Highlighted

Thin (NFS) VMDK backs up as thick (sort of), does not happen with VMFS

I have set up a new implementation of vSphere/ESX 4.1, NFS Datastores, plus BE 2010R2 with the AVVI.  When I run backup jobs through the AVVI they take a long time and the data size is the full size of the VMDK, not what is being used.  For example 100GB vmdk and 10GB used = 100GB data size.  I see the IMG file on the backup to disk folder grow to the size of the data in use and not get any larger, but it still goes through the entire disk and takes forever.  Ditto on duplicate to tape.  The same thing is *not* happening with thin-provisioned vmdk's on an iSCSI/VMFS datastore.  A VMDK with 100GB and 12GB used = 12GB data size, and a quick backup and duplicate.

Please tell me this can be fixed or I'm going to have to switch my datastores to iSCSI, which I would prefer not to do on my NetApp filer.  We are comitted to Backup Exec (3 years maintenance purchased) so I can't switch to any other solutions.  Let me know if you have seen this before and can help!  Thanks.

11 Replies
Highlighted

Just for the information of

Just for the information of any other forum members reading this, (orignal poster is already aware via a different communication method) this is a known issue that has been escalated to the Backup Exec engineering team.

In terms of the escalation though it would be useful to have a better idea of how often this configuration (NFS based ESX Datastores) is used and how many customers have seen this issue but not formally reported it to us.

As such please respond on this thread if you have seen these symptoms with this configuration as I will subscribe to the alerts.

Highlighted

I use some ESX on NFS

I use some ESX on NFS datastores (I have NetApp storage also), but currently these are test systems and I'm not backing them up with BackupExec.  I will be moving NFS into production in the near future, however so it would be an issue.  Currently I'm only having the issue with thin VMDKs on iSCSI.

Highlighted

I have since discovered

That this issue isn't NFS vs VMFS at all, but in fact it appears to be an issue with Windows 2008 R2 VM's only.  I can replicate the thick backup problem across vSphere 4.0 and 4.1, NFS and VMFS (iSCSI), multiple versions of VMWare tools.  Windows 2003 VM's are fine.

Highlighted

I have the issue with Windows

I have the issue with Windows 2003 vms.  I'll be opening a support case on this soon since it's not the standard NFS issue that Colin is describing above.

Highlighted

Hi there. Just want to add

Hi there.

Just want to add that I also have this problem.

It doesn't matter the OS on the VM. Even blank VM behave the same. I have the problem with RHEL, 2k3, 2k8, 2k8R2, Ubuntu, etc.

All backups of  VMs stored on NFS with thin provisioned disks result in BackupExec backing up the full disks, and not only the used space.

As soon as I move the VM to a VMFS datastore (iSCSI, FC, Local SATA), BackupExec backs up just the space in use. Move the VM back to NFS and the backup is again the full provisioned disk.

Its way to easy to replicate:

- Create a blank VM with ex.: 8GB thin provisioned disk on VMFS.

- Backup it with AVVI. Backup takes some kb (less than 100KB; this will match the in use space).

- Migrate the VM to NFS.

- Backup it with AVVI. Backup now takes approx 8GB (this will match the provisioned space).

- Migrate back to VMFS

- Backup it again with AVVI. Backup takes again some kb (this will match the in use space).

Installing an OS on the VM will result in the same behaviour, so you don't even have to waste your time trying.

All my hardware is on VMware and BackupExec HCL btw. I even tried a software NFS server running on a VM with the same results.

Btw, I also have an active support case with BackupExec. I'm experiencing this problem since vSphere 4.0 and BE 2010.

Regards.

Highlighted

I guess the following post

Highlighted

As previously stated we are

As previously stated we are aware of this issue

 

And just to clarify it is a Backup Exec/vStorage API problem with an NFS Datastore which will affect the backup of any VM stored inside the NFS datastore and is independent of the operating systems used inside the VMs themselves.

Highlighted

Updates?

Does anyone have any news on this topic.

I am also facing this.

Highlighted

  Same problem for me but i

 

Same problem for me but i use FC between ESX and SAN (and ethernet between esx and backup exec).

ESXi 4.1Backup Exec  2010

:-(

Highlighted

Same problem here, affecting

Same problem here, affecting 12 VM servers. All servers were backing up as thin whilst the VM license was in evaluation mode. As soon as I upgraded the licenses to ESX Enterprise 2 x socket licenses the backups started to backup the thin disks as thick.

If a solution cannot be found soon we will have to go to a different vendor for our backup solution. The backups are now being cancelled after 15-16 hours of running, they used to take 5 - 5.5 hours.

 

ESX 4.0 Update 1

Backup Exec 2010

Highlighted

If you are on Backup Exec

If you are on Backup Exec 2010 and not either Backup Exec 2010 R2 with latest hotfixes or Backup Exec 2010 R3 then you will still experience problems with NFS Datasrores and thin disks - although I am kind of surprised that installing a license would affect the issue.

Even if you are on 2010 R2/R3 with latest updates you also have to make sure that a specific registry key change is made (as per information in http://www.symantec.com/docs/TECH147674) to set "NTFS Used Sector Tracking Preferred" to 1