Creating virtual machine with existing data

[phlight]

Hi All,

I have virtualized nearly all of my servers except for the one and only file server.  I'm running vsphere 4 with 2 ESXi hosts and an Eternus 2000 model 100 SAN and using Veeam 6 B&R

Current setup

Physical server with 2 NICs

1 NIC to LAN

1 NIC to SAN using MS iSCSI initiator

Windows 2003

There are 2 volumes on my SAN that are connected via MS iSCSI initiator that contain the data for the file server.  The file server shares out files (NTFS) stored on the SAN and they have a lot of NTFS permission set on the files and folders (about 1.5 TB of data).

Ideal setup

1 virtual server

1 vNIC to LAN

2 volumes from SAN shared out via the server

Windows 2008 r2

My question

Can I just add the volumes to vsphere, create a VM, add the two volumes to the VM as 2 hard drives, setup the shares and be done with it?  I'm wondering where the permissions are controlled and if those steps can be done without wiping the volumes.  I don't have another 1.5 TB of free space on the SAN.  I do have a Netgear NAS that I use as backup target if needed but I'd prefer not to have to move the data around if it's not needed.

I hope I covered all my bases to make this situation easy to understand and hopefully answer :smileyhappy:.

Thanks,

[weinstein5]

Welcome to the Communiry -  Yes you would be able to present the NTFS to the VM hosted on you server through the use of a RDM (Raw Device Mapping). The permissions on the NTFS volumes are stored within the NTFS file system -

[rickardnobel]

Another option is to continue to use the Microsoft iSCSI client from inside the virtual machine. However, it will be a more "clean" solution to use RDM as weinstein suggested above.

As long as you do not copy the data around you will not have to worry about the NTFS permissions at all.

[cmcminn]

Hi phlight,

Welcome aboard!

You are going to need 2, possibly 3, volumes from your iSCSI SAN to make this work if I am reading correctly...

potential volume 1 - You are going to potentially need to have a volume (datastore) for your VM(s), your W2k8 server.  If you are going to add a second ESX server, putting the VMs on the shared SCSI storage would do you good up front.  This would give a ton more agility in your environment and some of the more advanced feature when you add a second ESX box.  Obviously, you could move these at a later date.

volume 2/3 - These are the volumes with your file shares... You have two great answers above.  Either use RDM or the MS initiator built into server.  You should be good either way here.

-CM

[phlight]

Thanks for all the quick replies!

@rickardnobel I was originally thinking of the Microsoft iSCSI solution but figured there had to be a better way.  I understand how using an RDM is a cleaner way.

For anyone else coming across this thread and using Veeam.

http://www.veeam.com/blog/rdms-explained-for-veeam-backup-replication.html

Key points:

The RDM must be in virtual mode and not use independent disks.

If the RDM is larger than the maximum VMDK size of the datastore that hold the VM’s configuration file; performing a snapshot on the VM with an RDM will fail.

Second question

I know that the max size of a datastore in vsphere 4 is 2 TB if configured with 8 MB blocks.  I'm not sure I understand this "If the RDM is larger than the maximum VMDK size of the datastore that hold the VM’s configuration file; performing a snapshot on the VM with an RDM will fail."

Does that mean the datastore that holds my Windows 2008 server with the 2 RDMs needs to have 8MB block size to allow for my 2 1.33 TB RDMs to be backed up?  All my datastores are using 8MB block size.

[rickardnobel]

phlight wrote:

Second question

I know that the max size of a datastore in vsphere 4 is 2 TB if configured with 8 MB blocks.  I'm not sure I understand this "If the RDM is larger than the maximum VMDK size of the datastore that hold the VM’s configuration file; performing a snapshot on the VM with an RDM will fail."

Does that mean the datastore that holds my Windows 2008 server with the 2 RDMs needs to have 8MB block size to allow for my 2 1.33 TB RDMs to be backed up?  All my datastores are using 8MB block size.

To be able to take VMware snapshots of a raw LUN the "extra data" must be stored somewhere. Since the raw LUN is not read/writable by ESXi (which does not the filesystem of the RDM LUN) the snapshot data is stored on an ordinary VMFS datastore.

In theory a snapshot could grow to the whole size of the disk being snapshoted, say if we took a snapshot, then formated the drive and replaced with totally different data. To be able to handle that the snapshot file must be able to in theory grow to this size. This means the VMFS LUN where the snapshot data is stored must be able to hold such large files, i.e. be using a larger blocksize.

If all your datastore is using 8 MB then this is not an issue for you.

[phlight]

Does the VMFS datastore needs to have enough free space on it to handle the amount of changed data (on a bit level?) between snapshots then?

[rickardnobel]

phlight wrote:

Does the VMFS datastore needs to have enough free space on it to handle the amount of changed data (on a bit level?) between snapshots then?

Yes, so it will be good to not keep that snapshots too long, since the raw LUN will be "read only" and all writes be done at a delta file on the VMFS datastore.