I have FAS8060, 3x 24SSD shelves. All this refurbished, second hand. I want to have dedicated disks for root volumes. I know they are partitioned root-data-data, but I don't like the idea mixing application data IO with disk operations related to root volumes.
NetApp automatically creation procedure left me with 10 drives per root volume/node. I have playing this days with manually assigning and creating aggregates and partitions.
What bothers me though is the following:
if I want to have dedicated drives not used for anything else but for root partition(volume) - how much drives do I have to put aside, having in mind that the root partition is always 53.88GB?
Yes, 53.88 is a correct figure for 'ADP: root partition 'size for SSDs for 24 Disks shelf for FAS8060 according to Hardware Universe.
And, the 'min root vol' size as per hwu.netapp.com is: If you are running 7-mode : Min root vol size = 250GB If you are running cDOT/ONTAP : Min root vol size = 350GB
Without ADP: If you can afford to lose 3 Physical SSDs per Node to be allocated for 'root aggr', then no issues and you can go ahead. However, for customers who have only 1 Shelf with 24 SSDs, you lose straight away 6 SSDs for 'root aggr', and you are left with 18 SSDs for Data Aggregate, out of which 2 will be gone for Parity (DP), leaving just 16 SSDs as data disks (considering just one RAID-GROUP).
Perhaps, this was the reason that NetApp introduced ADP (Advanced Drive Partitioning). Within this concept, each node would still have it’s own root aggregate, however ONTAP divides the Hard Drive (or SSD) into partitions (smaller root partition and a bigger data partition).
Again, as I said, if you could afford to lose 3 physical SSDs for 'root aggr/node' purpose then it's perfectly fine.
I perfectly understand the need for ADP. However with these SSD's which in my case are somehow unreliable, I have the following concerns:
Initially the system have chosen to put the root aggregates on the first shelf. 10 partitions per node for root aggregate. As a consequence - if I use the automatic creation procedure - It alwаys create two data aggregates.
Each one consisting from 24 x Partition from the first shelf (1 x 24 Raid-DP) plus 24 x 1drive (12 from the second shelf, and the other half from the third, as a second RAID-DP group). It is leaving one drive for spare, but you got the idea.
In this scenario, If I lose one drive from the first shelf - this automatically means that I am going to affect 1 root aggregate + 1 data aggregate. Not to mention that in some twisted scenario if I loose the whole first shelf I will loose everything...
I am already playing with migration of the root aggregates in searching for the best option.
And I am not really sure how to disable ADP and if this is OK in my situation.