Hi,

This may or may not be a Linux problem.

Here is the configuration :

two redhat linux 3.0U7 server configure in Oracle RAC Cluster.

The server are connected to a Compact HGS80 SAN with one hba each.

Oracle version is 10.2.0.3 for CRS and RDBMS.

The HBA are qlogic 2300.

The problem : I am log into one account through putty on one of the server. Every 5 minutes the system freeze/hang for about 5 to 10 seconds. I am in the process of typing a command and the display of the typing stop for 5 to 10 seconds.

My Oracle user also are experincing the same freeze/hang.

Did anybody experience a similar situation.

As I said there are many possible culprit :

bad kernel parameter
problem with the SAN or HBA card
problem with Oracle Cluster.

Thank for any response.

André

Hi André, plesase check the log messages(Linux and Oracle)to see if there's something mess. Monitor by "top" would also be useful when "freeze/hang". Or you can stop Oracle service, and watch the system to eliminate hardware issue.

Phillip

Hi,

Thank for the answer.

I did check the process with top and the culprit seem to be kscand.

I create a small script that log the time every five second. Every time kscand is the top process, I have a delay in the time. Here is an exemple :

Mon Mar 12 17:28:23 EDT 2007
Mon Mar 12 17:28:28 EDT 2007
Mon Mar 12 17:28:33 EDT 2007
Mon Mar 12 17:28:38 EDT 2007
Mon Mar 12 17:28:43 EDT 2007
Mon Mar 12 17:28:48 EDT 2007
Mon Mar 12 17:28:56 EDT 2007
Mon Mar 12 17:29:01 EDT 2007
Mon Mar 12 17:29:06 EDT 2007
Mon Mar 12 17:29:11 EDT 2007
Mon Mar 12 17:29:16 EDT 2007
Mon Mar 12 17:29:21 EDT 2007

To my knowledge kscand only check the memory. I don't understand that it take a couple of second to do it and freeze everything else.

André

Ok,I think I found a solution to my problem.

I have activated the vm.hugetbl_spool parameter in sysctl.conf parameter file.

With this parameter activated, all the sga goes in one reserved space and their is no need to check as often the memory.

I will continu to monitor my system and post a new update tomorow.

André

Hi,

I have solve my problem by activating the vm.hugetbl.spool parameter in sysctl.conf.

I don't understand why it solve it and why my Oracle Rac configuration needed it.

If anybody has a reason or a theory on the subject I would be interrested to here it.

André

Hi André,

Oracle SGA includes data buffers, share pool, log buffer and so on, which allocated memory are determined the performance of Oracle RAC. In your case, heavy memory load means you have to tune these parameters carefully, this job is interesting and sometimes need more patience. Well, congratulate for your fixing this issue, . If you want to know more in details, please go to OTN:

http://www.oracle.com/technology/

Regards,
Phillip

Hi,

Thank for the info.

André