ONTAP Discussions

Can we set Fractional reserve to Zero for the thin volume..?

anji_singathi
7,888 Views

Hi Al,

I need to know to before proceeding to set the above option on the below filer..

OnTap 7.3.6P2

Model FAS3050

I need to know whether i can set fraction reserve to zero for a new thin volume,..

Please advise..

1 ACCEPTED SOLUTION

diego_flaborea
7,888 Views

Yes you can. But i recommend you to monitore your environment closely.

VOL OPTIONS <VOLNAME> FRACTIONAL_RESERVE 0

View solution in original post

6 REPLIES 6

diego_flaborea
7,889 Views

Yes you can. But i recommend you to monitore your environment closely.

VOL OPTIONS <VOLNAME> FRACTIONAL_RESERVE 0

JCASON
7,888 Views

Fractional reserve and thin vs thick volumes have nothing to do with one another.

Setting a volume to thin simply consumes space as it is requested.  Setting to thick consumes space on provisioning. 

(Actually this isn't really accurate because WAFL never really thick provisions, it is really a question of whether the total volume size is reflected as consumed)

Fractional reserve is overwrite protection for your LUNs (and only LUNs, not CIFS/NFS volumes) if you are using snapshots.  It provides a way to guarantee free space in the volume for overwrite blocks.  It is important for you to understand that if you take a snapshot and the free space in the volume fills up from change blocks, if fractional reserve is set to 0 YOUR LUN WILL GO OFFLINE.

Your question can be answered several ways:

1. Can you?  Yes.  Should you?  Depends

2. Are you using a LUN?  If not, you should always have FR = 0.

3. Are you sure you'll never be taking a snapshot?  If you are assured of this, you should always have FR = 0 (but you must leave extra space in the volume past the size of the LUN, we typically use 1-3GB). 

I always encourage everyone to thoroughly test fractional reserve before making changes in their environment.  It behaves very strangely (for instance, it only actually reserves blocks that are allocated, so in a 1TB new LUN w/ 30MB of data, FR will only reserve 30MB on snapshot). 

Also regarding #3, be VERY sure you will never be taking snapshots.  In our environment we use NetBackup, which happens to leverage SnapDrive and the VSS to take a snapshot when it does backups.  This caused us to have some LUNs go offline after we set FR = 0 on LUNs we didn't expect to have snapshots on.  We had no idea snapshots were even being taken and it bit us.

aborzenkov
7,888 Views

2. Are you using a LUN?  If not, you should always have FR = 0.

If you are not using a LUN, FR is not relevant at all

3. Are you sure you'll never be taking a snapshot?  If you are assured of this, you should always have FR = 0

Another case is deduplication (A-SIS). And snapshots are indirectly used by other features (vol copy, flexclone, snapmirror, snapvault, ...). Otherwise the same as above - as long as no snapshots - FR setting is completely irrelevant.

JCASON
7,888 Views

You are correct about the FR with no LUN.  I actually meant, but poorly stated, that the default setting was FR = 0 with a volume and no LUN but it looks like that isn't the case either based on my environment (my cifs volumes have FR=100).  Probably the FR settings are copied from vol0 like most other things.

And you make a good point. Though it doesn't apply because this is referencing 7.3, once at 8.x vol move will become available, but vol move leverages snapmirror, which takes a snapshot.  So if you decide to provision your LUNs based on no snapshots and then use something that generates one, like vol move (or dedupe, etc.), you must remember to account for the change rate during the operation. 

Because again, if I start a vol move and your volume is improperly sized for the amount of change in the LUN, it has the potential to take my LUN offline.

jeremypage
7,888 Views

Pretty sure fractional reserve works for any file...I don't think LUNs are handled differently, the difference is that it's less likely to have a file that is as close to the size of the volume if it's not a LUN but folks running VMs over NFS will run into similar situations.

From the na_vol man page:

This option decreases the amount of space reserved for overwrites of reserved objects (LUNs, files) in a volume.

JCASON
7,888 Views

It is just for LUNs.

df -hr volname will show you the reserve allocated to that volume. If you do this for a NAS volume with FR=100, you'll see the reserve is 0.  For a volume with a LUN, it will show you the amount currently reserved. 

Public