Solved

Nutanix PoC – Community Edition - Hardware Disk setup

  • 25 September 2023
  • 6 replies
  • 1810 views
J
Badge

I am planning to make Nutanix PoC with the Community Edition. And if successful use Nutanix. The goal is to make a tree or four node cluster with local storage.

And how would it be best to prepare the tree nodes. The have buildin Hardware Raid controller. (Cisco UCS servers). I could put in up to 8 disks. They can be in same size or in deferent sizes. I would probably be some 1-2 TB SSD disks. I can make one or more Raid disks. (Raid 1, 5, 10). Would would make sense in a performance and redundancy perspective.

I have read the CE Getting started guide. Here there are mentioned:

  • Storage Devices (Max 4 HDD/SSD)
  • Cold Tier (500 GB or greater, maximum 18 TB (3x6 TB HDDs)
  • Storage Devices, Hot Tier Flash (Single 200 GB SSD or Greater.
  • Hypervisor Boot Device (32 GB pr. Node)

Witch storage needs to be fast? And should I have my hardware raid make the disks ?.

Would it make sense to build a raid 10 on eight equal sized disks and make logical drives fitting each of the above disks types?

icon

Best answer by ronaldvanvugt 25 September 2023, 22:20

View original
This topic has been closed for comments

6 replies

ronaldvanvugt
Userlevel 1
Badge +2

Hi Jan,

You don't need a RAID configuration, but a JBOD (Just a Bunch of Disks) configuration. The control VM/AOS of Nutanix is responsibilty of the disk redundancy and the control VM needs direct access to the disks/controller. Just put 1 SSD disk and 3 HDD disks and 1 boot disk in every node and configure the disk controller as JBOD.

Best regards,

Ronald

Ronald van Vugt
J
Badge

Tanks for the reply. Are there some dokuments about how to spec the disk hardware for the nutanix hosts in a 3 or 4 node cluster. And the network adapter as well by the way. I was thinking of two 10G ports on a dual port NIC. The memory/CPU probably is just more and faster are better :-)

The hardware are probably going to be either “Cisco UCS C220 M4” or “HP DL360-G9” with vendor supported hardware.

MartijnKonings
Badge +1

Hi Jan,

 

To be able to install the 3/4 nodes you need at least 3 disks in non raid (jbod) setup, like ronald mentioned. 1 for the cvm on fast ssd (the acropolis os software VM) which is the heart of the hyper converged infrastructure (hci) and controls everything. One data disk and one boot disk (this can even be an usb stick). These are the minimum requirements to be able to install the software on each node. After installation you can join these 3 nodes and configure them as a cluster. You will need at least 3 nodes for a rf-2 cluster 

 

documentation on CE installation and requirements can be found here : https://portal.nutanix.com/page/documents/details?targetId=Nutanix-Community-Edition-Getting-Started-v2_0:Nutanix-Community-Edition-Getting-Started-v2_0


regards,

martijn

 

 

 

Nutanix newbie
J
Badge

Thanks Martin. Check to the boot disk. I had figured that one out. It would probably be some high class SDCARD on the motherboard on the server.

 

The disk you mention for the cvm (acropolis os software) is this the “Hot Tier Flash” or an additional disk?

 

I had actually read the getting started ant this was my reason to ask for the different disks purposes. What I can not figure out is how the cluster handles a disk failure on one of the 3 or 4 nodes in the cluster:

  • Boot disk, well yes the node are not starting and is of course dead.
  • Cold Tear: Are the node dead or can it operate in a degraded way with data from the other nodes.
  • Hot Tear: Are the node dead or can it operate in a degraded way with data from the other nodes.

And would it make a node more robust on operation if more disks for the different types of disks uset in operation (Hot and Cold tear I guess).

 

I know a lot of questions. It might all be answered in some documentation I have not seen. If there are a “tech” dokument describing this just point me in the direction.

ronaldvanvugt
Userlevel 1
Badge +2

Hi Jan

Every Block will be written 2 or 3 times to different nodes (you can configure this, but for 3 times you need at least 5 nodes). 
 

of one disk fails, the CVM will get the missing block from a disk of another node. After a while AOS will also start the self healing proces and will write the missing blocks again to others disks. 
 

You can read more about the technique of Nutanix in the Nutanix bible, https://nutanixbible.com
 

best regards 

Ronald 

Ronald van Vugt
MartijnKonings
Badge +1

Hi Jan,


Like Ronald mentions. The storage in a cluster is one big distributed storage pool system, which spreads it's data across all your nodes. So data gets written to the all the 3 nodes in your cluster. This is all done by the Controller VM (CVM). You can read about this in the Nutanix Bible: 

 

 

Above picture shows the architectrure. (source: nutanixbible.com) You need to have your data disks (all flash or mixed) redudant so at least two of each. But the CVM all handles the writing. The boot can be 1 disk (usb flash) since it's only needed to boot the hypervisor (AHV for example). The CVM runs the AOS software. This is the VM which runs on top of the AHV hypervisor and controls everything.

 

The 3 disk example I gave earlier offcourse is not realistic since your node will die instantly after one disk fails and it has a single point of failure there. But for a lab if you want to test out the software in a single node cluster for example that is minimum to get it installed and running.

In a 3 node cluster you can lose 1 node and everything keeps running. After the node dies you will lose the redundancy so need to recover ASAP and after recovery the CVM rebuilds the data itself and you are redundant again. You can see the status for this on the prism overview page:

 

Prism is the administrator interface of a Nutanix cluster.

Regards,
Martijn

Nutanix newbie
Terms & Conditions