I might be missing something, but I've not found definitive answers on the following:
With a FAS2240-2, can I take a single DS4243 or DS2246 with 24 drives and configure two aggregates, each comprising of 12 disks, one assigned to controller A and the second aggregate to controller B? In other words, can a shelf be configured to serve individual disks to specific controllers or are shelves themselves bound to a single controller?
Can a single aggregate comprise of disks in both a FAS2240-2 and a disk shelf? I can see reasons why this may not be desirable (such as the ability to migrate a tray to another filer), but can it be done and is it a supported configuration?
Do the answers to the above, pertaining to the FAS2240-2, also apply to the FAS2220?
Thanks for any help!
Yes to all the above. Netapp uses disk assignment and a shelf can be mixed between controllers any way you want. 12/12 5/19 or all to one controller. One consideration is that when a shelf stack does have a mix of both controllers, then that stack does not utilize disk auto assignment. So even if enabled and you replace a failed disk, just remember to manually disk assign that drive to the correct controller that needs a spare.
Thanks for the replies.
It's a theoretical design at the moment. I thought I read somewhere that an ideal RAID group size was 16 (14+2) disks (is this right?). Therefore, in a FAS2240-2 with two shelves of 24 disks, I could have an aggr0 comprising of 32 disks (2 x 16 disk RAID groups) assigned to controller A and an aggr0 comprising of 16 disks on controller B. This was dependent on being able to split the 48 disks in the shelves across the two controllers.
Would this be a good idea? The challenge here is that there is no space in the shelves for spares. Presume these could be internal?
On a related note, do people reserve disks for system volume on each controller? Would it make sense to dedicate disks for this (internally?) such as a 3 disk aggregate?
I've not had luck finding any design documents that cover this subject, so any pointers would be appreciated!
an ideal RAID group size was 16 (14+2) disks (is this right?)
Not really - it is more important to have RGs of the same (or similar) size, but it could be anything in the supported range.
there is no space in the shelves for spares. Presume these could be internal?
Spares could be anywhere as long as there are assigned to the right controller (they are not system-wide, just controller-wide, so you need spares on both heads)
do people reserve disks for system volume on each controller?
It can be debated, but generally on smaller systems it doesn't make sense - root volume sits together in the same aggregate with user data.