Documenting a virtualization project

Experiences in Virtualization

Follow up: Oracle troubles

Posted by martijnl on March 23, 2007

I will be writing a “lessons learned” post about our project in the near future but for now I will follow up on the Oracle problems we had last month.

What we learned from using Oracle in a VI environment is that memory settings turned out to be highly critical in relation to the performance of the VM. We didn’t have this problem on a physical machine, at least not as intrusive as this, but when Oracle doesn’t have enough memory available it starts very heavy swapping to the virtual disk. This results in a 100% CPU load and processes start to hang with an unresponsive application as the result for the users.

Adding memory to the amount the server needs (indicated by checking the swapsize in the VM) has fixed these problems.

One Response to “Follow up: Oracle troubles”

  1. Jean-Marc Desvaux said

    We will migrate from Oracle 8 / HP-UX 10 64 bit to ESX 3.0.1/Oracle Linux/Oracle EE 10R2.
    I have installed Linux from the Oracle distribution download, installed 10.2.0.1 from Metalink download and patchset 10.2.0.3 on top.
    We have also installed 2 ESX patchset related to 64 bit guests, namely 1410076 and 8173580.
    My oracle home and my datafiles are on an NFS mount (one for each) from a NetApp filer.
    Server is a Dell 2950 Quad-core X5355 with 8GB RAM, half granted to the Oracle Linux VM.
    No other VMs are powered on on this server yet.
    Dell openManage 5.1 installed on ESX (on the Service Console).
    Before installing the Oracle patchset and the ESX patches, I experienced one ESX pink screen and one Linux crash.
    But since installation of patches and OpenManage and VMtools, no more problems yet (3 days only after).
    Test when the same Oracle instance is mounted on a Linux machine without Vmware shows that we loose a performance of around 10%. But we think it’s ok and balance by the advantages of Virtualization.
    Oracle with basic init parameters needs a min 1.5GB to start and work properly and I don’t see much more memory being used when run on a VM, it nearly the same.
    By the way, the cost-based optimizer of 10gR2 is completely different from the rule-based used previously and when some queries run 8times faster, some are slower than before…

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

 
%d bloggers like this: