VMware Cloud Community
brantn
Contributor
Contributor
Jump to solution

Thick to Thin sVMotion

I am trying to reclaim zero space on several Windows VMs. I have already ran sDelete and completed a storage VMotion however the VMDK stays at full storage usage even though it is thin. I can run punchzero and reclaim the space however I am looking for a way to do it without the down time. We are running ESXi6.5 with a mix of VMFS5 and VMFS6 datastores.

https://vmware.agentaccess.it.com/solution/SOL-2337

The above article and many others state that it should be possible so I am wondering what I am missing in the process.

Reply
0 Kudos
1 Solution

Accepted Solutions
brantn
Contributor
Contributor
Jump to solution

I have located the issue and what is now working for me. It appears that some of our VMs were already attempted to be made thin by vmotion prior to sdelete being run. Block size and VMFS version doesn't matter in this case.

So the process goes

1.) Verify VM is Thick Eager Zeroed if not vMotion to temp datastore and provision as such.

2.) Run Zeroing.

3.) vMotion to permanent datastore and provision as thin.

4.) Refresh datastore view to show storage usage correctly.

Thanks for the help.

View solution in original post

Reply
0 Kudos
6 Replies
rcporto
Leadership
Leadership
Jump to solution

You will may need to use the vmkfstools with the puchzero option, since looks like you're migrating between VMFS with the same block size, and based on the following VMware KB article, it will not reclaim null blocks, see: Storage vMotion to thin disk does not reclaim null blocks (2004155) | VMware KB

---

Richardson Porto
Senior Infrastructure Specialist
LinkedIn: http://linkedin.com/in/richardsonporto
Reply
0 Kudos
brantn
Contributor
Contributor
Jump to solution

True but correct me if I am wrong there is not a way to change block size as it is hard coded 1MB. I have got it to work on a few vms but haven't been able to figure out the difference yet.

Reply
0 Kudos
rcporto
Leadership
Leadership
Jump to solution

With release of VMFS5, the block size is fixed to 1MB... For that virtual machines that works, can you confirm if the initial VMDK was thick? And for the virtual machines that didn't work, the initial VMDK was already thin?

---

Richardson Porto
Senior Infrastructure Specialist
LinkedIn: http://linkedin.com/in/richardsonporto
Reply
0 Kudos
brantn
Contributor
Contributor
Jump to solution

These disks were all thick when I started however it is a possibility that some of them started out as thin then were converted to thick.

Reply
0 Kudos
brantn
Contributor
Contributor
Jump to solution

I have located the issue and what is now working for me. It appears that some of our VMs were already attempted to be made thin by vmotion prior to sdelete being run. Block size and VMFS version doesn't matter in this case.

So the process goes

1.) Verify VM is Thick Eager Zeroed if not vMotion to temp datastore and provision as such.

2.) Run Zeroing.

3.) vMotion to permanent datastore and provision as thin.

4.) Refresh datastore view to show storage usage correctly.

Thanks for the help.

Reply
0 Kudos
dcdaniel
Contributor
Contributor
Jump to solution

This worked for me, thank you! Just to clarify 'run zeroing' means:

1. Download SDelete... you can google it or get it here if this link works https://docs.microsoft.com/en-us/sysinternals/downloads/sdelete

SDelete - Windows Sysinternals | Microsoft Docs

2. Run (from CLI) 'sdelete64.exe -z c:' (or sdelete.exe on 32bit systems, change drive letter as appropriate for the drive you're zeroing)

Hope this helps someone!

Reply
0 Kudos