B2600, HP-UX 11performance problems (HELP!!!!)

[Denton Williams]

Hi All,
I am having some performance issues on one of my systems and it always seem to be using its resources to it's fullest. Some of the things I find strange are that there seems to be a lot of Event monitor notification mails to the root account. Does that mean it is using a lot of resources to monitor the system? Also whenever I do the "find" command it almost brings the system to a complete halt. Other things to note is that the machine is our Oracle server running two Oracle instances but does not do a lot of processing, and it is a 1 processor, running OS 64 bit and 1024 MB memory. The model of the machine is B2600 with two disks (18 and 16 gigs). Any suggestions as far as performance turning will be greatly appreciated.

Thanks,
Denton.

[Mark Grant]

1GB is not much for Oracle. How is the SGA set?

Regardless of that, you really need to take noice of those e-mails, they are telling you what is wrong.

You are probably swapping a lot. Try

"swapinfo -mt" and if any of the figures for swap are not 0 (except in the total field) prepare to get more RAM.

[Steven E. Protter]

http://www1.itrc.hp.com/service/cki/search.do?category=c0&docType=Security&docType=Patch&docType=EngineerNotes&docType=BugReports&docType=Hardware&docType=ReferenceMaterials&docType=ThirdParty&searchString=UPERFKBAN00000726&search.y=8&search.x=28&mode=id&admit=-682735245+1065034641015+28353475&searchCrit=allwords

I will add that leaving vx_ninodes at 0 in the kernel causes a lot of expensive I/O.

make sure dbc_max_pct and dbc_min_pct are low and not to far apart. Mine are 6 and 2.

SEP

[Denton Williams]

Thanks for you inputs guys. Just to give you guys some feedback of how my system is set.

This is what I get when I do swapinfo -mt

Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMT RESV PRI NA
dev 2048 242 1806 12% 0
- 1 /dev/vg00/lvol2
reserve - 979 -979
memory 700 169 531 24%
total 2748 1390 1358 51% - 0 -

As far as my Kernel parameters are concerned, they are as follows:

Name Value
dbc_max_pct 50
dbc_min_pct 5
vx_ninode 0

Let me know what you guys think about this informantion.

Thanks,
Denton.

[Alzhy]

If this is an Oracle 9i (or even 8i) that you're trying to operate in 1 1 CPU (speed?) / 1 GB machine - then it's only good for a certain number of users and processing... Give Oracle9i justice -- move it to at least a a more current CPU system... or up to 2 CPU's...


What does your ipcs -a say? It wil give you an indication of how your SGA is sized and utilized....

[Joshua Scott]

Hi,

Just a thought, but it seems to me that using a single-processor workstation as an Oracle Server could be a source of the problem. The b2600 wasn't really meant to be a server.

-Josh

[Alzhy]

I was once involved in troubleshooting an Oracle problem. After spending quite some time analyzing some perf data .. I found out the instance was hosted on a Sun V100 -- a desktop not really meant as a server -- 1024G RAM, 2xIDE disks and 1 550 Mhz CPU. It started out well when just 5 people were using it.. but when that quintupled to 25, that's when problems started.... that instance is now happily being hosted on a 2 x Opteron 800 Mhz Chip Linux system with 8192 MB of RAM, U320 SCSI disks .. and very happy user community...

[Mark Grant]

Denton,

Looking at the output from your "swapinfo -mt" you are swapping. Swapping is the scurge of system performance, it is the reason your "find" takes so long. Those poor disks can't quite cope. If you are interested in seeing how much you system is swapping do an "vmstat -S 1". under the colums "si" and "so" should be lovely zero's. However, I suspect that when performance is down you won't be having many of those. Of course, you could have the database set to use too much shared memory or you could have a memory leak or an ill behaved process running about but, one way or another, you are going to have to get some RAM back.

[Michael Steele_2]

Here's part of your problem, as others have pointed out above:

dbc_max_pct 50
dbc_min_pct 5

Try:

dbc_max_pct 10
dbc_min_pct 5

Also attach:

sar -u 5 5
sar -v 5 5
sar -d 5 5
vmstat 5 5
sar -b 5 5