Hello, we want to reclaim some space from some thin provisioned vmdk's. In the past we would just storage vmotion them to different datastores and we would get any freed up space back.
But since VMware 5 we haven't been able to do that as you would need to have different block sizes on the datastores. We no longer have that option.
So the reason SvMotion doesn't work is due to the newer Datamover, only the old legacy datamover would reclaim space during a SvMotion.
So I have a large VM around 32 TB with two 16 TB Vmdks. . Its a Windows 2012 Server, about 6 TB was deleted in Windows and ide like to get that space back as the VMDK is still 16TB.
So my options are as follows
Uses Sdelete -z on the VM
Then I can
1. Power off the VM and use vmkfstools command to get the space back, I think it was -k option. But this would require a long period of downtime so not an option with such a big VM
2. Use VMware Standalone Converter - This would require downtime to sync over to the new vm, but the downtime would be shorter.
3. Force the use of the old Datamover FSDM over FSDM3. Then use Storage vMotion - No Downtime
Ide like to get peoples opinion on option 3 as Ive only seen a couple posts on doing it, and im wondering if there is a reason why im not seeing to many people doing this as it seems on the surface as a good option.
So the procedure would be to Sdelete -z the drive. Then using PowerCLI I would run the following command
Get-VM -Name “vm-to-shrink” | Get-VMHost | Set-VMHostAdvancedConfiguration -Name “VMFS3.EnableDataMovement” -Value ([int64] 0)
This will force the use of the older datamover.
Then I can Storage vMotion the VM to a different datastore with no Downtime and the vmdk in theory will hopefully be 6 TB smaller.
Then I can use the command to put it back to the newer DataMover.
Ive tested this on a small VM and it works, plan on testing on a couple other small test VMs.
My understanding is the DataMover is only used for Storage vMotions and Cloning and such. So we don't use Storage DRS, so I don't see a downside to using the old Datamover temporarily, the downside would be the old datamover is slower as it reads all blocks. Not sure if our Veeam Backup uses the Datamover, I think they have there own
But Im a little nervous as im not seeing to many people using this method and im wondering if I missed something
The plan was to do it over Thanksgiving weekend, we operate 24/7 but that weekend would be a light weekend
Incase you need the info
We are running VMware 6.0 U3 with VMFS 5 Datastores
Wrote allot, sorry
Thanks in advance for your opinions
Mike
Hi, how'd it go? Unfortunately the situation is still the same in 2022 and vmfs6 in vmware 7 is still the same. When I had vmware 5.5, I had one datastore with vmfs3 on which I had a different blocksize. This is not possible anymore and as far as I know there is no method to ONLINE shrink a thin allocated disk. So your method described here is still a topic... Thank you.
Jan.