Aggregates made up of disk spanned accross Shelves

[TWEET2ARUN]

HI friends,

I am confussed while creating an aggregate i want to know is it a best practice to create a single aggregate from a single disk shelf only or can we create a single aggregate having multiple RG whose disks are spanned accross different disk shelves.

Regards

Arun

[resqme914]

It doesn't seem to matter anymore which slots/shelves the disks are located.  Best practice is to make sure the disks are the same type, speed, and size, and that your raid groups composing the aggregate have the same number of disks.

[martin_fisher]

Hi Arun, in a nutshell, it doesnt (and shouldnt) really matter.. the aggregates can span multiple shelves, as long at the disk are the same really. Also think of it another way, say you did create  1 aggregate on 1x shelf of FC and you had another shelf of FC with another aggr, You then have a disk failure, or maybe 2x disks fail in 1 aggregate. ONTAP will assign the all the spares available (of the correct type), to rebuild it. The planned disk layout would go out the window, as spare disks of the same type are used to rebuild the aggr.

Martin

[lmunro_hug]

I prefer having my aggregates designated to specific set of shelves disks where possible, I then know which disks are in what aggregate just by looking at the system.

Also, if a shelf lost power or connectivity for some reason (example) I would rather have 1 aggregate go offline rather than multiple.

Luke

[billshaffer]

If you have enough disks/shelves, you can play with your raid group size and set the raid groups up so that a shelf failure will only take out a max of 2 drives per raid group, preserving data.  The only caveat there is that, like Martin pointed out, a disk failure will pull a spare from wherever and use it, thus ruining the protection.  I get around this in my environment by doing a disk replace <old spare/new data disk> <new spare> after I swap out the bad drive.