HP Performance

[mingzheng huo]

hello:
I met a performance problem .My Database is oracle9.0.1 for HP-UX11.0. When the DB Server is running,the job queue process uesed all of my CPU time.
If I set JOB_QUEUE_PROCESSES parameter to 1, it will take CPU time 96 percent, if I set it to 6, it also take so much.
So my application become very slow.
Why the job queue process takes so much CPU time, how can I resolve this problem.
Thanks

[Ravi_8]

Hi,

can u reveal RAM size and swapinfo -t o/p ?

[mingzheng huo]

Hi:
Very glad to meet you,this is the command result.
# swapinfo -t
Kb Kb Kb PCT START/ Kb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 2097152 1564 2095588 0% 0 - 1 /dev/vg00/lvol2
dev 1843200 1225808 617392 67% 0 - 0 /dev/vg00/lvol11
reserve - 2235408 -2235408
memory 1117812 302892 814920 27%
total 5058164 3765672 1292492 74% - 0 -
thanks.

[David_246]

Hi,

The info you provided us shows how bad it is looking right now.
You currently have a system with aprox 1-1.5 Gb of internal memory. In total you have defined 3.8 Gb of Swap whereof on one device 67% is used.

So, instead of using (the fast) memory it will swap onto disk. A disk is very slow compared to physical memory. as soon as you start Oracle your system is die-ing because of the high swap. You can see this using the command sar -d 5 100 (for example). The swap disk will be very busy now.

So, a couple of solutions possible :

- increase physical memory (will cost money)
- decrease the memory oracle claimes (per instance)
- decrease the amount of instances

When you issue the following program:

#!/bin/ksh
# This script is written to retreive the total memory claimed
# Written by D.M. van der Geer # 11 februari 2003

/usr/bin/clear
t=0
for i in `UNIX95= ps -e -o vsz=Kbytes -o ruser -o pid,args=Command-Line | sort -rnk1 | grep -v Kbytes | awk '{print $1}'`
do
t=`expr $t + $i`
done
echo "\n In totaal is er $t Kbytes aan memory geclaimed \n"

echo "Swapinfo :"
/usr/sbin/swapinfo

echo "\n"

You can see how much memory is claimed by the system.
If you want to see only the memory claimed by oracle you can change the UNIX95 line into :

UNIX95= ps -e -o vsz=Kbytes -o ruser -o pid,args=Command-Line | sort -rnk1 | grep -v Kbytes | grep oracle | awk '{print $1}'

Maybe you can give us some of this info. I don't think we can give you more options than we just did :(

Regs David

[Steven E. Protter]

You have kernel configuration problems most likely, and other performance issues.

I've attached a script to collect data over a period of time and give you the ability to analyze it. The script was originally developed by HP and enhanced by me. Neither party takes responsibility for it, though it does work and has yet to damage any system after wide distribution in itrc.

Increasing message queue sizes seems to me to be your most immediate concern.

SEP

[mingzheng huo]

thanks a lot.
I will try it soon.

[mingzheng huo]

Hi:
There is 3 instance in my HP , but one instance is strange. when it's running , it will make a process and take most CPU time.
Can u tell me what the process is?
TTY PID USERNAME PRI NI SIZE RES STATE TIME %WCPU %CPU COMMAND
? 2940 oracle 236 20 52416K 3348K run 7:01 95.98 95.81 ora_j000_GYGBSI