i thought DNFS bypassed the kernel nfs calls/overhead.  what i see below just looks like subnetting your nfs traffic to a dedicated interface which many folks already do out of best practices.

also DNFS is not available in 10g RAC correct?  i thought it was a new 11g feature


--
Daniel Leeds
Manager, Storage Operations
Edmunds, Inc.
1620 26th Street, Suite 400 South
Santa Monica, CA 90404

310-309-4999 desk
310-430-0536 cell



-----Original Message-----
From: Joe Bishop [mailto:jbishop@west.net]
Sent: Sat 12/15/2007 4:06 PM
To: toasters@mathworks.com
Subject: Oracle Direct NFS client Re: Oracle on Netapp seemless redundant NFS mounts

Dear Toasters.

A performance and availability tip for anyone running Oracle on Netapp.

This is especially true for people running Linux out there with Oracle on NetApp.

To keep the performance of your Oracle Database fast while performing other high loaded NFS operations, consider Oracle's Direct NFS client.

Why have one NFS mount when you can have 16?  And if that isn't good enough, how about another 16 on a second dedicated subnet?

Oracle has created the Direct NFS client to provide greater reliability of an available NFS connection through the redundant direct NFS mounts.

Here is an example...


NetApp Server  
  eth8 10.0.0.100   ..................[common 10.0.0.0 switch]....................  other  n-tier traffic
  eth9 10.100.0.100   ..............[private 10.100.0.0. NetApp traffic switch]

    export /vol/vol1/oradata   10.0.0.200,10.100.0.200  rw,root


DBServer
  eth0 10.0.0.200  ...............[common 10.0.0.0 switch]
  eth1 10.100.0.200 ............[private 10.100.0.0 NetApp traffic switch]

  /etc/fstab

  10.0.0.100:/vol/vol1/oradata    /oradata


  /etc/oranfstab
  server:dnfsnetapp
  path:10.0.0.200
  path:10.100.0.200
  export:/vol/vol1/oradata mount:/oradata

NOTE: remember to adjust the libodm* files
   Linux command set is

     <shtudown DB>

     localhost]# cd $ORACLE_HOME/lib
     localhost]#  mv libodm11.so libodm11.so_stub
     localhost]#  ln -s libnfsodm11.so  libodm11.so

       <startup DB>

At this point you will see the Oracle DNFS mount

    select * from v$dnfs_servers;
        ID SVRNAME DIRNAME          MNTPORT    NFSPORT      WTMAX      RTMAX
---------- --------------- ------------------------ ---------- ---------- ---------- ---------- ---------- ----------
         1 dnfsnetapp /vol/vol1/oradata       4046       2049      32768      32768

And  the new connections (DNFS exclusive 10.100.0.0 subnet count 16 -e9 entries) on the NETAPP using netstat

With two subnets, you have eliminated any single point of failure between your database and the Filer.

And you might want to retry that RAC setup with the DNFS.

Cheers,

Joe Bishop