2016-01-14 02:14 AM
We have a customer asking about how space is being reported up in vCenter for their Netapp Datastores. They're looking at the datastore 'performance' tab, and looking at the 'space' view.
This view shows a pie-chart of datastore usage, broken down by file type. Alongside the usual 'Virtual Disks' and 'Free Space' category, they see an 'other' category. This is normal of course, and things like logs and metadata are counted here. However, on a FAS unit, this can also be influenced by the configuration of the flexvol.
In our case, we had one volume that initially had 'Other' space usage into the terabytes range! This was down to fractional reserve. Removing it eliminated most - but not all- of the 'Other' space (so for example on one volume there is still 26GB shown). After confirming there was nothing at a VMWare level taking up this space, I've been tasekd with investigating if there is anything else that can impact things from the Netapp side.
So, can anyone here help me with an explanation? SnapShots in this case total far more than the 'Other' space shown, so it isn't that (unless it's being mis-reported), and I'm struggling to find anything that would explain it.
Any help greatly appreciated!!
2016-01-20 03:52 PM
What mode and version of Data ONTAP are you using and what protocol do your vSphere Datastores use? IE 7-Mode or Clustered Data ONTAP, NFS, ISCSI or FCP? ONTAP version 8.X.X?
There are some bugs that relate to de-duplication which might be related to your issue:
Have you raised a support case for this?
2016-01-21 07:19 AM
It's a cDOT system, 8.3. Storage is being presented to the hosts by NFS.
No support case - in the first instance I just want to see if anyone had seen this or could offer any thoughts that might help explain this for us and our customer.
2016-01-21 04:57 PM
Okay so based on your ONTAP version you won't be effected by that deduplication issue. However if your your NFS volume is thin provisioned and has deduplication enabled (which most customers would do for vSphere datastores) then the dedupliction process will create a fingerprint database and change log (which is stored in the volume and is used by the deduplication process to compare hashes of each 4K block in WAFL). I think that might be where you are seeing the space discrepancy which vSphere is showing as "other" space. See following technical report for deduplication: