Hi Carl,
The TR-3322 that you mentioned in your email is a great start and has basic fundamentals of NFS protocol but it is rather an outdated one.
For the current and latest mount options recommendations, please refer to the following URLs - these options are well tested and have been validated with Oracle.
NetApp Database Portal FAQ available from the NOW site: http://now.netapp.com/NOW/knowledge/docs/bpg/db/oracle_faq.shtml
NOW site (recommended mount options): http://now.netapp.com/Knowledgebase/solutionarea.asp?id=kb7518
Oracle Metalink: metalink doc for mount options: -- note 359515.1
https://metalink.oracle.com/metalink/plsql/ml2_documents.showDocument?p_ database_id=NOT&p_id=359515.1
Regarding protocol performance comparison - please check out:
TR-3496: ORACLE 10g PERFORMANCE - PROTOCOL COMPARISON ON SUN SOLARIS http://www.netapp.com/library/tr/3496.pdf
Regarding more information on forcedirectio or llock on Solaris - please refer to: http://now.netapp.com/Knowledgebase/solutionarea.asp?id=kb6570
According to our experts: You should use DirectIO for the REDO Logs as it reduces CPU overhead on the host. There is no benefit by using llock and caching REDO Log IO. Bypassing the FS cache for REDO traffic is a good thing, especially for high SPARC CPU count systems and any AMD based systems even as of Solaris 10 U2. It has to do with CPU DTLB tear downs, etc. Using llock in addition to forcedirectio does not gain anything on Solaris, thus it's EITHER-OR.
We are also in the process of merging the content of TR-3322 with TR-3496 and we will remove TR-3322 from the web to avoid any further confusion.
Please let us know if you have any questions.
Thanks,
Shanthi Adloori Sr. Manager, Oracle Alliance Engineering 408.822.7828 | Network Appliance | shanthi@netapp.com
------------------------------------------------------------------------ -------- From: Sto Rage(c) [mailto:netbacker@gmail.com] Sent: Wednesday, May 02, 2007 11:31 AM To: Carl Howell Cc: toasters@mathworks.com Subject: Re: Solaris Oracle NFS mount options
Yeah, the Netapp documents are not consistent. This is what we have for our oracle mounts: nfs 2 yes rw,bg,vers=3,proto=tcp,hard,intr,rsize=32768,wsize=32768,forcedirectio,l lock <<--- For all data files, redo logs and arch log volumes. and no forceddirectio for the Oracle binaries volume nfs 2 yes rw,bg,vers=3,proto=tcp,hard,intr,rsize=32768,wsize=32768,llock
Is this correct? It seems to be working OK for us now, we are yet to go into production though, its only a few developers working on the database. BTW, we are on Solaris 10. -G
On 5/2/07, Carl Howell chowell@uwf.edu wrote: In TR-3322 it says that the mount point for the redo logs should have the forcedirectio option on and the mount point for the data files should have it off. KB7518, which appears to be more recent, doesn't say anything about mount options for redo log volumes, but I assume you should have either forcedirectio or llock on both the mount points for your redo logs and data files. Is this correct?
Thanks,
- Carl