Windows Server 2025 Hyper-V Workgroup Cluster with Certificate-Based Authentication

In this guide, we will walk through creating a 2-node or 4-node Hyper-V failover cluster where the nodes are not domain-joined, using mutual certificate-based authentication instead of NTLM or shared local accounts. Here we are going to leverage X.509 certificates for node-to-node authentication. If you don’t use certificates, you can do this with NTLM, but we’re avoiding that as NTLM is supported, but the general recommendation is that you deprecate it where you can. We can’t use Kerberos because our nodes won’t be domain joined. 

It’s a lot easier to do Windows Server Clusters if everything is domain joined, but that’s not what we’re doing here because there are scenarios where people want each cluster node to be a standalone (probably why you are reading this article).

Before diving into configuration, ensure the following prerequisites and baseline setup:

• Server OS and Roles: All cluster nodes must be running Windows Server 2025 (same edition and patch level). Install the latest updates and drivers on each node. Each node should have the Hyper-V role and Failover Clustering feature available (we will install these via PowerShell shortly).
• Workgroup configuration: Nodes must be in a workgroup. The nodes should be in the same workgroup name. All nodes should share a common DNS suffix so that they can resolve each other’s FQDNs. For example, if your chosen suffix is mylocal.net, ensure each server’s FQDN is NodeName.mylocal.net.
• Name Resolution: Provide a way for nodes to resolve each other’s names (and the cluster name). If you have no internal DNS server, use the hosts file on each node to map hostnames to IPs. At minimum, add entries for each node’s name (short and FQDN) and the planned cluster name (e.g. Cluster1 and Cluster1.mylocal.net) pointing to the cluster’s management IP address.
• Network configuration: Ensure a reliable, low-latency network links all nodes. Ideally use at least two networks or VLANs: one for management/cluster communication and one dedicated for Live Migration traffic. This improves performance and security (live migration traffic can be isolated). If using a single network, ensure it is a trusted, private network since live migration data is not encrypted by default. Assign static IPs (or DHCP reservations) on the management network for each node and decide on an unused static IP for the cluster itself. Verify that necessary firewall rules for clustering are enabled on each node (Windows will add these when the Failover Clustering feature is installed, but if your network is classified Public, you may need to enable them or set the network location to Private).
• Time synchronization: Consistent time is important for certificate trust. Configure NTP on each server (e.g. pointing to a reliable internet time source or a local NTP server) so that system clocks are in sync.
• Shared storage: Prepare the shared storage that all nodes will use for Hyper-V. This can be an iSCSI target or an SMB 3.0 share accessible to all nodes. For iSCSI or SAN storage, connect each node to the iSCSI target (e.g. using the Microsoft iSCSI Initiator) and present the same LUN(s) to all nodes. Do not bring the disks online or format them on individual servers – leave them raw for the cluster to manage. For an SMB 3 file share, ensure the share is configured for continuous availability. Note: A file share witness for quorum is not supported in a workgroup cluster, so plan to use a disk witness or cloud witness instead.
• Administrative access: You will need Administrator access to each server. While we will avoid using identical local user accounts for cluster authentication, you should still have a way to log into each node (e.g. the built-in local Administrator account on each machine). If using Remote Desktop or PowerShell Remoting for setup, ensure you can authenticate to each server (we will configure certificate-based WinRM for secure remote PowerShell). The cluster creation process can be done by running commands locally on each node to avoid passing NTLM credentials.

The core of our setup is the use of mutual certificate-based authentication between cluster nodes. Each node will need an X.509 certificate that the others trust. We will outline how to use an internal Active Directory Certificate Services (AD CS) enterprise CA to issue these certificates, and mention alternatives for test environments. We are using AD CS even though the nodes aren’t domain joined. Just because the nodes aren’t members of the domain doesn’t mean you can’t use an Enterprise CA to issue certificates, you just have to ensure the nodes are configured to trust the CA’s certs manually.

