recovering from a snapshot

[simon_164]

Anyone knows how to recover from a snapshot? if you have the data vdisk and then you find out it is corrupted. how to restore data from the snapshot. If somebody did it, can you be specific about all the steps.

[Ivan Ferreira · Accepted solution]

Running on? With newer versions of EVA firmware (V6), you can recontruct the original lun from the snapshot.

To restore data from a snapshot, depends of the snapshot type.

If the snapshot type is demand allocated probably is better if you create a new lun to restore the data from the snapshot.

Anyway, you can:

Destroy all data in your original lun, that could be by formatting (format/mkfs) the original lun.

Mount the original "now formatted" lun and the snapshot and transfer all data.

If you want a more "quick" way of restoring data, use snapclones. With snapclones, as vdisks are independent, you can just delete the previous snapshot and then mount the clone.

Additional operations may be required depending of:

OS type.
Vdisk data (As in boot from SAN configurations)

Ex.
Windows cannot mount the snapshot/snapclone on the same host (i think).

HP-UX will use another device name for the clone.

Tru64 Unix requires the same WWID to maintain device names.

[simon_164]

Well the Operating system is HP-UX and I don't have the new firmware. The snapshot is an on-demand one.
So you advice to remove all the corrupted data from the volume and then present the snapshot and copy the data from it?
Is it the best way?
Can you think of another?

[Uwe Zessin]

Watch out!
If you copy the data from the snapshot to the original virtual disk, you sort-of turn a demand-allocated snapshot into a fully-allocated one, if the vdisk contains much data.

Why? Because the snapshot must retain the original view at the time the snap was taken.

[Ivan Ferreira]

What Uwe said is exactly the care that you must have with demand-allocated snapshots. But basically your options are:

- Create a new lun, mount the new lun and snapshot and copy all data from the snapshot.

* Advantages: Your snapshot won't convert to a fully allocated snapshot. The storage controller does not have to copy the snapshot data and the "recovered" data at the same time.

* Disadvantages: You need more space.

- Destroy the file system in the existing original lun, mount the original lun and snapshot and copy all data from the snapshot.

* Advantages: You need less space than in the first option

* Disadvantages: As you copy the data from the snapshot to the original lun, your snapshot will increase in space used and your controller probably will do a lot more job by preserving the snapshot and copying the data to the original lun.


You must change the Volume Group ID to mount the snapshot and original lun in the same host, see:

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=1000416

[simon_164]

Thanks guys,

I guess I will ask for more disks and go for the snapclone since it is easier.

Thanks !!