VDAP - v-storage backups - HP Data Protector

[ja5e]

Hi all,

I've been asked to look into the possibility of backuping up our 200+ VMs using v-Storage API for Data Protector (HP Data Protector 6.2).

I've read a number of whitepapers, but still can't seem to find the answers to the following: -

• Licencing - Is is on a per-host basis as oppose to per-guest?

• Backup Process - I see that the process is based on shapshots, does anyone know if these are written into Data Protector on the fly or are these stored onto shared storage before transferring to Data Protector?

• Storage - we have very limited SAN storage to accomodate snapshots, so does the process create the snapshot, transfer it, then remove the snapshot sequencially per VM? I'm trying to get an idea of how many snapshots i might potentially have to store on a daily basis (will i have 200+ snapshots?).  Also, does anyone know how big the snapshots typically are?

Does anyone currently have a similar running scenario?  If so, any ideas how long it can typically take to backup around 200 VMs via VDAP?

Any help would be appreciated.

Thanks

Jason

[scottm_occ]

Interested in this is well, sorry I don't have any helpful input.

I have held off on updating to 6.2, as an HP consultant told me over the summer that it had some bugs still to work out with the VDAP module.

I am looking to revisit it.

[Jenni_S]

Hi Jason,

• Licensing - you will need 1 x B6965BA (DP Windows/Linux Online Extension) per ESX host
  
• Backup Process - provided you are running vSphere the method is on-the-fly utilising vmware's vStorage API (DP VE Integration)
  
• Storage - At the start of the backup job an ESX snapshot is created for each VM in the backup job. Then the VM's are backed up on-the-fly either via Ethernet, iSCSI or FC (depending on the configured backup transport mode) and once all vm backups are complete the snapshots are removed again. You can back up multiple VM's concurrently, the average throughput per vm detemines how many you can run concurrently without crashing the DP VE Agent. The size of the snapshot depends on how much your VM changes during the backup window (and also how many vm's you have in the same backup job as they all get snapped at the beginning and all snapshots are retained until the last vm backup is complete)
  

"Does anyone currently have a similar running scenario? If so, any ideas how long it can typically take to backup around 200 VMs via VDAP?"

I am a DP consultant and have implemented the VE Integration a lot, usually on environments of a similar size you mention above and what I've found is that the backup throughput varies greatly depending on the storage that the ESX datastores reside on, the transport mode, the available RAM on the backup host and the destination device.

Jenni

[Jenni_S]

Hi Scott,

You need to go to DP6.21 and also log a support call with HP and ask for patch QCCR2AE35623_TM2 - once you've got these installed the VE Integration is fairly stable.

Jenni

[caklov]

Only when VM snapshot handling mode is OFF (disabled) or the changed block  tracking is ENABLED the snapshot get's deleted after backup.

Otherwise it would depend on the snapshot mode how many snapshots are kept (eg in mixed mode up to 2 snapshots can be there).

As I understand the snapshot is kept to provide the base until next backup (incr or diff) will be taken.

Also I think that TM2 patch has been included in the latest patches for DP6.21 (6.20) - DP_WIN566.

Cheers.

Peter.

[Jenni_S]

Hi Peter,

Yes you are right, sorry I should have been clearer in my post. Full Backups delete the snapshot at the end of the backup but if differential/incremental backups are required then the conditions you state must be true before the snapshots are removed.

The QCCR is indeed included in GR patch DPWIN_00566 (but this GR patch was not yet released at time of my original post)

Thanks for the clarification 🙂

jenni

[caklov]

Thanks Jenni,

and yes can confirm the backups and restores with that patch are working like a charm. Also tested diff and incr backups with no issues.

But one thing I'd like to ask - do you know if there's any way to get backups/restores of particular files done? Except of VCB as that's not gonna be used in newer versions of ESX vSpheres.

Cheers.

Peter.

[Jenni_S]

Hi Peter,

With DP6.21 have 2 choices for restoring individual files using Data Protector.

1) Put a disk agent inside the VM and back up all the files across the network (slow but free 🙂

2) Purchase a DP Vmware GRE license per ESX host and use Data Protector Granular Recovery Extension for Vmware

With GRE what happens is as follows:-

You need to install the GRE Web-Plug in on the VC and this adds an extra Tab into the Vmware Console called HP Data Protector as well as adding an additional Administration section - HP Data Protector (Admin). Then for each VM that you have backed up using VE Integration you can go to the VM in the VMware console, select the HP Data Protector Tab and browse the VE backups that were run for that particular VM. You then select the backup you wish to restore an individual item from and request that the vmdk file be restored. Then the Vmware administrator uses the HP Data Protector (Admin) section to actually restore the vmdk dump to a "Mount Proxy" (Ideally this Mount Proxy is the same server as your VE backup host as this will ensure the quickest restore). The Mount Proxy must have sufficient storage space to hold all the vmdk's that you are restoring.

Once therestore is complete the vmware operator can browse this vmdk restore (again through the HP Data protector Tab in the VMware console, select the VM that is applicable for the restore) and can select individual files to be restored. They can be restored back to the same place or to a different place and you can choose to preserve the tree or not.

The Vmware GRE part code is TB737AAE, a Granular Recovery Extension (GRE) license is required for each ESX server that hosts one or multiple VM virtual machines during backup and which files are recovered using GRE.

Of course there is nothing stopping you manually restoring the vmdk to disk and using vmware converter to get this back into vmware under a different name and IP address then simply locating and copying the requested restore files back to the original vm. If budget is tight then this is the way I would do it, if budget is available I would recommend GRE it works really well.

There is an HP whitepaper available for Vmware GRE which i've attached so check it out if you think that this might be of interest to your organisation.

cheers,

jenni