High load average low cpu usage

[Emiliano_1]

Hi
I have this problem on my application server,
an High load avarege of CPU but a low utilization.
My hardware is a partition os SD64 (no vpar)
4 CUP 875 MHz end 14GB RAM.
The problem is on tomcat 4.0.4 end java 1.4

output of sar -u
15:19:14 %usr %sys %wio %idle
15:23:04 18 22 1 59
15:23:09 15 17 0 68
15:23:14 21 27 0 51
15:23:19 21 20 0 58
15:23:24 28 23 0 49

Average 16 20 1 63

output of sar -d
15:25:12 device %busy avque r+w/s blks/s avwait avserv
15:25:17 c0t0d0 2.59 0.50 3 22 5.08 13.41
c0t1d0 0.60 0.50 1 4 8.81 10.59
c1t0d0 1.20 0.50 2 18 4.57 9.01
c1t1d0 0.20 0.50 0 3 8.40 8.84

Average c0t0d0 3.13 1.04 4 24 8.43 15.06
Average c0t1d0 0.06 0.50 0 0 7.23 8.65
Average c1t0d0 2.36 1.06 3 21 8.79 13.72
Average c1t1d0 0.03 0.50 0 0 7.23 8.66
Average c3t5d1 0.36 10.56 1 33 6.02 2.84

output sar -v

15:55:33 text-sz ov proc-sz ov inod-sz ov file-sz ov
15:55:38 N/A N/A 160/4120 0 920/6104 0 1758/30010 0
15:55:43 N/A N/A 160/4120 0 920/6104 0 1759/30010 0
15:55:48 N/A N/A 160/4120 0 920/6104 0 1757/30010 0
15:55:53 N/A N/A 159/4120 0 919/6104 0 1753/30010 0
15:55:58 N/A N/A 159/4120 0 919/6104 0 1756/30010 0
15:56:03 N/A N/A 160/4120 0 920/6104 0 1761/30010 0
15:56:08 N/A N/A 160/4120 0 920/6104 0 1762/30010 0

output of vmstat
procs memory page faults cpu
r b w avm free re at pi po fr de sr in sy cs us sy id
263 0 0 706495 1898435 20 4 0 0 0 0 0 7056 17948 19195 17 18 65
263 0 0 706495 1898435 11 1 0 0 0 0 0 6817 15558 21289 16 25 59
263 0 0 706495 1898435 5 0 0 0 0 0 0 5882 12360 21081 10 23 67
2 0 0 701464 1898435 13 1 0 0 0 0 0 4867 11019 19105 12 17 72
2 0 0 701464 1898435 6 0 0 0 0 0 0 7024 13276 20594 19 21 59
2 0 0 701464 1898435 2 0 0 0 0 0 0 5674 11278 20520 14 23 64
142 0 0 700700 1898435 0 0 0 0 0 0 0 6933 12995 21645 18 23 59
142 0 0 700700 1898557 0 0 0 0 0 0 0 6949 14060 20002 24 20 56
142 0 0 700700 1898557 0 0 0 0 0 0 0 7712 16697 20184 24 21 55
142 0 0 700700 1898119 19 9 0 0 0 0 0 6530 15579 22282 13 25 61

In attach the output of top

Thanks in advance Emiliano

[Coolmar]

Please see the following thread:

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=48314

I have found that the load average is somewhat misleading and is not necessarily what you think. I have had very high load averages in the past on virtually idle systems.

[Steven E. Protter]

Shalom,

2 ? 299 ttmgeo 152 22 1784M 1347M run 609:43 182.86 182.54 java

I think its a java problem.

I'd patch java because its making other process wait or is bound to the cpu when it does not need to be.

SEP

[Ivan Ferreira]

Your load indeed is a little high, but it depends of the "normal load" of the system. I have some systems with normal load average of 50 (Tru64) but with heavy cpu usage.

In your case, you can see the same ammount of job in the user and system cpu times. Probably your application is doing too much system calls for some reason. I don't see pi/po so this should not be a memory shortage (swapping/paging) problem and you don't have iowait.

Application tracing should be done.

[Chan 007]

Emiliano,

Can you please elobrate what is the problem.

Check sar -Mu (which will give you useage of all CPUs and then average.

Just check the process run by "ttmgeo"

I don't see any problem with your top.

Chan

[Emiliano_1]

I have check the patch level installed on the server, there are GOLDAPPS11i end GOLDBASE11i June2003, then I have find the java versione installalled on the system Java 1.4.1.05. On site Hp Java I have controll the matrix for Java end Patch Bundle end this patch PHKL_32457 end PHKL_34534 are not istalled.

In attech the outout of sar -uM end sar -q


Thanks

[James George_1]

Hi

Please check the value of dbc_max_pct and dbc_min_pct in your kernel .

make sure it is 10 and 5 respectively.

some times it is set to 50 and this will hold 50 % of your memory ... when it is 10 and 5 , it will release the meory and there will be big diffrence in the performance of your application server.

Rgds/ James

[nanan]

Hi
As you can see in the TOP report.
the system cpu power focused on jaba application.
and most of CPU power consumed on system call. it means that you need to try the java application version and patches

Probably application tuning part or OS patch..

No resource problem...
Regards..
nanan