load average above 4
David L. Gehrt
dlg at mail.inanity.net
Wed Dec 14 00:27:54 UTC 2005
> I was copying some large files (1 GB each) via scp and noticed the
> load went up. I also noticed this in the log:
> "mtrr: type mismatch for fd000000,800000 old: uncachable new: write-combining"
I have no insight into the mtrr message, but read on the "elevated load
average" I discovered many years ago that load average is only a partial
(and often rough) indicator of the load on the system. The load average
is a report ONLY of the number of processes that are currently runnable.
A better indication is the percentage of CPU utilization. The figures
below indicate that the CPU is idle 36% of the time, running in user
space (applications) 37% of the time and in system space (OS) 17% of
the time.
What this tells me is the copy applications are spending most of the
time waiting, probably for applications to complete or become
interruptible or for I/O to complete so that they can continue.
> top:
> top - 14:30:27 up 2:24, 5 users, load average: 4.47, 2.65, 1.16
> Tasks: 112 total, 4 running, 108 sleeping, 0 stopped, 0 zombie
> Cpu(s): 37.1% us, 17.2% sy, 0.0% ni, 36.3% id, 8.7% wa, 0.7% hi, 0.0% si
> Mem: 4025072k total, 4009104k used, 15968k free, 3108k buffers
> Swap: 2040212k total, 152k used, 2040060k free, 3838652k cached
>
> vmstat:
> procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu---
-
> r b swpd free buff cache si so bi bo in cs us sy id wa
> 5 0 152 17104 3448 3837272 0 0 252 294 68 453 8 7 84 1
> 2 1 152 16856 3440 3835980 0 0 14990 38994 1398 7416 35 17 44 4
> 3 0 152 15728 3432 3839108 0 0 14608 94 1385 7387 33 15 44 ?
If you check man vmstat you will see that the system has been idle 44%
of the time over the last to reporting intervals. My guess is that the
system remained pretty responsive even though the load ave look high to
you.
The first two columns report runnable (but waiting) processes blocked in
uninterruptible sleep.
Another useful statistics program for you t0 consider is iostat (1).
Does this help?
dlg
>
> Any tips to some tests to figure out whats going on? I am using kernel
> 2.6.9-22.0.1.ELsmp
More information about the fedora-list
mailing list