vparmodify command delay

[Jayaprakash Subramanian]

Hello All,

I have a scenario to migrate the Oracle application along with memory from one vPar to another. If I migrate the Oracle DB from one vPar to another, the memory should be moved along with. The memory is a floating memory.

But When I migrate the oracle package using SG, the primary node takes a while (>20 seconds) to release its memory. So the package got started on the adoptive node without Memory migration.

But the same case works vice-versa. So it seems there is a performance issue with the primary node, which takes a long time to release the memory with vparstatus command. The user defined command to delete the memory is "vparmodify -p -d mem::4192:f

Is there anything to fasten the vparmodify command activity.

Thanks,
Jayaprakash S

[Manix]

Please checked if the Luns have been zoned correctly by SAN team.

Run ioscan -fnC fc and then check the HW attached to the HBAs using ioscan -fnH of your fc card >

If insf -eC disk is not able to create device new files ,most likely it`s a SAN issue.

Post the above output.

Thanks
Manix

[Manix]

Opps i am sorry ..in correctly posted !!
please ignore.

Regards
Manix

[Torsten. · Accepted solution]

You should get the white paper

Configuring and Migrating Memory on vPars

from

http://h20000.www2.hp.com/bizsupport/TechSupport/DocumentIndex.jsp?lang=en&cc=us&taskId=101&prodClassId=10008&contentType=SupportManual&docIndexId=64255&prodTypeId=18964&prodSeriesId=4146102


to get more information.


The system needs some time to free up the memory.

[Jayaprakash Subramanian]

Hi,

Thank you for the suggestion. The same case works fine when migrating the package from the adoptive node to the primary node.

As per the document, the vpar will take some time to release the floating memory. But both the vpars were identical here.

Is there any kernel level parameters there to configure them?

Thanks,
Jayaprakash S

[Manix]

This is pretty much expected behavior as memory can`t be freed at instance until the processes accessing it are exited , i don`t think that there is a kernel tunable to fasten the same operation.

Thanks
Manix