vSAN stretched clusters have had a long history of success with our customers. The distributed architecture of vSAN allows for stretched clusters to simply be a variation of a single-site vSAN cluster, but one that provides full site-level resilience of your VMs and data across two data sites.
With vSAN for VMware Cloud Foundation (VCF) 9.0, we’ve enhanced stretched clusters in several ways. One significant improvement is the ability to have a vSphere cluster spanning two sites “site aware” so that it can properly consume the storage of a stretched vSAN storage cluster. Let’s look at this in more detail.
What’s New
With the introduction of vSAN storage clusters (previously known as vSAN Max) in vSAN 8 U2, administrators could build a vSAN cluster and treat it like a storage array. You could enjoy all of the benefits of storage provided by vSAN while the vSphere clusters consuming the storage could be scaled and managed independently.
Much like aggregated vSAN HCI clusters, vSAN storage clusters could run in a stretched cluster topology due to its inherent understanding of geographic sites using “fault domains.” But there was a gap. vSphere clusters mounting the datastore had no understanding of fault domains. This resulted in a topology that would not handle a stretched cluster topology properly.
vSAN for VCF 9.0 removes these limitations. Now you can have disaggregated compute and storage resources stretched across two data sites with the same behavior as a stretched vSAN HCI cluster. The enhancements include the following:
- Uses optimal data path from VM instance to backing storage. In a fully operational state, a stretched vSphere cluster mounting the datastore of a stretched vSAN storage cluster has two paths to the data. Without any site awareness within the vSphere cluster, this impacted the ability to provide consistent performance and maintain efficient use of the ISL. With the use of fault domains, now the VMs will have an understanding of the optimal data path to transmit its storage I/O.
- Proper site failover of VM instances in the event of a site failure. Previously, if a vSAN storage cluster was configured as a stretched cluster, the data remained resilient and available during a site failure, but the VM instances did not. The introduction of fault domains for vSphere clusters mounting a vSAN datastore will allow for HA to restart VMs in the remaining site upon a site failure.
Figure 1. Stretched vSAN storage cluster providing site resilient storage to stretched vSphere clusters.
Note that the witness site for the vSAN storage cluster is not shown in the illustration above. This is simply for clarity purposes. A stretched vSAN storage cluster will need a third site for a witness. The stretched vSphere clusters that mount the datastore do not need any type of witness site or appliance.
Why it Matters
Why is this so important? The ability to stretch a vSphere cluster for use with a vSAN storage cluster addresses an often unspoken truth about VMware Metro Storage Clusters (vMSC). Metro storage clusters, which are backed by traditional storage arrays, have a reputation of being difficult to deploy, operate, and troubleshoot. Configurations are complex, and often specific to a given storage vendor, with strict requirements for the storage arrays at each site.
vSAN stretched clusters offer site level resilience that is easy and flexible. With stretched clusters extended to a disaggregated topology using vSAN storage clusters, you can have all of the benefits of metro storage clusters without the headache and expense.
Simple Configuration
The steps to configure a stretched vSphere cluster to mount the datastore of a stretched vSAN storage cluster are easy and intuitive. In this example we have a vSAN storage cluster that is stretched across two data sites: Milan and Turin. We have a very small 2-host vSphere cluster spanning both sites that will be mounting this datastore, but we must configure it so that it has the same level of site awareness as the vSAN storage cluster. The basic steps will include:
- Configuring the vSphere cluster as a stretched vSAN compute cluster.
- Mount the remote datastore.
Configure vSphere Cluster
The first step will prepare the stretched vSphere cluster to mount a vSAN datastore with site awareness. Highlighting the vSphere cluster, and selecting Configure > vSAN > Services will allow you to choose “vSAN Compute Cluster” which is simply a name for vSphere clusters that are configured to mount a vSAN datastore.
At this point, one can choose a single-site compute cluster, or a stretched compute cluster. For this example we will choose the latter, as shown below.
Figure 2. Choosing a stretched ‘vSAN Compute Cluster’ from the UI.
Reminder: A “vSAN Compute Cluster” is simply an in-product reference to a vSphere cluster that is mounting the datastore of a remote vSAN cluster. The hosts in this vSphere cluster have a thin layer of vSAN activated allowing them to communicate with the datastore in an intelligent way. The servers in a vSAN compute cluster only need to comply with the compatibility guide for vSphere, not vSAN. They may be referenced as vSAN compute clusters, or vSphere clusters interchangeably.
Next, the workflow will allow us to define the names of the fault domains representing each site of the vSAN compute cluster, and which hosts reside in their respective sites. Selecting descriptive site names that match your stretched vSAN storage cluster will help you self-document your topology, and assist in future configurations and troubleshooting. Once the fault domains are named, the hosts physically located in each site can be associated with the respective fault domain.
Figure 3. Selecting the site names and host associations for each site.*
* Hosts shown in Figure 3 do not reflect the minimum host count required for VCF.
Mount Remote Datastore
Once the configuration of the compute cluster is complete, it will give the option to mount a remote datastore. The first step will be to simply choose the datastore residing on the stretched vSAN storage cluster.
Figure 4. Select the datastore of the desired stretched vSAN storage cluster.
The next step will be to configure “site coupling.” This associates the fault domains of the stretched vSAN storage cluster to the fault domains of the stretched vSAN compute cluster. In the image below, we can see how proper naming can be helpful in this step. In this example, once the server site of “Milan” is coupled with the client site of “Milan” the other site coupling occurs automatically.
Figure 5. Configure site pairing of the stretched vSAN storage cluster to the stretched vSphere cluster.
At this point, the workflow will complete after a few compatibility checks ensure proper configuration.
Figure 6. Check compatibility when mounting the remote datastore.
Once the datastore is mounted, it will be available for use with the VM instances created on the compute cluster. Do not forget to create the appropriate host groups, VM groups, and VM/host rules in DRS to ensure the proper distribution of workloads using DRS. For more information, see the “Cluster Settings – DRS” section of the “vSAN Stretched Cluster Guide.”
Figure 7. The mounted datastore of a stretched vSAN storage cluster to a stretched vSAN compute cluster.
Recommendation: Ensure the Auto-Policy Management feature is enabled on the vSAN storage cluster. This will create a cluster-specific default storage policy based on the topology and host count of the cluster. It takes the guess-work out of determining the ideal storage policy for a cluster.
For more information on vSAN stretched clusters, see the “vSAN Stretched Cluster Guide.” See the “Design and Operational Guidance for vSAN Storage Clusters” for more recommendations on the use of vSAN storage clusters.
Summary
With the very latest updates to vSAN in VCF 9.0, you can now provide a fully site aware stretched cluster arrangement using vSphere clusters and vSAN storage clusters. Since vSAN storage capacity is a part of your VCF licensing, it is a great way for you to eliminate some of the complexities and hardware dependencies of VMware Metro Storage Clusters.
***
Ready to get hands-on with VMware Cloud Foundation 9.0? Dive into the newest features in a live environment with Hands-on Labs that cover platform fundamentals, automation workflows, operational best practices, and the latest vSphere functionality for VCF 9.0.
