Hello.
I got the HSZ40's online... but now I have a different issue. I wrote a
script to test the raw speeds of the disks as configured on our server.
Here is the pseudo-code:
for each disk (list of disks)
umount $disk
newfs -o time $disk
mount $disk /tmpmnt
dd if=/dev/zero of=/tmpmnt/junkfile bs=64K count=262K
time cat /tmpmnt/junkfile > /dev/null
time dd if=/tmpmnt/junkfile of=/dev/null bs=4K count=4M
time dd if=/tmpmnt/junkfile of=/dev/null bs=16K count=1M
umount $disk
newfs -m 0 $disk
mount $disk /tmpmnt
dd if=/dev/zero of=/tmpmnt/junkfile bs=64K count=262K
time cat /tmpmnt/junkfile > /dev/null
time dd if=/tmpmnt/junkfile of=/dev/null bs=4K count=4M
time dd if=/tmpmnt/junkfile of=/dev/null bs=16K count=1M
end for
what I saw, running iostat, was that for a while I got results, and about
700-800 bps and 99-100 tps on the proper devices. But then I saw the page
fault count go through the roof (on the time'd commands), the IO's go to
essentially 0, and the time to write 16G went from 250-300 seconds to
1200-1500 seconds.
What happened? Was csh leaking? after I stopped the script and did all of
the commands by hand, everything was cool. So SIGINT'ing the script made
the IO's go back up and the faulting went away... has anyone seen this
before? This is an 8200 with 2GB RAM, 12 20GB arrays hanging off 3
hsz40's hanging off of 3 kzpsa's, and two CPU's.
Thanks,
Ry
--
Ry Jones 206.455.2652 x 172
rjones_at_hbsi.com, rjones_at_wicker.com (
http://www.wicker.com)
HBSI, 411-108th Avenue NE, Suite 800, Bellevue, WA 98004
This message does not represent the views of my employer in any way shape
or form.
Received on Thu Mar 07 1996 - 03:36:30 NZDT