Question

Mixing HDDs in a node

  • 24 March 2021
  • 4 replies
  • 665 views

Badge

We have 3 nodes with 2x960GB SSD and 4x4TB HDD.

We need to expand storage (and storage only as it is going to be used to store cold data - we do not need additional processing power or RAM)

Question 1: We simply added two 12TB HDDs to each node. Now each node has 2x960GB SSD, 4x4TB HDDs, 2x12TB HDDs. Everything is working fine, but we were told that this is not recommended. What is the problem with this configuration

Question 2: Assuming the configuration in Q1 is not recommended and we want to end up with each node having one type of HDDs. We were told that we can upgrade the existing 4TB HDDs with 12TB drives. What the right procedure would be:

  a) Mark one 4TB drive for removal, remove it, insert 12TB drive in place of removed 4TB drive. Repeat 4 times for a node. Resulting configuration - 6x12TB HDDs per node

  b) Add two more 12TB drives first, then mark for removal and remove four 4TB drives (one by one). Resulting configuration - 4x12TB HDDs per node (preferred configuration by us)

Thank you!


4 replies

Badge

Added 12TB drives to the nodes on Saturday. They are just finishing populating with the data… The plan is to mark one 4TB drive for removal on Thursday night. Cannot risk doing it during the week. Will report back once it is successfully removed

Userlevel 7
Badge +34

How did things go @IgorBinder ?

Badge

This is great. I hoped for this to be the case.

Unfortunately Nutanix engineer insisted that we need to remove a drive first and then replace it with the new one, which would imply that after loosing a drive Nutanix expects it to be replaced (as a regular RAID) which would not make much sense.

Thank you for confirming.

Will report back after successfully following your instructions

Userlevel 2
Badge +2

Hello, 
from what I have read and what I use, Nutanix recommends the same disks across the entire cluster. The believe the reasoning behind this is how the replication works, but I am not 100% certain. For example: if you have 4x 12tb disks per node across a 3 node cluster, this is fine. So is having 6x 12tb disks per node across the entire cluster. The issues come when you mix and match drive sizes. (other clustering software recommends this as well as software raiding programs such as ZFS. in ZFS if you have a 8tb 8tb 8tb and 10tb drive, the software will partition the 10tb drive as an 8tb drive. mixing and matching drive sizes becomes problematic when calculating where to put the data and can lead to corruption of data or data loss depending on how the data is in the cluster). 

Tl;dr keep your drive sizes the same to avoid headaches :joy:


here’s how I add / remove drives in my homelab: 
1. add the 12tb drives.

  1. initialize the 12tb drive in the hardware tab.
  2. wait for the data to migrate over.
  3. when the drive seems to stablize, go to the hardware tab and remove a SINGLE 4tb drive.
  4. wait as the ​​​​​1 drive is being removed and data is being replicated. 
  5. once the drive is marked as completely remove, power off the node. → this is optional if you have Hot swap available, but i’d recommend it for homelab use. 
  6. remove the drive. 
  7. rinse and repeat per drive. 

Hope this helps,

College Student

Reply