Reading the documentation, it says that I need to do a shutdown of the filers to remove the shelf. Is there any way to avoid this shutdown? Would doing a cf takeover and cf giveback work? My plan is to :
It won't work and it will lead your filer to panic mode or it will reboot by itself. because of multiple disks missing from the controller. The safe way to do it without downtime is, remove ownership from all the disk belong to the shelf that you want to remove, power off the shelf and remove the cable.
1) destroy the volume and aggregate (if necessary)
2) remove ownership
3) make sure all the disks belong to the shelf that you want to remove are unowned disks (disk show -n)
4) power off the shelf that you want to remove
5) remove the cable
I did not do this exactly as stated as I wanted to experiment a bit. I am confident that it would have worked as well. The way that I did this was to do a cf takeover from A, removed the cable from B while it was booting up, did a cf giveback, then did the same on the other controller.
Just a CF takeover and giveback wont work - as the cabling to your shelf which has to be removed - will still be valid from the partner head and will continue serving data(I am assuming Mulitpathing is implemented in your case which is Netapp best practice).
If you do a TO and then power off the shelf - the disks will go missing from the partner (which is accessing it though MP)and filer will panic.
You should shutdown the filer and then remove the shelf.