mcsg related issue

[apurba_k]

dear all,
we have 2 k360 servers and mcsg loaded in it (hp-ux 11.0)pkg1 is running in one server and pkg2(oracle) on other..now our observation is ..as the oracle process crosses 85-90 plus ..the floating ip given to the oracle package ..stops responding from few windows NT workstation and HPUX workstation ..but we can ping to the physical ip of the server .oracle version is 8.0.6.we have recently upgraded to 11.0 but the problem was in 10.2 also but not frequent .
pls help me sorting this issue out
warm regards
Apurba

[Steven E. Protter]

I little more data about your situation would be useful.

Following commands:

swapinfo -tam either for the cluster or the individual machines.

How much physical memory is present.

You might want to use a script I'm attaching to collect some performance data to point out possible performnance issues.

Also posting up a performance document from an oracle tuning expert at HP.

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&mode=id&admit=82735245+1053636168960+28353475&searchCrit=allwords&printable=true

Based on the limited data you gave, most specifically the upgrade of the oracle application, a supposition can be made.

You upgraded your application, which was not on a properly tuned box or didn't have enough memory in the first place. The situation became worse because of the voracious resrource needs of the Oracle products.

I actually believe you are using two products, the oracle database and the Oracle application suite 11.00

I am told the oracle application suite runs better on 8.1.7.0 or above of the database.

You also might want to make sure you are running all of the oracle required patches for the OS and Service Guard available at http://metalink.oracle.com

SEP

SEP

[Massimo Bianchi]

HI,
from what you say this issue is related to the oracle configuration.


you shold change (and bounce the db) the following parameter to allow more connections:

STANDARD VALUE
process=96
session=80


when you hit the max, there are no messages and you can connect any more, it depends on the first to connect.


You must remeber:

-process=session*1.2
- for each process oracle will reserve a semaphore, so keep an eye on you kernel parameters.

HTH,
Massimo