You wrote: “Oracle’s Unbreakable Kernel is based off of RedHat’s Kernel.” – this is actually not the case. The Unbreakable Enterprise Kernel is derived from a recent version of mainline Linux, we try to minimize the number of patches we apply on top of it. The UEK will be updated to a new version from mainline every 12-18 months and is fully maintained and supported by Oracle.

With regards to the missing vmv_pvscsi driver – this was indeed a bug (as noted in the release notes) that was addressed in later errata releases of the Unbreakable Enterprise Kernel ( ELSA-2011-2010 for OL5/6: http://oss.oracle.com/pipermail/el-errata/2011-March/002008.html and http://oss.oracle.com/pipermail/el-errata/2011-March/002014.html)

Even if it is, I don’t know of anyone running a production Oracle system on Solaris x86 – Solaris on Sun/Oracle Hardware yes, but not Solaris on Intel.

My personal recommendation these days is to go with Nahelm class or better x86 CPUs running RedHat Enterprise Linux under VMware vSphere.

Good luck!

Anyone please give some lead.

regards

I have a support case open about oracle processes keeping file handles on dropped asm disks. This seems to be a bug that they don’t get fixed (even after several patches)

But I dont understand how the asmlib helps you to solve this?
The rdbms processes are still going to access the block devices directly.

Thanks for the pointers to the articles.

You can find the statement about ACFS / ADVM not supporting the UEK kernel in the release notes and on Oracle support notes:

http://download.oracle.com/docs/cd/E11882_01/relnotes.112/e16778/toc.htm#sthref1
mos note: Certification Information for Oracle Database on Linux x86-64 [ID 1304727.1]

You are right about the hard partitioning, but when used you can no longer make use of live migration and after a failover you need to manually pin the cores again.
In my opinion this severely limits the benifits of using virtualization.

You can also add ASMlib to the list of thing oracle will no longer support under RHEL6.

Although UDEV do replaces ASMlib for most installations, we still need ASMlib to get rid of stale file handles problems left by the RDBMS.

I rarely touch RAC these days, but it is on my to-do list to get more familiar with it.

ADVM / ACFS – I haven’t heard anything about ASM dynamic volume manager or ASM cluster file system not being supported when using a UEK kernel… can you send me some links on this so I can do my own research?

Yes, you are correct, RHEL 6 / OL 6 is (still) not supported for Oracle DBs – I’ve got to think they’ll do that one of these days. You might want to check out Kevin Closson’s comment on my “On Oracles commitment to Linux” post a few days ago for his (and to a lesser degree, the oaktable private list) thoughts on the matter.

Regarding PVSCSI, you may want to check out http://www.vmware.com/pdf/vsp_4_pvscsi_perf.pdf which is a performance white paper from VMware on the performance benefits on PVSCSI vs. LSI Logic SCSI adapters. You should note that white paper was based on vSphere 4.0 and PVSCSI in vSphere 4.0 could actually decrease performance compared to LSI if the VM is pushing less than 2000 IOPS consistently. This was rectified in vSphere 4.1 .

I personally don’t use RDM volumes unless some clustering solution (for example a quorum drive with microsoft clustering) requires it. VMware and EMC both say there isn’t a performance advantage and the disadvantages (no snapshots, no thin provisioning, etc) outweigh any possible benefits, IMHO. So my iSCSI/NFS/FC VMware datastores have the VMDKs of my Oracle VMs on them, using a PVSCSI SCSI adapter if I’m running vSphere 4.1 or higher.