2012-05-08 08:50 AM
I have a FAS2040 with dual controllers and 36 x 15K SAS 600GB Drives. This FAS2040 was originally setup for a large VDI deployment. All of the usable drives are lumped into one large aggregate. The second controller is basically just used for HA. The FAS2040 is now used for VMWare Cluster running only servers, over NFS shared storage. Should I go through the trouble of splitting the system into two aggregates, one on each controller? I already know what I will have to do to perform this action. Luckily there are only 6 VMs currently on the VMWare cluster, so now is the time to decide before it grows (already outstanding requests for several more servers, one being Exchange 2010). My thoughts behind this is that with 15K SAS drives, performance will not be an issue with the drives if we split them between controllers. Adding the second controller as active will give more performance in the long run.
Solved! SEE THE SOLUTION
2012-05-08 09:05 AM
The second controller is basically just used for HA.
So are you saying it has something like 3 drives assigned to it to keep root volume on it? If that's the case, you can either keep it as is (your usable capacity is probably slightly higher in this combination), or do some juggling (basically destroy the 'big' aggregate) & re-balance the disks across controllers, possibly giving benefit of balancing workloads across both of them - but you need at leas two datastores (or other non-VMware workloads) to use this benefit.
2012-05-08 09:17 AM
So are you saying you have dedicated aggregates on both controllers? That's a bit odd on a fairly small system.
I'd say 'usual' way of doing things from scratch is as follows:
- 18 disks assigned to each controller, with 2 hot-spares per controller
- single RAID-DP aggregate per controller, containing both user data & root volumes.