Largest possible aggregate

[allenford]

Hi,

I have a FAS 3070 with 9 disk shelfs. The shelves each have 14 300GB FC disks, for a total of 126 disks. I am interested in creating the largest aggregate possible. What would be the most effective Raid Group size to use? What are the total number of possible disks that can be used in an aggregate?

Thanks for your assistance,

Allen

[reinoud7]

I have not really calculated, but you have following limitations:

• aggregate size is max 16 TB (in 7.2., this is including the parity disks)
• raidgroup is (DP) is max 28 disks
• there is no limitation about the amount of disks in an aggregate

When you go for a raidgroup of 25 disks, you can 5 create raidgroups and 3 aggregates. You loose then 10 disks for parity and you only have 1 disk as a spare disk. I think (can someone do better?) this is a capacity based optimum. I personally think this is the best solution. I never do this on this way. I try to use the default raid group and for FC disks I will have at least 2 or 3 spare disks for that amount of disks.

[kusek]

Great question Allen,

From a best practices perspective, using 300GB disks, in order to maximize your aggr while butting up against the 16TB current maximum, you're best off creating an aggr with 15Disk Raid Groups

What you will end up with, is 59 disks in a single aggregate providing you the maximum amount of available space.

By this math, you can end up with 2 aggr's eating up 118 disks, leaving sufficient room for spares (3) and a separate root aggr for the system

If this is a cluster, you'll be close on disk count, as you'd need 1 more disk to have dedicated RAID-DP root vols, 59disk Aggr's and 3 spares for 9 shelves

Hopefully this helps address your question Allen,

Christopher

[benn]

Two 28 disk group after rightsizing is just about the 16TB limit but it will span 4 or 5 shelves and u can only make one.

From a redundancy standpoint 4 groups of 26 disks will yield 2 14tb aggregates spanning 48diskd (+4 for parity) each. Manually assigned 13 disks on each shelf leaving 1 disk as a spare on each shelf will speed rebuild time in event of a disk failure. 8 disks are lost to parity and 8 are left as spares. This leaves 14 disks (one whole shelf) for the root volume and any partner parity volumes. It may not be the most efficient use of space but its very fault tolerant.

Break down

Shelf 1 (reserved for root and partner parity)

Shelf 2 and 3 group1 aggregate 1

Shelf 4 and 5 group2 aggregate 1

Shelf 6 and 7 group3 aggregate 2

Shelf 8 and 9 group4 aggregate 2