I have a feeling that majordomo is handing lists over to
bulk_mailer, which in turn is sorting and batching lists
just fine before handing the process off to sendmail.
I have witnessed sendmail deliver 6-8 messages per second
on a well connected single intel p5-133 processor based
system running BSDI unix. At that rate, your 1000 queued
messages could get delivered in 2-3 minutes..
The biggest bottleneck you can encounter in any mailing
list scenario is bad addresses in the list. Do all or
most of the lists have their subscribe_policy set to
+confirm? An address with a bogus domain could "defer"
all the messages in that batch. Multiply this by 20 lists,
with one bad address each, and the average number of
messages per list per day, it does not take long to get
1000 messages in the queue.
Dan Liston
Norman Walsh wrote:
>
> I'm really at a loss, suggestions most welcome...
>
> I'm working with a Sparc 20 with 2 cpus and 256Mb of memory.
> The loadavg never seems particularly high. The ISP that it's
> connected to claims to have lots of bandwidth and never more
> than about 40% saturation. I'm running Solaris 7, sendmail
> 8.8.8, majordomo 1.94.5, and bulk_mailer on it. I installed a
> caching only DNS server.
>
> It's running about 20 or so lists, a couple with roughly 2,000
> users and fair volume. And it just cannot keep up.
>
> Given that the machine isn't working very hard, I'm inclined to
> think that it's a sendmail or other configuration
> problem. Installing bulk_mailer and a caching DNS server got the
> turnaround time down from half a day to half an hour, but there
> are lots of lists out there doing better than that. And the queue
> has already grown to more than 1000 messages...
>
> Help!?
>
> Be seeing you,
> norm
>
> --
> Norman Walsh <ndw@nwalsh.com> | One must look for one thing only, to
> http://nwalsh.com/ | find many.--Cesare Pavese
References:
|
|
