Trucluster slow performance

[eddouik]

Hi all and thanks in advance for your help,
I have 2xES45 (4xCPU/One With 16Gb RAM and the other with 4Gb)running Trulcuster 5.1B with PK3 and running Oracle 8.1.7 and some other applications developped localy .
I suspect problem in kernel parameters , can someone tell me how can i find ideal parameters for my plateform ? Or can someone tel me what can i do ?
I see that the PIDs are very high number ! Is this the main cause of the problem .

[Hein van den Heuvel]

Welcome to the HP Tru64 ITRC Forum!

>> I see that the PIDs are very high number ! Is this the main cause of the problem .

NO, Blame it on the users. It's always the users. If they were not using the system, then it would not be slow. Correct?

YES it is the problem. Not the PIDs themselfs but because the stamement suggests that there is insufficient understanding on about the workings of a Unix box.

MAYBE actually... If the systems are creating too many processes too rapidly.


Seriously now, this is way to big a problem with way to little detail to solve here in this forum.

nd just in case you wonder, a large PID has ZERO, NONE, NADA system performance impact.
They coudl simply mean that the system has been up for a long time, or they could be a tell tale sign of excessive activity which may or might not be avoidable.

Please consider getting serious contrating / consulting help (me! :^)

Focus on the ORACLE side of things.
Once the OS is set up halfway reasonable, one generaly can only improve performance by only a few procent, where are Oracle tuning my readily impact the performance by 100%.

Check out the following 'best practices' document:

http://h71028.www7.hp.com/ERC/downloads/4AA0-9949ENW.pdf

Collect information (with collect!)

- Are both systems doing similar things?
- Both slow?
- Typical usage pictures (peak, average):
---- CPU Load
---- MEMory load
---- IO Load
---- NETwork load
- What are top resource users? Oracle? FTP? NFS?

Please provide a better problem description if you want better help!

Best regards,
Hein van den Heuvel
HvdH Performance Consulting.

[Ivan Ferreira]

You can run sys_check -perf to check kernel parameters and the running system. You should also check the Tru64 Performance Guide available at the doc page, it as a whole section that describes the tuning configuration recommendation for oracle specifically.

[Hein van den Heuvel]

As your are building information to help others help you an important input factor will be 'slow compared to what' and 'did it change'.

Has this application beeen running for a while, or is is new?

Did it used to run with reasonable performance?

What changed?

Did the go slow gradually, or does it seem to have fallen from a cliff

That better you define slow, the easier it will be to improve things

Cheers,

Hein van den Heuvel
HvdH Performance Consulting

[eddouik]

Thanks all,
I can see with the command "ps alxww" extracted from the output of "syschec -perf" that the value of VSZ is about 3.2GB for oracle processes (and 20Gb for the kernel idle)
Please see attachment .

[Hein van den Heuvel]

Well, that's a start. Not much, but it's a start. Grepping for intense CPU uses ( more than 1 hour... see below) shows a hand full of oracle tasks, notably an export (and its slave) at over 3 hours?! That would slow stuff down.

Looks like a 4GB SGA, which is find for the 16GB config, but too much for 4GB config.

But why not ask Oracle how it is doing (SQL> SHOW SGA, statspack (or at least utlbstat, utlestat).

Still, no clue about active/average usage.
I like the 'monitor' tool. But some vmstat and iostat, or sar or collect might help some. But like I said, there is so much to look at before we would even know that where to look that you may be better of getting some dedicated help. From the data provided so far it seems you team does not even know where to start at this point. A good education opportunity!

Regards,
Hein.

C:\Temp>perl -ne "print if /\d+:\d+:\d+|VSZ/" tmp.txt
UID PID PPID CP PRI NI VSZ RSS WCHAN S TTY TIME COMMAND
0 1572864 0 123 38 -6 20.4G 615M * R < ?? 16:00:59 [kernel idle]
0 1573052 1572865 0 44 0 2.02M 120K badf5e60 U ?? 01:11:14 /sbin/update
100 1574256 1572865 498 44 0 13.6M 7.1M * R ?? 2-01:11:51 /usr/bin/../bin/../opt/java118/bin/../bin/alpha
/native_threads/java -Dtomcat.home=/usr/internet/httpd/tomcat org/apache/tomcat/startup/Tomcat
103 1581418 1572865 6 44 0 3.87G 2.4M * S ?? 01:13:33 ora_lgwr_dbtest
103 1587496 1572865 0 44 0 3.87G 3.6M * S ?? 03:36:30 oracledbtest (LOCAL=NO)
103 1848643 1572865 53 44 0 3.88G 12M * R ?? 01:10:07 ora_p000_dbtest
103 1854645 1854730 107 44 0 23.8M 13M * R ?? 03:33:39 exp FILE=/backup/Backup_Sonatel/SavSonatel_0603
2007.dmp LOG=/backup/Log_Sonatel/Log.Sonatel06032007 FULL=N OWNER=EPPIX CONSISTENT=N BUFFER=800000
103 1855201 1572865 23 44 0 3.87G 5.1M * S ?? 01:06:22 oracledbtest (LOCAL=NO)