VMware Data Recovery - Disk Retention/Restore

[Ward8124]

Hi all,

I'm running a proof of concept environment to trial VM DR and so far having a lot of joy in testing the features however I came across a potential show stopper for prospective users/clients who might be considering using this solution but wanted to run it by you guys/girls to see if there was a way round the problem.

My problem is that some clients might not have large storage arrays to store the backups on the LUNs/RDMs presented to the Recovery Server so would probably look to reducing the retention settings to allow a certain amount of daily/weekly jobs on disk. In addition to this I can foresee customers wanting to back the backup data LUNs to tape to adhere to company policies/laws etc.

So in light of this, if I wanted to backup the datastore/rdm containing the backup data from the vDR server to tape and at a later stage carry out a restore (i.e. a VM/file which has expired beyond the retention set by vDR server); how would I go about importing the files/data back on to the server and then see these backup objects in the vDR server?

Is it possible or a functionality I've missed or is there a work around for this?

Your thoughts and suggestions are always welcome!

Regards

Don

[alefestaedist · Accepted solution]

The process to recover a missing job is to restore in the lun/RDM the catalog and the backup files, that it means all the files in the VmwareDataRecovery folder since the vDR read the catalog will find again all the backups and will be able to restore the files.

Different is the case where you deleted accidentally the vDR, basically the main problem is that the vDR do not refresh information automatically from the restore point. It would be nice to "force" this reading but to do so one should be able to log on to the appliance.

alex

[alefestaedist]

I'm doing some test and what I see is that you have to create the same folder/share in order to help the vDR to recover the files..I suppose that inside the console there is a way to "force" the recover will post later a solution.

[Ward8124]

Hi guys, you are both correct, I've managed to find some information on these forums that provides a procedure of pretty much exactly what we/I/you are trying to achieve.

Please refer to this link for the procedure as provided in a document on this forum:

http://xtravirt.com/xd10131

The only issue is that the VC DR server requires to be shutdown while a backup is taken (according to the guide) so this will need to be scripted as pre-script and post-script.

I hope this is of use to anyone else thinking about the same scenario.

Regards

Don

[HMUSA]

I'm right there with you.  Just starting up with vSphere, small business.  Though I put the VM's into production right away.  We're using direct attached storage, so I can't access the datastore over the network. Hopefully the powers that be will see this is a demanded feature: getting the vDR server image off the datastore onto removable storage. 

Have you tested pushing the datastore onto a network drive somewhere else, like smb:\\server\share ?  vDR gives a message saying it might not be reliable, so I chickened out and have it on the DAS.  But it would allow easier backup to tape from there outside of the VM datastore.

If that doesn't fly, I'm looking at scheduling the power-down of the vDR / a third party tool like Veeam FastSCP scheduling the copy off datastore / schedule the power-up of the vDR / and a traditional backup system which puts that onto tape or a rotation of portable USB drives.

I think that would be my pipe dream: a scheduled component to vDR which puts the necessary files onto a USB drive rotation (connected to one/any of the Hosts in a vSphere Datacenter.  I could see people wanting tape, but that's a whole 'nother ballpark of compatibility.  USB is pretty ubiquitous though, and most of my stuff could fit within current 2.5" drive capacity: <1TB.

[HMUSA]

Okay, I've hit upon USB Passthrough.  Maybe I can connect one of my USB drives to a host.  Auto-mount it onto a VM, and use a third party utility to copy the powered-down vDR image from the datastore onto the USB drive.  That way it doesn't have to go out onto the network as my previous post's plan would.

OR maybe I could pass the USB drive through to the vDR machine itself as its data location.  Has anyone tried that?  It wouldn't be getting the whole vDR VM, but it would get all of the deduplicated, compressed, historic backup data.  I think having a rotating set of USB drives would confuse DR though, and that's kind of the whole point.

Any ideas/experiences much appreciated.

[alefestaedist]

Though you may use VTL if you got a storage that support this tecnology, so you may simply present a disk to the vDR but it will be saved to the tape directly, this way you will avoid the risk of "confuse" the appliace changing the usb from time to time.

[Ward8124]

Ok I've managed to get this sorted to what I'd like and it's fairly simple. Basically I present an RDM in virtual compatibility mode and use this as the repository. I use my third party backup software to backup the repository only (default vmwaredatarecovery folder) to tape. If I need to revert back to a restore that is past the retention I set in the configuration I carry out the following:

1. Roll out a new OVF to install a new data recovery server

2. Configure the new VCDR server as per the guide with a different IP and DNS name if possible - if not shutdown the existing server (as long as it's not running any tasks that is) and use it's existing IP/DNS

3. Open the new VCDR server using the plugin on the VI Client

4. Mount the RDM or Network share of where you restored the data from the tape to.

5. Perform an integrity test on the restored repository

6. Open up the restore tab and you should see the restore points along with their time/date stamps and "should" be able to restore these.

I've tested this and it does work everytime but can't say it will for you.

HMUSA - In my opinion trying to backup the repository to a USB Drive presents a big security risk unless you have an encrypted USB as these can go walkies fairly easily and in addition to this I think you will have space and performance issues in reading and writing from this media. It does of course depend on the amount we are talking about in relation to performance though. I can't say rotating the drives will be a good idea unless you perform an unmount of the drives and remount of the new one and effectively wipe the slate clean but again I can see problems with:

1. Performance as you will be effectively having to do a full back up everytime you do this and then incrementals there after (de-dupe)

2. I've seen and had problems mounting new storage areas as the backup jobs fail due to the the destination repository being different which in turn means you need to re-do all the jobs and select the new destination.

3. I wouldn't want to rely on these USB hard drives as the media in which to keep critical backups as I've been burned before when these drives fail and therefore all is lost.

4. Using USB pass through would cause the host to use more resources to write the data on so therefore may impact the machines on this host.

5. I can't remember off the top of my head but I'm not entirely sure if the the VM i.e. the DR server, is vMotioned it will still keep it's mapping to the USB but will need to check that.

Just a few thoughts should you go ahead with this approach.

[alefestaedist]

Good job....thanks.

Alex