Need Official Guidance: Two-Node Cluster + Witness Deployment Requirements, Supported Hardware, and Licensing

Hi Atsuyori,

1. Yes a dedicated host is required. The witness for a two node cluster is different then a witness for 2 complete clusters for disaster recovery. the witness for a two node clusters can be in the same datacenter. 
2. There is no "standalone AHV’, this is a misconception lots of people have. AHV is only available on nutanix hci, so a CVM (a nutanix cluster) is always required for AHV. You can run the witness for a two node clusters on every type of hypervisor where the download is available for. (ahv and esx)
3. Again, there is no standalone AHV.
4. It is AHV and/or ESX. For AHV this means any nutanix cluster you already have running. For ESX this is any vmware/broadcom environment you already running. 
5. Normally a two node cluster is only deployed if there is a normal nutanix cluster running in the main datacenter, and then the witness is placed there. 2 node clusters are typically not being used as the main cluster. 
6. Here you can find all information for the witness vm: https://portal.nutanix.com/page/documents/list?type=software&filterKey=software&filterVal=Witness%20Service%20(formerly%20Witness%20VM)

So to make this easier, get yourself a three (3) node cluster which is the minimum for proper Nutanix cluster without all the hussle a 2 node cluster brings you. 

Dear JeroenTielen

Thank you for your detailed explanation.

We have already completed our evaluation of scalability, HA, DR, operations, and ROI, and have decided to proceed with a Two-Node Cluster with Witness.

The issue is that both the Nutanix sales and pre-sales teams in Japan and Singapore refuse to discuss a supported Two-Node + Witness design and continue recommending a three-node cluster instead.

I am not looking for alternative architectures. I am simply trying to identify an officially supported configuration so that we can proceed with procurement and support planning.

Could you please clarify the following based on your experience?

Is hosting the Witness VM on an independent Single-Node Nutanix Cluster a supported configuration?

Have you personally deployed or supported this architecture?

If so, what hardware and licensing were required for the Witness hosting platform?

Is the requirement for a dedicated host an official Nutanix requirement or a best practice recommendation?

Any information based on real deployment experience or Nutanix Support validation would be greatly appreciated.

Thank you.

Additionally, I would like to explain the architecture I am trying to validate.

My understanding is based on the Nutanix documentation:

• "Nutanix also offers the option of two-node cluster. By configuring external witness VM in a separate failure domain, the configuration provides resiliency features of a three-node cluster." [Web-Consol...Prism-v7_6 | PDF]
• "Witness VM may reside on any supported hypervisor, and it can run on either Nutanixor non-Nutanix hardware." [Web-Consol...Prism-v7_6 | PDF]
• "The Witness VM can be deployed either off premise or in a different physical platfom on premise." [Web-Consol...Prism-v7_6 | PDF]
• Single-node clusters are described as a supported deployment model in the current Web-Console-Guide-Prism-v7_6. [Web-Consol...Prism-v7_6 | PDF]

Based on the documentation above, I am trying to determine whether the following archtecture would be supported:

Production Environment

Two-Node Cluster

- Node A

- Node B

Separate Physical Platform

Independent Single-Node Nutanix Cluster

- Witness VM

To be clear, I am not looking for guidance on whether I should deploy a three-node cluster.

Our evaluations regarding scalability, HA, DR, operational requirements, and ROI have already been completed, and we have decided to proceed with a Two-Node Cluster with Witness.

The challenge I am facing is that both the Nutanix sales and pre-sales teams in Japan and Singapore have been unwilling to discuss a supported Two-Node + Witness architecture and continue redirecting the discussion to a three-node design.

I am simply trying to identify an officially supported architecture so that I can proceed with system design, procurement, licensing, and support planning.

If anyone has deployed a production Two-Node Cluster with Witness, I would greatly appreciate learning:

• Where the Witness VM was hosted
• Whether AHV was used
• Whether the architecture was accepted by Nutanix Support
• Any licensing or infrastructure requirements that were identified during the deployment

Any real-world deployment experience, supported reference architecture, or feedback from Nutanix employees would be greatly appreciated.

Yes I have built and design this. But then the two node cluster was in an edge location and the witness was hosted in the main datacenter. If you are going for a single node nutanix cluster just to host the witness that doesnt make any sense. Then you have bought 3 nodes already and I would always suggest to do a three node cluster. 

If you are really going forward with the two node cluster. The witness doesnt need to be on a different location. Can be hosted in the same dc as well. This is different (offcourse) if you are using witness for (for example) disaster recovery. 

But, as sys-admin to sys-admin, safe yourself the hussle with a two node cluster and get yourself a proper 3 node cluster. 

Dear Jeroen

Thank you for your reply.

I haven’t purchased anything yet.
I’m currently verifying the details before making a purchase.
While all three nodes must have identical specifications when configuring a three-node setup, I believe that at least one node should be able to run on the minimum specifications required to host the Witness.
As mentioned earlier, I understand why you’d recommend a three-node setup, but we’ve thoroughly reviewed the official guide and decided to set up a two-node plus Witness configuration instead, so we no longer need advice on that matter.
I simply want to confirm the officially supported configurations for our upcoming procurement.