From time to time I am in the situation where I need to get data back from storage-side snapshots of cluster shared volumes. I suppose I just never figured out a way to do it right, so I always needed to:

  1. expose the shadow copy as a separate LUN
  2. offline the original CSV in the cluster
  3. un-expose the LUN carrying the original CSV
  4. make sure my cluster nodes have detected the new LUN and no longer list the original one
  5. add the volume to the list of cluster volumes, promote it to be a CSV
  6. copy off the data I need
  7. undo steps 5. - 1. to revert to the original configuration

This is quite tedious and requires downtime for the original volume. Is there a better way to do this without involving a separate host outside of the cluster?

  • You can have multiple CSV mount points/volumes per cluster so why not just add it as an additional CSV? – joeqwerty Oct 31 '13 at 20:45
  • @joeqwerty because it won't let me. As it is a snapshot of an already present CSV, it has identical IDs (whichever the cluster manager would be checking), so it is not listed as a volume for addition as long as I do not offline and un-expose the original CSV. – the-wabbit Nov 1 '13 at 9:08
  • OK, understood. Is this FC or iSCSI? If iSCSI what about attaching it to a VM? – joeqwerty Nov 1 '13 at 13:37
  • @joeqwerty It is iSCSI. Can I simply attach a CSVFS volume to a non-cluster host and expect it to mount as NTFS? – the-wabbit Nov 1 '13 at 13:51
  • I'm not a Hyper-V expert and I'm certainly not a CSV expert but when I remove the CSV from my test cluster for my iSCSI LUN it shows up on my host as an NTFS volume and I'm able to access the files on it. I think CSVFS is an additional file system "driver" or "filter" that's only used when the volume is in use as a CSV, so in other words it's another layer of abstraction at the file system level. If the volume isn't being used as a CSV then the underlying file system is plain old NTFS. – joeqwerty Nov 1 '13 at 14:13

Sometimes vendors can integrate with the windows "previous versions" feature to allow you to grab data from snapshots. If that's supported with your storage, that would be the ideal. Otherwise, you'll have to either figure out a way to mount the shadow copy in read-only without having it cause trouble to the server or the R/W production volume.

  • My storage doesn't support this. As far as I know, the "previous versions" stuff is not supported for CSVs anyway (shadow copies in general are). The main trouble here though is that due to the overlap in IDs between my snapshot LUN and the "original" LUN Windows would not let me add the snapshot volume to the cluster along with the original one present. Knowing which IDs need to change on the snapshot to allow for this would solve the problem already. – the-wabbit Nov 1 '13 at 9:15
  • How about a read-write snapshot? Can you change the ID on the snap, allowing you to import it? – Basil Nov 1 '13 at 17:59
  • I could, if I just knew what to change. – the-wabbit Nov 1 '13 at 19:42

In doing some research and testing it appears that the underlying file system of a CSV is NTFS, so you should be able to add the LUN for the snapshot to a VM via iSCSI and access it as you would any other NTFS volume and grab whatever files you need off of it. This eliminates the need to down your cluster and offline the active CSV LUN.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.