Hi
I am currently in the process of reclaiming disk space back from our NetApp FAS8020 Array running 7-mode 8.2.1.
All of our flexvols are VMware datastores using VMFS which are all thin provisioned volumes. NONE of our datastores are presented using NFS. On the VMware layer we have a mixture of VMs using thin and thick provisioned disk, any new VMs created are normally creating using thin provisioned disks. Our VMware environment is ESXi 5.0.0 U3 and we also use VSC 4.2.2.
This has been quite a journey for us and after a number of hurdles we are now able to see reclaim of volume space on the NetApp, this resulting in the free space returning to the aggregate.
To get this all working we had to perform a few steps provided by NetApp and VMware. If we used NFS we could have used the disk reclaim feature in VSC but because that only works with NFS volumes this wasn't an option for us.
- NETAPP - Enable lun set space_alloc to enabled - https://kb.netapp.com/support/index?page=content&id=3013572. This is disabled by default on any version of ONTAP.
- VMWARE - Enable BlockDelete to value 1 on each ESXi host in cluster - http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=2007427. This is disabled by default on the version of ESXi we are running.
- VMWARE - Rescan the VMFS datastores in VMware and update the VSC settings for each host. Set recommended host settings. Once performed check delete status is showing as 'supported' esxcli storage core device vaai status get -d naa - http://kb.vmware.com/selfservice/search.do?cmd=displayKC&docType=kc&docTypeID=DT_KB_1_1&externalId=2014849
- VMWARE - login to ESXi host, go the /vmfs/volumes and datastore where you want to run disk reclaim and run vmkfstools -y percentage_of_deleted_blocks_to_reclaim
- NETAPP Run sis start -s -d -o /vol/lun - this will rerun deduplication and delete the existing checkpoints and start afresh.
Whilst I believe we are seeing savings on the volumes we are not seeing the savings at the LUN layer in NetApp. The volume usage comes down and with dedupe on I would expect the volume usage to be lower than the datastore usage but the LUN usage doesnt go down.
Does anyone know why this might be the case. Both our flexvols and LUNs are created using thin provisoned and space reserved in unchecked on the LUN.
