VBR6: zeroed disk blocks and .VBK files

[Frosty]

Just a query about how VBR 6 handles zeroed disk blocks. Given the following scenario:

(1) a VM is backed up and certain disk blocks are in use (so they are stored in the .VBK file)
(2) some time later when files are deleted from the VM and the disk blocks are zeroed (e.g. via SDELETE.EXE)

which of the following would be true:

(1) the next backup will remove those blocks from the .VBK and the size of the .VBK will shrink; or
(2) once those blocks cycle out of the incremental backups (not referenced in the current backup or in any reverse incrementals), they will be removed from the .VBK and the .VBK file will shrink; or
(3) those blocks will always be referenced in the .VBK, at least until such time as we run a fresh, clean, Full Backup and create a brand new .VBK file

[Gostev]

(4) these blocks will be marked as unused and will be reused by the new data during the following runs. There is no way to shrink the file without recreating it...

[Frosty]

Thanks Gostev. If I understand your (4) correctly, its a variation of my (3), so if I want to shrink my .VBK files I will need to run a Full Backup.

New question: does this principle also apply for:
(a) disks which were once backed up and since removed; and
(b) VMs which were once backed up and since removed?

Is a fresh Full Backup required in those cases also if the .VBK file is to reduce in size?

[Gostev]

Correct.

[Frosty]

Thanks. If I can make a suggestion for a future release of the product, it might be helpful if each backup job reported on the amount of "unused disk space" in the .VBK file at the end of the job. This would help in assessing whether a Full Backup would bring a benefit by shrinking the .VBK file.

[Fetaboy]

Dear All,

I'm using Veeam Backup 6.1 patch 1

I tried to shrink the vbk file without any luck.

I backup the file server which is thick 1TB. Once was grow up to 800GB the used space so the vbk file was increased. (It is normal)
But when I removed the "big files" from FS we had only 300GB used space.
The problem is the vbk file is still 800GB, never shrink.

I did active full backup on the existing job. Nothing happend.
I created new job with full backup. Nothing happend.

The backup still 800GB (vbk file).
Seems to be the veeam still see the previously used space.

Anybody has idea?

Thx

[foggy]

Ferenc, you need to run sdelete on the VM first to zero all unused blocks. Please check this topic also for better understanding. Thanks.

[Fetaboy]

Thanks for your quick answer.

I've never used sdelete. Could you give me some explanation about it?
Should I run sdelete on VM or on the vbk file?

Can you write some example? I have to run on live system.

[dellock6]

Please keep in mind, sdelete is NOT to be executed on a thin disk, it will inflate it to the maximum size you configured it.

Luca.

[foggy]

You should run it inside the VM, over the disks you are going to zero unused blocks on. Something like "sdelete -c C:\" (anyway, the tool has embedded help available by using the /? key).

[Fetaboy]

The actual disk is 1TB thick provisioned on EQLogic datastore. This is the E: drive on Fileserver which is W2008R2.

On Windows I've seen only 300GB used space not 800GB from 1TB.
I don't really understand. So the problem is coming from VMWare not from Veeam?

[foggy]

The "problem" is coming from how NTFS works.

[Fetaboy]

I've got it now. Thanks for you guys.

One more question. Can I run on live system? The users using the FS. Will they feel any performance drop down?

The drive is thick provisioned lazy zeroed. I think is that ok for sdelete.

[Vitaliy S.]

Yes, feel free to run it on the live system. I haven't noticed any performance impact on my Windows Servers while playing with this command.