Oracle 12c, ASM and NFS

We are seeing a lot of customers who are concerned that NFS has been desupported in RAC with 12cR2. We got the same concerns with 12cR1.

What actually happened is Oracle removed support for certain RAC configuration files directly on a filesystem, NFS or otherwise, with new 12cR1 installs, although older 11g installs could be upgraded to 12cR1 in place. Oracle has totally desupported files directly on a filesystem for 12cR2 RAC.

Instead, ASM must be used for all RAC installations. This does not create a complicated management requirement for ASM. The RAC installer will prompt the user for the paths to some files on NFS. These are then configured as a basic ASM instance with the RAC related files stored inside. The administrator can essentially forget that ASM exists after installation unless they plan to use ASM for the databases themselves.

Here’s what it looks like when complete:

root@jfs3-a ~]# cd /crs
root@jfs3-a crs]# ls -l
total 36096508
-rw-rw—- 1 oracle dba 39641088000 Oct 18 06:38 ocrvfdgdisk0

I’m using NFS with just a one-disk ASM diskgroup. Yes, there are some ASM processes running, but it was all set up automagically when I installed RAC. If I wanted extra reliability, I could have made this a 3-disk group using three different NFS shares from different controllers, but on the whole one is fine for a local cluster. Previously there were a bunch of files in an NFS share that I ignored. Now there are a bunch of files encapsulated in an ASM diskgroup based on some NFS files that I am ignoring.

The installer has options for all of this. The only real difference is you have to use “dd” to make a big, blank file on the NFS shares you use for CRS/voting, and set the permissions. You then point the installers to those giant pseudo-LUNs rather than pointing the installer to the filesystem itself. The installer does the rest.

2 thoughts on “Oracle 12c, ASM and NFS

  1. Funny you should mention that this is the exactly issue we are running into and Neal (our sales Engineer is going to try and setup a call with you for later this month to go over our oracle stuff in broader terms.

    Does oracle/netapp have any best practice on this ? In previous versions we would create 3 /votingdisk volumes. Now the dba’s keep wanting to mirror this with disks based on those volumes and then an extra big one for the GIMR database.

    Questions without answer (Mostly due to Oracles exceedingly vague documentation.)

    Which redundancy? ( it feels wrong to use external based on old rac deployments)
    How big are the crs/voting disk files ?

    Which redundancy for GIMR ?
    How big should it’s disk files be ?

    And Lastly can all these stupid files be sparse as writing zeros to an AFF seems a waste of time all around.


Leave a Reply to Jeffrey Steiner Cancel reply

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

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

Google photo

You are commenting using your Google 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 )

Connecting to %s