Why is system load so high ?

[David_246]

Hi,

I currently have a system load over 3.
The system is barely doing anything. what can it be ??

top :
Load averages: 3.02, 3.19, 3.12
174 processes: 151 sleeping, 23 running
Cpu states:
LOAD USER NICE SYS IDLE BLOCK SWAIT INTR SSYS
3.02 0.4% 0.8% 0.6% 98.2% 0.0% 0.0% 0.0% 0.0%


sar -u 1 5 :
HP-UX ??? B.11.00 U 9000/800 04/10/03

18:09:35 %usr %sys %wio %idle
18:09:36 1 0 0 99
18:09:37 0 1 0 99
18:09:38 0 0 0 100
18:09:39 0 0 0 100
18:09:40 1 0 4 95

Average 0 0 1 99

swapinfo:
Kb Kb Kb PCT START/ Kb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 3145728 0 3145728 0% 0 - 1 /dev/vg00/lvol2
reserve - 306412 -306412
memory 1619860 453376 1166484 28%


Anyone an idea. F.Y.I. the server has only one processor, so load shouldn't reach a system load of 3.

Regs David

[Patrick Wallek]

System load has nothing to do with the number of processors. The load avg. is the average number of jobs in the run queue during the last 1, 5 and 15 minute time periods.

1 kernel parm to check that can cause abnormally high load average is 'timeslice'. Make sure it is set to 10. If it is set to anything else, especially 1, reset it, rebuild your kernel, and reboot.

[S.K. Chan]

The load just shows you got processes in the run queue waiting for something. To find out use glance instead and look at Global Wait screen. Run glance with -h and it'll give you various screens.

[Bill McNAMARA_1]

I've often had single proc systems with load > 7. Response is okay too !

example:

root @ cactus / #
root @ cactus / # uptime
7:35pm up 98 days, 5:52, 25 users, load average: 7.62, 7.52, 7.53

What you won't see on a single proc is %CPU greater than 100%

Later,
Bill

[David_246]

Hi,

timeslice is already at 10
We don't have glance :( But sar should do the same as glance right ?

Bill, how can a system with one proc ever reach a load of 3 when it's doing nothing ?
There is no IO, USR or SYS, nothing that could keep the CPU busy.

I know that systemload is a complicated counting of many processor dependend issues.
I also know (from my Sun background) that the more CPU's you have the higher the systemload can be.

Can you please give me an explanation where it gets the high number from, as I can't see any activity on the system.

Your help is much apreciated (and rewarded :).

Regs David

[someone_4 · Accepted solution]

Hi david,
Glance does allot more then just sar,top. If you have some application cds it has a free 30 day trail of glance and you can see for your self.


Richard

[Michael Tully]

Part of the problem could be that you have 23 processes running. WHat are they?

COuld you attach the output of top -f please?

[David_246]

Hi,

Went home, so couldn't reply any earlier.

It's a shame, but the system is running at a systemload of 0.03 again. so there is nothing to compare it with anymore.

Thanks for your responses, I'll retry it when it's re-occuring.

Best Regs David

[Bill McNAMARA_1]

It was probably just having a nightmare..
Try this command next time, to show the top ten cpu hogs..

UNIX95= ps -e -o "vsz pcpu ruser pid stime time state args" | sort -rn |head -10

credit from this link: http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0x836cc1c4ceddd61190050090279cd0f9,00.html