Build separate cluster for UCS C240 and HX240C

[zed_theNewBie]

I am newcomer to this forum so please bear with me.
At work, we are planning to deploy a Hyperflex cluster.

Cluster-01
5 x UCS C240 Compute-Only node

Cluster-02
10 x HX240C - Converged node

Based on the guides I've read so far, I can easily deploy an HX cluster with all servers using HX240C.
But I can't find a document that would tell me if I can deploy an HX cluster with all servers using UCS240 Compute-only node.

It is a must that the UCS C240 servers have to sit on their own cluster separate from the HX240c servers.

My question, is it possible to deploy the above clusters using HX Data installer?
Any suggestion, on how to build and deploy the 2 clusters?
Any links or documents that I can read  and follow through to achieve the above cluster setup?

Thank you.

[RedNectar · Accepted solution]

Hi @zed_theNewBie ,

Firstly let's make sure everybody is equally confused about the word cluster when it comes to HyperFlex configurations.

1. The UCS cluster - is nothing more than a pair of fabric interconnects
2. The HyperFlex cluster is a group of n Converged (storage) nodes (where n>=3) along with a group of m Compute Only nodes (where 0<=m<=n using a standard Advantage licence, 0<=m<=n*2 using a Premier licence).
   So 10 Converged nodes will work fine with 5 compute only nodes - but only if they are all in the same HyperFlex cluster.
3. Oh, and the HyperFlex Cluster is sometimes referred to as the HyperFlex Storage Cluster or the HyperFlex Management Cluster - essentially they are all the same thing, although both the  HyperFlex Storage Cluster IP address and the HyperFlex Management Cluster IP address are on different VLANs/Subnets
4. Assuming you will manage these nodes in vCenter, remember that you will create a cluster of ESXi hosts in vCenter, which will manage the collection of Converged and Compute Only nodes.

Now back to your question:

At work, we are planning to deploy a Hyperflex cluster.

Cluster-01
5 x UCS C240 Compute-Only node

Cluster-02
10 x HX240C - Converged node

Are you trying to configure:
A. ONE HyperFlex cluster and TWO UCS clusters? Or

B. TWO HyperFlex clusters and TWO UCS clusters? Or

C. Are you really just trying to put all the servers into ONE UCS Cluster with ONE HyperFlex Cluster being managed by ONE vCenter cluster?

If you want to put them all into one cluster (answer C), then this is the process:

1. Make sure you have done your planning and know about all the VLANs and IP addresses you'll need. This is one very good planning resource: https://hxpreinstall.cloudapps.cisco.com/#/deployments - and a cheeky search for rednectar tag hyperflex may yield more interesting results!
2. Download the installer VM and deploy it to an EXISTING vCenter (or ESXI host or even your own PC running VMware Fusion/Desktop)
3. Make sure you've configured your Fabric Interconnects
   1. with management IPs (and a cluster IP) before you start, and those IPs are visible to the Installer VM.
   2. via UCS Manager, config the FIs with uplink ports (and made them port-channels if appropriate)
   3. BE CAREFUL assigning server ports on the FI - MAKE sure you enable the port where server1 is connected first - wait until it is discovered, then enable the port for server2 and so on for all 15 servers.  Failure to do this will result in the servers being assigned random numbers in UCS Manager
   4. Make sure you have configured all VLANs you need on the upstream switches from the Fabric Interconnects
4. Run the installer by browsing to the Installer VM IP and logging in as root - you'll be asked to change the password from the default cisco123
   • Now the strategy is to install ALL the converged nodes first - ignore the compute only nodes. When that has completed, run the installer again and do the expansion.
5. Click Create Cluster, and install the HX Cluster on the 10 Converged nodes (give yourself at least a day)
6. When the cluster is build and healthy, come back to the installer and this time click Expand Cluster

I would HIGHLY recommend practising on Cisco dCloud first, there is a good lab there that takes you through both an install and an expansion

Now back to your specific questions

Based on the guides I've read so far, I can easily deploy an HX cluster with all servers using HX240C.
But I can't find a document that would tell me if I can deploy an HX cluster with all servers using UCS240 Compute-only node.

You can't deploy a HyperFlex Cluster with just compute only nodes - the VMs have to live SOMEWHERE

It is a must that the UCS C240 servers have to sit on their own cluster separate from the HX240c servers.

• The UCS C240 servers:
  • MUST be in the same HyperFlex Cluster as the HX204C 
  • MAY be on the same UCS Cluster (PREFERRED)
  • MAY be in separate UCS clusters, but that will require a lot more work

My question, is it possible to deploy the above clusters using HX Data installer?

See above

Any suggestion, on how to build and deploy the 2 clusters?

See above

Any links or documents that I can read  and follow through to achieve the above cluster setup?

dCLoud is your friend.  If you do not have rights to access dCloud or book the lab I linked above, PM me and I'll book one for you then share it.

I hope this helps.

---

Don't forget to mark answers as correct if it solves your problem. This helps others find the correct answer if they search for the same problem.

---

[zed_theNewBie]

Thanks RedNectar  aka Chris Welsh for your very thorough explanation and you even go beyond that to assist a newbie like me.

It definitely answers all the questions I have in mind.

I think I have now a very clear picture of what we're going to implement.