2012-07-15 11:07 PM
We've been trying to move a few SAS draws between different netapp storage systems. The disk shelves contain the aggregates and volumes are wanted to move. I'm just wondering if it's possible to move the shelves and miraculously see the aggregates and volumes on the new netapp controller.
2012-07-16 02:35 AM
Yes, it is. You will need to assign disks to new controller after which aggregates become visible. It is advisable to rename aggregates and volumes in advance to avoid name conflicts with new system.
There are several considerations when doing it though, and it requires downtime at least on source system.
2012-07-16 03:14 AM
As mentioned above, you will want to rename volumes (where possible). If not then when you move the shelves, and the existing aggregates are brought online, any volumes with identical names will be renamed, to avoid conflict.. eg, vol0 would be renamed vol0(1).
Before you move the shelves you should remove disk ownership from the old system and on the system the shelves are to be added to, turn off auto disk-assign, so they can be added manually, when you are ready.
2012-07-16 04:44 PM
I'm just not quite sure what you mean by "remove disk ownership from the old system and on the system the shelves are to be added to". How is this possible if we want to retain the aggregates and volumes on the shelves that we are moving to the new storage cluster. I was planning to just turn the shelves offline without doing anything prior as the aggregate and volume names are different to the existing aggregates and volumes on the new storage cluster.
After the shelves are connected up to the new storage cluster, I guess I'll need to reconstruct the old aggregates by assigning exactly the old disks to the reconstructed aggregates. The only issue I have is wouldn't the filer asks us to zero the newly added disks.
Thanks for your input.
2012-07-16 05:54 PM
ONTAP does not zero drives unless you issue a disk zero spares command or if a non-zeroed spare is used (for a failed drive or added to an aggregate). It is easier to disk remove_ownership (or disk assign -s unowned) the disks in the aggregate prior to connecting to the target controller then you can assign them after moving. To remove ownership and the shelves the system that owns the disks needs to be brought down since there is no hot removal of shelves... Like already mentioned if a head swap, then just reassign all at once after swap...but if not all disks, I prefer to maintenance mode boot and remove ownership before connecting to the new controller.
2012-07-16 10:10 PM
Removing disk ownership does not destroy information on disks (although in this case I may not be needed - rather it makes sense to immediately reassign disks to target system).
You need not to create aggregates on target system - doing this will indeed destroy all previous disks content.
Did you consider ordering this job from NetApp or partner professional service?
2012-07-16 10:18 PM
Data ONTAP will automatically start zeroing disks if you add non-zeroed spare to aggregate.
There is no need to remove disk ownership in this case. As source system must be shut down anyway to remove shelves, it is easier to assign disks in maintenance mode on source to correct target system before removing shelves and then simply add them online to target. One just have to be careful to add all disks at once. But it is near to impossible to assign disk ownership to all new disks at the same time, which will result in scary “errors” about incomplete aggregate and possibly aggregate reconstruction.
There are other ways to do it, that why I said initially that there are some points to consider. And it is better done by professional service which has experience with NetApp.
2012-07-16 10:39 PM
There's no urgent need for this as we can migrate this volume by volume to the new storage system. Just wanted to see if this is an option later on down the track if we don't have any spare storage space.
2012-07-16 10:54 PM
By the sound of it, it's not possible to do what I was initially wishing for:
Sorry, I didn't make myself clear. Basically, we wanted to move a few aggregates with volumes (not entirely) to our new storage cluster to relief the load on the old system. The aggregates and volumes are all inside 2 SAS shelves. My intention was to simply (not so simple now) move the shelves and connect them up to our new storage cluster and at the same time keeping the old storage online at all times. If possible, this will save me time from migrating all the data to the new storage which is time consuming.