On Fri, Nov 12, 1999 at 07:21:58AM -0600, Matthew Callaway wrote:
> Dave Sill wrote:
>
> > >It's obvious that 150 MB of mail is a lot to process on such a pokey
> > >little machine, but it seems a bit odd for the machine to completely
> > >choke and die.
> >
> > If you push an underpowered system running antiquated software to the
> > breaking point, don't be surprised if it breaks.
> >
> > Is that 16 MB RAM parity, ECC (:-), or pot luck? Do you have adequate
> > swap?
> >
> > If qmail croaks a system, either the hardware or the OS is buggy.
> >
> > -Dave
>
> The swap space is 50 MB, which seems adequate to me.
Well, 10MB Message for 15 Users makes 150MB vs. 16MB Ram + 50 MB Swap makes
64MB. So if qmail tried to deliver all the messages simultaniously, it will
run into trouble, and could make the machine *seem* to be crashed, while it
was just really, really busy. This specially happens with IDE disks, which
are not running in DMA-Mode, since IO on this Disks costs quite a lot of cpu
time.
> I realize that the machine is old, and is running software that has
> updates, but the point is that a heavily loaded mail program shouldn't
> *kill* a machine. I would understand slow performance. How can you be
> sure that the hardware or the OS is buggy, and not qmail? I'm not that
> familiar with qmail. Could it be that qmail could be reconfigured to
> handle mail without bringing the system to its knees?
Under any good OS (not Win95, Winnt?), software running in userspace (that
is, all programms, including init) should _never_ be able to crash the
machine. If the machine crashed under heavy load and/or heavy IO, it seems
that your hardware is buggy. Consider, that non-ECC RAM _could_ cause
errors, as well as IDE-Disks.
Greetings, Florian Pflug
