Re: killing 2.0.31-pre9

ADAM Sulmicki (adam@cfar.umd.edu)
Tue, 9 Sep 1997 19:11:30 -0400 (EDT)

->> Here is another easy way to 'kill' pre9.. at least worked for me
->> It does not literally kill it but makes it pretty unusable ..
->>
->> first
->> dd if=/dev/hda5 of=/dev/null (it is around 5 gb data)
->> make -j zImage ( 3.0.31+pre9+serial.console)
->> ping -f foo (local ethernet, just two hosts)
->> now to kill it replace the last line /w
->>
->> ping -f -s 1024 foo
->>
->> uptimes goes to 80.0 .. the swap usage increases..
->> soon you can notice zobime proceses and in end
->
->How much RAM did you have during this test? I just did it on 2.0.31-pre9
->SMP on my Tomcat II with 64mb and saw load avgs in the low teens but
->nothing catastrophic. I have much smaller disks (500mb SCSI on NCR 810),
->but did a dd reading each simultaneously.

sorry, but unfortunatelly after I did sent the email I did realize that I
did not mention the most important element of the 'kernel torture suite'
I had been then also running "Netscape Communicator 4.03 Pro for Linux" ..
a great program which can kill mac via one touch of button, make Windows
lose it is mind ...also you have run it on some 'busy' web site where is
lot java and animated gifs etc..

so summariging it is :

dd if=/dev/hda5 of=/dev/null (it is around 5 gb data)
make -j zImage ( 3.0.31+pre9+serial.console)
ping -f -s 1024 foo (local ethernet, just two hosts
netscape ( at www.cnn.com )
and some 2nd make -j ( a Xfree86 is good thing )

for me at this point would being to appear lot zombies, load would be
around 50.0 .. driving it up to 80.0 should cause it the problems start to
appear.. overall you may want run top and try make sure that all kind of
resources are equallly busy.. ie user space, kernel and I/O etc.... ..
I'll do more test tonite and let you know ..

as of ram : 64mb EDO ram.. and 130mb swap..
Cyrix PR200+
7GB EDIDE ok hdd

-Adam