Vmware Datastore usage different from LUN

[defrogger]

Hello, not sure if this is the right place for this post or not

We have a Fiber SAN FAS3240, we are running Vmware 5.5 with Vmware View VDI.  I noticed that the Space on the VMware Datastore is about 60% free, however the space on the LUN reportred by Netapp is only 9% free. 

The way its setup is 1 Volume at 1 TB, 1 LUN at 1 TB, and the Vmware Datastore is 1 TB.

Ive been reading other posts and cant figure out if I should be worried or not.  Should I only go by what Vmware says is free and I have 60% free.  Meaning the space on the LUN from Netapp isnt showing actual free space due to Zeroed out Blocks or something like that. 

Or is there really only 9% free and if I add more VM's i could run out of space?

Ive read some articles about having to reclaim space, if this is needed how would i go about it?  We are using Netapp 8.14 7 Mode.

I know this lun has had allot of VM's removed over the last year, so is that space free or not?

Thanks, hopefully that made sense

[defrogger]

Forgot to mention, there is no thin provisioning on the Volumes or LUN side of things

[rwelshman]

NetApp only knows when a block has been modified, it doesn't know if it was deleted by Vmware. So what Vmware see is the actual available space within the LUN.

Do you have any snapshots on the volume? Because that will also skew the usage that you see on the NetApp side.

[defrogger]

No Snapshots.

So I talked to NetApp Support today.  My understanding of what is happening is that this.  I should go by what Vmware is showing in terms of how much free space there is.

However, the space reported by NetApp on the LUN is saying what you are saying in a way, that Netapp doesnt really look back to see if Vmware has deleted anything until it reached 100% used and then it goes back and looks for free blocks.  So technically it will find blocks that are free but there is a performance hit while its doing this. 

So its recomended to do space reclamation with Scsi unmap in vmware.  Or use Storage Vmotion to move everything off the LUN, delete it, and then recreate it, and then move the VM's back. 

So it looks like i have the space, so it just becomes a question of how much of a performance hit will i take when Netapp sees it as 100% full. 

So ill look into the space reclamation thing or using storage vmotion to move things around.

Thanks

[marcel_juhnke]

Hi,

you could use the UNMAP primitive from ESXi to reclaim unused space in the LUN as described in http://www.netapp.com/us/media/TR-4333.pdf on page 142: Dead Space Reclamation Using UNMAP

With the UNMAP primitive, the storage array can be notified about blocks that have been freed after VMs have been deleted or migrated to another datastore. In VMware vSphere 5.0, UNMAP was run by the vmkfstools –y command, which specified the percentage of blocks that should be freed. In vSphere 5.5 the vmkfstools –y command is replaced by the esxcli storage vmfs unmap command, which specifies the numbers of free blocks.

Regards

Marcel

[fedaynnetapp]

Hi,

It happens the other way around to me.

I'd already run the UNMAP primitive and a lot of unused space was reclaimed.

In my case, the LUN is thin provisioned and all VMWare VM have thin provisioned hard disks.

There are just a couple os small in size snapshots.

How on earth we have more space within the LUN than the Datastore?

Thank you.

[fedaynnetapp]

Hi,

I missed to say that dedup is enabled for the Volume that holds that LUN.

There's only that LUN within the Volume.

But I've always though that when lokkin at free space numbers you haven't to take into account dedup.

It seems that space reclaimed with unmap commands is released back into the aggregate, so no gain for the datastore.

https://community.netapp.com/t5/Data-ONTAP-Discussions/Free-Space-difference-between-NetApp-and-WmWare/td-p/76172

Thank you.