> > > disable spam
> > > > checks it will clear the queue within about an hour, but
> > > what i wanna
> > > > know is how i can keep it running fast, even with spam
> > > checks as our
> > > > customers have grown to love this feature.
> > >
> > > Kelly
> > >
> > > You need more and faster boxes! Two dual 1GHz processor 
> Linux boxes 
> > > with 1GB or 2GB of memory should cope with that load and some to 
> > > spare. If the two boxes are MX'd to your mail
> > > domain(s) then you also have some resiliance should a server fail.
> > >
> > > This site handles more than 90K incoming messages a day. 
> If you also 
> > > count outgoing email then our Mail Hubs handle more than 200K 
> > > incoming/outgoing messages a day.
> > >
> > > With just MailScanner + McAfee AV software running, we could cope 
> > > with that message load shared across 4 x Sun SPARC Ultra-5 boxes 
> > > running Solaris 7 and sendmail. These have 266MHz 
> (approx) CPUs and 
> > > 384MB memory.
> > >
> > > Two of the four Mail Hubs are significantly busier than 
> the others 
> > > and we could not run SpamAssassin on these two without 
> building up 
> > > large backlogs and even refusing incoming connections at 
> peak times. 
> > > For this reason we had to disable the use of SpamAssassin.
> > >
> > > Like you we wanted to run SpamAssassin. The solution was 
> to replace 
> > > each of the four Sun boxes with dual 1GHz Intel processor 
> boxes (2GB 
> > > memory). We run RedHat Linux 7.2 in place of Solaris. I simply 
> > > recompiled our existing sendmail under Linux and run it with the 
> > > existing sendmail.cf file. Building the Redhat system and 
> installing 
> > > MailScanner, McAfee software and SpamAssassin was easier 
> than when 
> > > working with Solaris!
> > >
> > > For key infrastructure servers I would always recommend 
> using RAID; 
> > > in our case we mirror all disks (RAID 1). Each server has 4 disks 
> > > providing a mirrored set of 2 disks; the first is used as 
> the system 
> > > disk + sendmail log disk + local applications disk while 
> the second 
> > > disk of the set is the sendmail spool disk.
> > >
> > > I hope this info is of some help.
> > >
> > > Quentin
> > > ---
> > > PHONE: +44 191 222 8209    Computing Service, University 
> of Newcastle
> > > FAX:   +44 191 222 8765    Newcastle upon Tyne, United
> > > Kingdom, NE1 7RU.
> > > --------------------------------------------------------------
> > > ----------
> > > "Any opinion expressed above is mine. The University can get its 
> > > own."
> > >
> >
> 
> 
> 
> ===========================================================
> David Lancaster
> ITS ESS
> 