#Disk assignment and aggregate failover

1 messages · Page 1 of 1 (latest)

fervent raft
#

Hi everyone,
I am new to the Netapp world so please help me out with this question.
I am using a Netapp A150 to setup a block storage for the esxi host.
There are 2 nodes and each node has 6 disks each of 894GB capacity. Initially, 2 aggregates were created with RAID-DP and the total capacity came out to be 6.5TB with 3.29TB for each aggregate(Each aggregate had 11 disks with one disk as spare and each disk had 415GB as the new capacity)

Now, I have to chuck that plan and assign all the 6 disks from one node to the other, so that I avoid doing Double parity for each aggr and then use node aggregate failover in case the node with 12 disks(after reassignment) goes down.

How do I go about it?
I started with storage disk assign and gave the ownership of all the disks in one node to the other.
Do, I need to format the disks? I am still unable to create one single aggregate using the node with 12 disks. What am I missing here?

past nexus
#

to clarify, you can't assign all physical disks to one node. the system won't work .

#

Here's a visual of how your disks are partitioned out -

image.png
#

you can create one single aggr on node 1 - but you don't gain anymore space.

muted roost
#

You need to assign the disk partitions, not the actual disks themselves. Each Disk will have 3 partitions, you need to ensure the P1 and P2 partitions are all assigned to a single node, these are the data partitions used to make your data aggregates.

The CLI will be your best avenue.

storage disk partition show - Will list the partitions
storage disk partition remove-owner - Remove the ownership from the partitions assigned to Node 2
storage disk partition assign - Assign the partitions to Node 1

You need to be in advanced privilege mode to see the "disk partition" command (set advanced)

I will typically use the partition show command to get all the partitions I want to unassign, then create myself a text file with all those partitions listed and the remove-owner commands followed by the assign commands.

I usually remove ownership and then assign immediately. So just do everything for one disk at a time (otherwise you might run into disk auto-assignment complications, and thats another story).

past nexus
#

OP - Here's the script -

set d
<yes>
storage disk option modify -node * -autoassign off
disk removeowner -disk x.x.x -data true
disk assign -disk x.x.x -data true -owner NODEx
disk show -fields data-owner, root-owner
storage disk option modify -node * -autoassign on
set admin

fervent raft
#

@past nexus @muted roost Firstly, thanks a lot for the suggestions. Also, is it possible to do the following? I might be sounding extremely dumb with this idea though, but I’ll give it a try. Out of the 12 disks, have 9 disks with root-data partitions in RAID-4 on node A as we actually do not need that high of a redundancy. And then, 1 as hot spare and the remaining 2 part of node b in RAID 0. Have Active/Passive setup. The node B will have just 2 disks to preserve root aggregate(root partition) of node B. What I want is one large aggregate and have a node aggregate failover in case node A fails. But the question that arises is, what happens once node A is back online and then node B is down. Can just the aggregate failover take place and the 2 disks of node B remain there? Is there an effective way to go about this?

unique shore
#

ONTAP will not let you put P1 and P2 in the same raid group. Whether all partitions are assigned to one node (1 aggregate with 2 raid groups with 4 parity partitions) or across both (2 aggregates with 1 raid group each and 2 parity partitions each) you end up with the save usable capacity.

If you do not leave one data partition from P1 and one from P2, ONTAP will have a fit.

past nexus
stuck maple
#

@fervent raft you should use ADPv2 with root/data/data partitions and RAID-DP for the optimal redundancy and resiliency. Then all disks are partitioned with data partitions for each node and all 12 disks are usable by all nodes.

muted roost
#

Correct, as @unique shore said the P1 and P2 partitions can't be in the same raid group. But you could make two raid groups in the same aggregate, which will end up with the same parity usage anyway (both raid groups will use RAID-DP). Its really up to you, you'll have the same amount of usable capacity as having one aggregate per controller, it'll just be one contiguous pool. Some customers like this on small systems as it removes the choice of which aggregate to place a volume on.

But also remember that by doing this, Node 2 will be idle almost all of the time. It's compute and network resources will only be used doing a storage failover event, so its unused resources you've paid for.

stuck maple
#

When using ADPv2... your layout will look like this:

unique shore
#

Wow @stuck maple that’s old! ONTAP hasn’t shipped like that for quite some time! Same picture imagine six in the left and six on the right (instead of all twelve to the left)

stuck maple
#

@unique shore .... I was using that as an example of the partitioning and not necessarily using it for drive placement! 😋

fervent raft
#

In case, you want to know why I need this, below is the requirement. So, a volume from a source netapp is going to be snapmirrored to my netapp. The volume can be greater than 4TB. My netapp has 12 disks of max capacity of 896GB. Now, after the root partition is given to each disks, the usable space is 832GB out of which 416GB is taken by each data partitions p1 and p2. Now, after all this and then parity, double parity and hot spare, I am left with 6.58GB of capacity. Now, if I have 2 aggregates of 3.29TB each, volume with more than 3.29TB cannot be snapmirrored. Hence, I plan on using all the disks in node 1. But, give a disk maybe to node 2 for the Ontap OS to run in the root aggr. So, I thought of giving all the p2 and p3(its root-data1-data2 partitioned) to node 1 but the root partitions of 6 disks will be with node 2 and the other 6 with node 1. Is that a better approach? Or is it better to do that in RAID-4 setup. Btw, this doesn’t require high redundancy. Just 1 parity with one spare should be fine. I’m just trying to learn by doing different things, but I don’t know if I am breaking something by doing this. Please help!

past nexus
#

I don't even think you can do RAID 4 root aggrs anymore. but looking at the math...
You'd lose 5 disks up front - leaving 7 for data - gives you a 5 TiB Aggr. if you skip having a spare you'd have 5.9 TiB aggr.

ADPed with all partitions assigned to node 1 you have a single aggr at 6.5TiB.

I wouldn't over complicate this and leave it as ADP and have all partitioned assigned to node one.

fervent raft
#

@past nexus This is how it looks like. Also, is it a bad thing to use root-data partition scheme instead of root-data1-data2? Also, why do I have 22 entries instead of 11? Is that because P1 and P2 can’t be in the same RAID group? Does that look normal. And, apologies for asking too many questions.

IMG_4514.png IMG_4513.png
past nexus
#

You have 22 data partitions. 11x2 + HS.

#

also v1 was for FAS systems. AFF uses v2

#

but with 12x960 drives. v1 gives you the same useable as v2

#

(with all drives on one node)

fervent raft
#

@past nexus Understood. Do I go ahead and use that setup? I want an active/passive setup. I am not sure if this kind of setup is being used elsewhere.

past nexus
#

It's not the default, but people do use that type of config for various reasons.

fervent raft
#

@past nexus Also, is there any advantage when it comes to the performance or efficiency when using the ADPv2(r-d2 partition scheme) as opposed to Root-data? As you mentioned, I see the capacity is the same when I use either of those, but does one perform better than the other?