I just had a strange case where kswapd started eating up 100% CPU
time. I haven't been able to reproduce it, but it seemed to occur when I
did a "dd bs=4096 </dev/fb0 >/scratch/image" in X while writing a CD (4x)
from the same filesystem--the dd took 10-20 times longer than usual, and
kswapd's run time from ps also pointed to the same time. At the time,
roughly 6MB of real memory was free, but most in-use memory (~800MB) was
cached data; only 3MB or so of swap was used. The system itself seemed to
remain stable, though I rebooted shortly after I discovered the problem.
System is as follows: (if more info is desired, please contact me
directly at [email protected]'m not subscribed to the list)
Kernel: 2.4.13 (i686, SMP)
CPU : Dual Pentium II 400MHz
Memory: 896MB RAM, 576MB swap
IDE : 1 HD (data, including /scratch, and 512MB swap)
SCSI : 1 HD (root and 64MB swap), 1 CD-R drive
Video : 3dfx Voodoo 3 (PCI), 16MB VRAM (3dfx framebuffer driver enabled
for virtual consoles, but X hits the hardware directly)
--Andrew Church
[email protected]
http://achurch.org/
Hello all,
I have this problems too.
Any sugestions?
Regards
Marcus Grando
> I just had a strange case where kswapd started eating up 100% CPU
>time. I haven't been able to reproduce it, but it seemed to occur when I
>did a "dd bs=4096 </dev/fb0 >/scratch/image" in X while writing a CD (4x)
>from the same filesystem--the dd took 10-20 times longer than usual, and
>kswapd's run time from ps also pointed to the same time. At the time,
>roughly 6MB of real memory was free, but most in-use memory (~800MB) was
>cached data; only 3MB or so of swap was used. The system itself seemed to
>remain stable, though I rebooted shortly after I discovered the problem.
>
> System is as follows: (if more info is desired, please contact me
>directly at [email protected]'m not subscribed to the list)