NTFS target and staging to tape considerations ?

[StixForBrains]

I've read many recommendations here that my target for backups should be NTFS rather than VMFS (stability, easier restore etc). I am currently replicating to a SAN at a hot-site. I also want to local backups on FC SAN and then move them off to tape storage. I'm using a Veeam VM to do the replications and backups. I have a physical server (win srv 2003 with EMC networker) with scsi tape library. The physical server has a fibre HBA and I can expose my san luns to this physical server. My veeam VM has a RDM lun attached and formated as NTFS as the target (virtual appliance mode)for backingup the other VM's on the SAN. I think many people here are doing something like this.
My questions are about the next stage:
- how are you mounting this lun to physical server to to backup the VBKs to disk?
- how are you not currupting NTFS?
- are you manually removing it from the Veeam VM? or Scripting it somehow? or leaving it visible to both OSs?
- should the target NTFS RDM lun be a basic disk or dynamic disk?
- other considerations?
- are others doing this successfully without corrupting NTFS? What tool(s)?
- if the physcial srv is seeing lun read-only what has this done to your backup methodology are files not being tagged after backup?

As a side issue I have some concern with the whole concept. As I'm not really happy with my Windows box being able to see all my san luns (automount is disabled but but i hate to think what culd happen) but with IBM DS4700 san its limited way of mapping luns to hosts I have no choice. I've had this verified by a couple storgae specialists. Its the only way that esxi hosts (veeam VM) can see the source and target luns and for the physical srv to also see the target lun is to be in the same san "hostgroup" as the esxi hosts, at which point the physical srv can see everything. I can't put the ntfs lun in two different groups or two different hosts. Jus one host or one one host group. So I worry about windows somehow crashing and corrupting my VMFS luns. But I don't want to spool to tape across network either.
Instead my veeam vm could share NTFS lun as a network share, and the physical srv could grab files that way but then the last stage disk-to-tape happens over network instead of over fibre (yuck) with accompanying congestion and speed issues.

Thoughts? Recommendations?
Thanks in advance
Tony

[StixForBrains]

I came across these two articles/discussions that have a few nuggets re:accessing the same Raw LUN from two different OSs:
http://serverfault.com/questions/16427/ ... stem-setup
http://serverfault.com/questions/16427/ ... stem-setup

But they are more about full Clustering. What I'm hoping to do (disk-to-disk-to-tape scenario) should be simpler I hope.
Anyone with thoughts on my original NTFS questions

My questions are about the next stage:
- how are you mounting this lun to physical server to to backup the VBKs to tape?
- how are you not currupting NTFS?
- are you manually removing it from the Veeam VM? or Scripting it somehow? or leaving it visible to both OSs?
- should the target NTFS RDM lun be a basic disk or dynamic disk?
- other considerations?
- are others doing this successfully without corrupting NTFS? What tool(s)?
- if the physcial srv is seeing lun read-only what has this done to your backup methodology are files not being tagged after backup?

[StixForBrains]

Bumping this question again. Never heard from anyone.
Can anyone using an NTFS pRDM as a target comment on these questions above, and how you connect your LUN to a physical windows server without corrupting the lun or conflicts, or how you mount it "read-only"?

[tsightler]

Take a snapshot of the pRDM that is presented to your Veeam VM, mount the snapshot to the physical server to dump the files to tape. This could be easily scripted. I've been doing variations of this for a decade, with NetApp, EMC, and Equallogic storage. I'm sure the same can be done with the IBM DS4700, but I have limited experience with that storage system.

[StixForBrains]

Thanks Tom,
I've certainly understood other's here to say they connect to THE rdm lun in question. But I prefer your method. I have Flashcopy. Don't have Volume copy but if Under stand ibm flashcopy correctly its suitable for this purpose. Just means I'll need to free up some additional drives equal to my existing backup luns or probably more likely invest inanother expansion rack. The other good thing about this, with the IBM limitations on how host group access is defined I'll feel safer about the physical machine not trashing any luns. I know there is lots here about preventing the physical server from writing signatures to all your luns but just giving it access to one lun will certainly reduce risk of the bizzare. I'll test it on a small backup thanks.