#I build in live and virtual (fusion) I get this behavior in a AFF-C400/with 60x 15.3TB SSD's.

1 messages · Page 1 of 1 (latest)

plush frost
#

When I have only configured in Fusion 48 disk (with 2 shelves) I have 2 root spares and 4 data spares. (ADPv2)
When I add a new 3th shelves the data spares are gone. Assuming that the other 12 disks are spares.
But when I configure (mixed) a aggr with 6 or 12 disk (claiming all disks) I dont get the data spares back in the ADP setup.

Is this normal behavior in Fusion? Now I have only root spares.

In real live there are two extra raidgroup of 2x 5 disk and 2 spares (1 per controller) build on the 3th shelf.
The ADP config is as configured in a full 2x 48 data partioned raidgroup (2x 24)

Final, when I auto create a aggr on the 3th shelf it configures 1 spare (still no spares on data partition)
1 spare is configured for 1 controlle assumable (which is not a good thing)

What are the spare rules in this scenario(s)?
The point here is that with 15TB SSD you loose available capacity easily.

unreal talon
#

Maybe because ADP only works up to 48 drives?

plush frost
#

I know but its about the spares

lyric halo
#

@unreal talon system initialization only works on up to 48 drives. You can use it in as many as you’d like.

#

There are plenty of issues with fusion that in my opinion, just like you are finding @plush frost that it doesn’t quite behave like a real physical system does

plush frost
#

Thnx i noticed, but the behavoir is also (fysical) when you have a 48 drives (actually it's more 48 partitions (but ok)) are random over the 60 drive (but its still aware of the shelf separation). When a 49 or more drives come in to play the 48th spare partitions for data will be used for data. And the spare (role) is moved to the 49th or more disks. Is that behavior by design? The two extra data partition are then added to the usable capacity. So you must be aware of the spare configuration/redundancy. The position of a drive in a shelf with ADP does not matter so it seems.

lyric halo
#

I just did a fas8700 with 60 drives. That is currently setup with 2 aggregates. Each aggregates has a max raid group size of 28 and 28 disks each, two spares each

You cannot use the automatic way. You need to manually build the aggregate. There should be 12 drives assigned to each node to start. There will be two whole spares left at the end.

aggr create -node local -maxraidsize 28 -diskcount 28 -disktype FSAS -aggr myaggr_01 -encrypt-with-aggr-key true

#

I will have plenty of UNUSED root partitions. But what I’m giving up is way less than the parity tax.

#

Anything more than filling up the raid group will use whole drives. When I go to add a shelf to this system in two months, I am simply going to add 28 drives to each raid group. The second raid group will contain all whole drives and no partitions

#

There is no reason to partition the next set. I need the extra parity drives for another raid group.

wooden shadow
#

I have 192 disks on 2 C400 HA pairs, all using ADP:

partition show

...576 entries were displayed.

disk show
192 entries were displayed.

#

so its 48 drives per node?

#

I did manually copy partition layout from root-data1-data2 drives to any not partitioned

unreal talon
#

adp only supports 48 drives per ha-pair for root slices

dusty trout
#

so unless you want to have 196 spare root slices lying around I don't see a point 🙂