I've done this kind of thing a few times and by far the biggest hassle is dealing with the server side of things. (i.e. getting all of the hosts to look for their luns/cifs/nfs vols on a new head)
Roughly, here are the steps I took for moving some shelves to a new head a while back.
* create new igroups/iscsi security on the destination head
* get a new /etc/exports file ready ahead of time (on destination head) with the new vols that will be exported.
* get new /etc/exports file ready ahead of time (on source head) that has the moving volumes removed from it, so it doesn't complain about missing vols.
* unmap luns that are going to move to new head.
* boot into maint. mode and offline the moving aggregates
* shutdown the filer
* move shelf to new head and cable up
* disk assign
* aggr online
* aggr rename
* map back luns to hosts
* copy over new exports file and export the new vols (exportfs -a)
On Wed, May 18, 2011 at 7:28 AM, Steve Losen <scl@...> wrote:
I need to physically move some disks/aggregates to another
controller. Do you see anything wrong with this procedure?
1) upgrade DOT to the same version on both heads (7.3.3)
2) shut down and power off both heads
3) physically move the disk shelves
4) boot the destination head in maintenance mode and use
"disk assign" to take ownership of the moved disks
5) boot the destination head normally
6) aggr online the foreign aggregates
I am concerned about step 4 above. Will it work or do I need
to boot the source filer in maintenance mode and change the disk
ownership to the destination head before moving the disks?
In step 5 some of the foreign aggregates will have names that
conflict with native aggregates. I presume that when ONTAP
boots it will automatically rename the foreign aggrs as
necessary. Or must I rename them on the source filer beforehand?
I know that some of you do this routinely, so I appreciate
any words of advice.