Pentuim Pro 300 Mhz
128 M RAM
BusLogic SCSI controller (tape drive controller)
EATA (DPT) Raid Controller (hard disk controller)
3Com Vortex Ethernet
began acting strangely after the upgrade to 2.0.34 (which was compiled
with GCC 2.7.2.1, BTW). Basically, there are *some* tasks that take
forever to complete. For example, when sendmail is run from pine, or
as newaliases, the elapsed time has gone thought the roof. System and
user time remains small; CPU utilization is very small (averages 90%
idle), system load varies, but is usually in the 0.3 to 0.7 range. For
example. I just ran newaliases and:
System Time: 0.04
User Time: 0.04
CPU Utilization: ~ 14%
System load: ~ 0.4
Elapsed time to complete command: 28 seconds!
It began exhibiting this behavior when I upgraded from 2.0.32 to 2.0.34.
It almost appears that the way the kernel is handling the allocation of
resources to processes is way our of whack. I have a 90% idle CPU, but
the above run, for example, the CPU utilization for this command was so
low as to show 0%. Other things happen very quickly: I can compile a
kernel in about 4 minutes. Sometimes even running sendmail or newaliases
can happen in a second or two: once it took close to a minute.
My first thought was that there was something strange happening with the
system (sendmail is linked to libdb and libc dynamically), but examination
of the system calls (via strace) reveal no time hog, so I'm forced to
believe that it's a kernel issue. Unfortunately, It doesn't seem that
anyone else in the world is experiencing this problem. I've discussed it
on some of the lists that I'm on, and no one else is seeing this behavior.
Now, I compiled this kernel with a processor type of Pentuim (not
PPRO), and I was getting ready to try specifying a PPRO processor
type, but at this point I'm just grasping at straws.
Does anyone on the list know of anything in the 2.0.34 kernel that
could cause this behavior?
Thanks in advance.
--Mark
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu