Efficiently Scale HCI with vSAN HCI Mesh

Hyperconverged Infrastructure (HCI) has solved many challenges with complex cloud infrastructure deployments for IT businesses over the years. Combining virtualised storage, compute, network, and management into a unified system, offers simplified management, streamlined operations, scalability, security, and out-of-the-box infrastructure automation, delivering rapid scalability to meet business-critical demands.

However, it can be challenging to right-size the hosts in your clusters to achieve the most efficient ratio of storage and compute. The very nature of HCI is that it scales linearly, meaning that additional hosts are required even when not all resources in the cluster are efficiently utilised. For large-scaled HCI environments, this often leads to additional costs for hardware, licensing, utilities and results in a significant amount of underutilised resources being wasted.

What is HCI Mesh and, how can it help?

HCI Mesh, introduced in vSAN 7.0 U1, is a software-based solution that provides disaggregation of storage and compute resources by enabling you to remotely share and consume storage between your vSAN clusters without needing additional hardware. For example, with HCI Mesh, vSAN clusters can consume spare storage from remote vSAN clusters within the same vCenter, delivering even more flexibility and scalability to your vSAN installations. HCI Mesh leverages native vSAN RDT protocols, preserving vSAN features and performance benefits.

Now, this is where things get cool! As of vSAN 7.0 U2, clusters without storage can be configured as an HCI Mesh Compute Cluster and can mount and consume storage from remote vSAN clusters, and best of all, HCI Mesh Compute Clusters do not require a vSAN license!

How to configure a HCI Mesh Compute Cluster

HCI Mesh configuration is very straightforward, provided you have met all the prerequisites and networking requirements. Check out the HCI Mesh FAQ. I will now show you how simple the steps are to configure HCI Mesh in my home lab.

In this demonstration, I’ll be mounting a remote vSAN datastore to a Compute-Only cluster by configuring it as an HCI Mesh Compute Cluster. The Datastore I’ll be sharing is from vSAN-Cluster-01.

Here you can see, there are no Datastores configured for hosts in Compute-Cluster.

Select the cluster, select Configure > vSAN > Service and press Configure vSAN, this will begin the vSAN configuration wizard.

Select the configuration type, “HCI Mesh compute cluster” and press Next.

Review the configuration and press Next.

On the Compute Cluster, the “Remote Datastores” option should now be available. Press “Mount Remote Datastore” to start the wizard.

The Mount Remote Datastore wizard will present available vSAN Datastores that can be mounted. Select the vSAN Datastore you would like to mount and press Next.

The wizard will perform a compatibility check against the remote Datastore and highlight any potential issues; if everything checks out ok, press Finish to complete the configuration.

And that’s it! You should now see the compute cluster has successfully mounted a remote vSAN Datastore and is available for consumption.

Before you continue, you may need to tune your High-Availability (HA) settings for clusters mounting remote Datastores. The failure response for the setting “Datastore with APD” should be set to “Power off and restart VMs” either aggressing or conservative. For most scenarios, conservative will be the preferred option.

Now, you may exhibit the following warning on your HCI Mesh compute cluster, “The number of vSphere HA heartbeat datastores for this host is 0, which is less than required: 2” Since the only storage attached to this cluster is a remote Datastore, I think this can be safely ignored and can be suppressed.

To suppress the Insufficient Heartbeat Datastore warning, you can apply the following advanced option to your cluster. Under Cluster HA Settings, Select the Advanced Options and add the following option “das.ignoreInsufficientHbDatastore” with value “true”. and press OK.

Note: Given how flexible and straightforward HCI Mesh is to configure, some of you may be tempted to build a beefy vSAN cluster and share it out with multiple compute clusters. Well, just because you can doesn’t mean you should! I highly recommend you familiarise yourself with HCI Mesh design and performance considerations before deploying this in anger to use with your tier-1 production applications.

Deploying a VM on a Remote Datastore

Now that we have configured HCI Mesh and mounted a remote vSAN Datastore to a cluster configured as an HCI Mesh Compute Cluster, let’s test deploying and running a VM on that Compute cluster using a Remote Datastore we mounted earlier.

For this, I’m going to clone an existing VM in my lab and select the Compute Cluster as the destination.

I then select the Remote vSAN Datastore we mounted earlier.

Review the deployment, Finish and, power on the VM.

And there you have it! A VM running on a Compute-Only cluster using a Remote vSAN Datastore. This truly opens up some amazing possibilities for the future of HCI.

Note: The environment for this demonstration was built in under 10 minutes with nested virtualisation leveraging the instant-clone technique.

I hope you found this helpful. Feel free to comment if you have any questions.

Leave a Reply