Uptime Load Average

[Pattabhi]

Hi All,

Can someone explain in simple terms the meaning of "load average"

What is a normal load and above normal load?

I happened to see the below load on one of my servers today and yet i was able to loginto.

# uptime
12:57AM up 32 days, 21:50, 5 users, load average: 7500.85, 1848.33, 640.99
# uptime
12:57AM up 32 days, 21:50, 5 users, load average: 5374.93, 1729.23, 626.93
# uptime
12:57AM up 32 days, 21:50, 5 users, load average: 4945.25, 1700.67, 623.47
# uptime
12:59AM up 32 days, 21:51, 6 users, load average: 1417.72, 1324.83, 573.73
# uptime
12:59AM up 32 days, 21:52, 6 users, load average: 1417.72, 1324.83, 573.73
# uptime
12:59AM up 32 days, 21:52, 6 users, load average: 934.61, 1218.92, 558.01
# uptime
12:59AM up 32 days, 21:52, 6 users, load average: 859.89, 1198.78, 554.92
# uptime
12:59AM up 32 days, 21:52, 6 users, load average: 859.89, 1198.78, 554.92
# uptime
01:01AM up 32 days, 21:54, 6 users, load average: 226.74, 918.24, 507.74
# uptime
01:01AM up 32 days, 21:54, 6 users, load average: 208.61, 903.07, 504.93
# uptime
01:02AM up 32 days, 21:55, 5 users, load average: 59.93, 703.39, 464.58
# uptime
01:03AM up 32 days, 21:55, 5 users, load average: 46.90, 669.14, 456.92
# uptime
01:03AM up 32 days, 21:55, 5 users, load average: 46.90, 669.14, 456.92
# uptime
01:06AM up 32 days, 21:59, 5 users, load average: 1.60, 332.40, 361.87
# uptime
01:06AM up 32 days, 21:59, 5 users, load average: 1.48, 326.91, 359.87


What should i infer out of this output

[Ganesan R]

Hi Pattabhi,

This is the simple term meaning of "load average"

The load average is the average number of processes in the queue waiting
for CPU time, over the last 1, 5, and 15 minutes.

So now you could understand easily...

[Pattabhi]

Hmm that make it a bit more clear now.. Thank you

[Simon Hargrave]

That's a ridiculously high load average you had there, but it seemed to settle pretty quickly. Basically it looks like there were several thousand processes running at the same time.

My guess would be that you have a shell script that got into a loop and kept spawning itself or another process.

I'd run "sar -q" (assuming you're collecting sar data) and see what time it started to ramp up, then check your crontabs for what started around that time.

[Dennis Handly]

What were you doing when you had a load average of 7500.85?

[Steven E. Protter]

Shalom,

Quite clearly you had a lot of processes waiting on some resource for a long time.

I've never seen a live, responsive system with numbers looking like that.

http://www.hpux.ws/?p=6

That link contains scripts that might help you identify the problem, however if the system is responsive and there is no user complaints, no action may be required.

SEP

[Pattabhi]

Guys,
Thanks for all your sugesstions.

I am not sure what kicked of such a high load avg, as u can see it was just within a span of 8 minutes and was back to normal. we were infact doing a test on memory leak

Well is it just the avg number of processes or does it also include the amount of memory and cpu utilised over the last 1,5 and 15 minutes.


Thanks,
Pattabhi RAman

[Patrick Wallek]

>>Well is it just the avg number of processes
>>or does it also include the amount of memory
>>and cpu utilised over the last 1,5 and 15
>>minutes.

Just processes. CPU and memory have nothing to do with the load average.

The 'uptime' man page also briefly describes the load average.

[Pattabhi]

Thanks for all your sugesstions