Adding a second FC shelf - how to size the RAID group

[ihkberlin]

Hello,

i want to expand an exisiting shelf (14x300GB, FC) with another one. One aggregate should span both shelves.

The Raid size of the existing aggregate now is 16. When i add the second shelf should i change the raid size and if yes to what?

What are the pros and the cons?

How are optimal raid sizes are calculated?

Thanks in advance!

[Darkstar]

We always use RAID sizes of 13 or 14 disks. So, for example, set the RAID size to 14 and do "aggr add <aggrname> 14"

you will get one 13 disk and one 14 disk raidgroup, which is the best option performance-wise (raid groups should be around the same size)

if you have (or want) 2 spares you could set the raid size to "13" and do "aggr add <aggrname> 13" to get two 13-disk raidgroups

And you can use "aggr add -n" to preview the new layout without actually adding the disk

hope that helps,

-Michael

[nigelg1965]

Using the extra disks to just expand the existing aggregate so it spans both shelves, is the right way to do it. Raid group size can't be change after the aggregate is created, so you are stuck with 16. Adding the extra 14 disks means you would start a new plex (raid group). Allowing for 2 spares overall that would leave the new plex at 10 disks with two of those being the parity ones, luckily in Ontap 7.3 onwards the spares are not counted in the 16.

What are the pros and the cons? smaller the raid group the greater the protection.

How are optimal raid sizes are calculated? There are recommended limits depending on disk type (SATA is 16, I think, SAS / FC 28) and an overall limit on the size of the aggregate.

[Darkstar]

nigelg1965 wrote:
Raid group size can't be change after the aggregate is created, so you are stuck with 16. 

This is wrong. You can change the RAID group size any time you want. "aggr options <aggrname> raidsize <n>".

You can even add disks to other raid groups than the last one created by adding the "-r" option to the "aggr add" command.

I wouldn't recommend making RAID size bigger than 16 or so, except if you have specific reasons to do so. Otherwise the rebuild times will get bigger and bigger very quickly

-Michael

[nigelg1965]

You are, of course, right, I mis-read the on-line manuals.

[DARRENBRENT]

is the raid size number modification completely non disruptive to luns/users on the aggregate/system?

Thanks

[eric_barlier]

Yes, its completely transparent/non disruptive.

[jakob_bena]

hi,

i have found a report from netapp. there is a description about how to configure 32 / 64 aggregates.

http://www.scribd.com/doc/57428791/Storage-Subsystem-Technical-Faq-Sep-10

[eric_barlier]

I would do a RG size of 14 unless you re confident you ll be adding more shelves later. By using RG size 14 you get

to fil the shelf up with the aggregate and its balanced with the other shelf. NB: you ll need to set the new RG size before you add disks if you decide to use 14.

Make sure you have 2 spare disks of each disk type if you have different disk types.

[scottgelb]

I'd like to know your current aggr status -r for that aggregate... if you have 14 drives in one shelf you don't likely have all 14 in the aggregate (especially if multiple spare and a separate root aggr) so even set to raidsize 14 it will fill out the existing rg0 raid group by default (the current raidsize of 16 doesn't mean you have 16 drives in the rg yet).  If you want to create a new raid group and ignore the raid size fill out, you can use -g new on the aggr add command or to fill out an existing raid group...and as already mentioned also with -g rgN to add to an existing raid group.  If you have 2 spares already (Eric's recommendation which I agree with) then making a new raid group of 14 makes sense on the new shelf.  Typically, we try to keep all raid groups even size in an aggregate for performance reasons or within 1 drive of each other so knowing the current layout of your aggregate would help in making a recommendation.