Certificate Requirements and Template Configuration

For clustering (and related features like Hyper-V live migration) to authenticate using certificates, the certificates must meet specific requirements:

• Key Usage: The certificate should support digital signature and key encipherment (these are typically enabled by default for SSL certificates).
• Enhanced Key Usage (EKU): It must include both Client Authentication and Server Authentication EKUs. Having both allows the certificate to be presented by a node as a client (when initiating a connection to another node) and as a server (when accepting a connection). For example, in the certificate’s properties you should see Client Authentication (1.3.6.1.5.5.7.3.2) and Server Authentication (1.3.6.1.5.5.7.3.1) listed under “Enhanced Key Usage”. 
• Subject Name and SAN: The certificate’s subject or Subject Alternative Name should include the node’s DNS name. It is recommended that the Subject Common Name (CN) be set to the server’s fully qualified DNS name (e.g. Node1.mylocal.net). Also include the short hostname (e.g. Node1) in the Subject Alternative Name (SAN) extension (DNS entries). If you have already chosen a cluster name (e.g. Cluster1), include the cluster’s DNS name in the SAN as well. This ensures that any node’s certificate can be used to authenticate connections addressed to the cluster’s name or the node’s name. (Including the cluster name in all node certificates is optional but can facilitate management access via the cluster name over HTTPS, since whichever node responds will present a certificate that matches the cluster name in SAN.)
• Trust: All cluster nodes must trust the issuer of the certificates. If using an internal enterprise CA, this means each node should have the CA’s root certificate in its Trusted Root Certification Authorities store. If you are using a standalone or third-party CA, similarly ensure the root (and any intermediate CA) is imported into each node’s Trusted Root store.

Next, on your enterprise CA, create a certificate template for the cluster node certificates (or use an appropriate existing template):

1. Template basis: A good starting point is the built-in “Computer” or “Web Server” template. Duplicate the template so you can modify settings without affecting defaults.
2. General Settings: Give the new template a descriptive name (e.g. “Workgroup Cluster Node”). Set the validity period (e.g. 1 or 2 years – plan a manageable renewal schedule since these certs will need renewal in the future).
3. Compatibility: Ensure it’s set for at least Windows Server 2016 or higher for both Certification Authority and Certificate Recipient to support modern cryptography.
4. Subject Name: Since our servers are not domain-joined (and thus cannot auto-enroll with their AD computer name), configure the template to allow subject name supply in the request. In the template’s Subject Name tab, choose “Supply in request” (this allows us to specify the SAN and CN when we request the cert on each node). Alternatively, use the SAN field in the request – modern certificate requests will typically put the FQDN in the SAN.
5. Extensions: In the Extensions tab, edit Key Usage to ensure it includes Digital Signature and Key Encipherment (these should already be selected by default for Computer templates). Then edit Extended Key Usage and make sure Client Authentication and Server Authentication are present. If using a duplicated Web Server template, add Client Authentication EKU; if using Computer template, both EKUs should already be there. Also enable private key export if your policy requires (though generally private keys should not be exported; here each node will have its own cert so export is not necessary except for backup purposes).
6. Security: Allow the account that will be requesting the certificate to enroll. Since the nodes are not in AD, you might generate the CSR on each node and then submit it via an admin account. One approach is to use a domain-joined management PC or the CA server itself to submit the CSR, so ensure domain users (or a specific user) have Enroll permission on the template.
7. Publish the template: On the CA, publish the new template so it is available for issuing.

Obtaining Certificates from the Enterprise CA

Now for each cluster node, request a certificate from the CA using the new template. To do this, on each node, create an INF file describing the certificate request. For example, Node1.inf might specify the Subject as CN=Node1.mylocal.net and include SANs for Node1.mylocal.net, Node1, Cluster1.mylocal.net, Cluster1. Also specify in the INF that you want Client and Server Auth EKUs (or since the template has them by default, it might not be needed to list them explicitly). Then run:

certreq -new Node1.inf Node1.req

This generates a CSR file (Node1.req). Transfer this request to a machine where you can reach the CA (or use the CA web enrollment). Submit the request to your CA, specifying the custom template. For example:

certreq -submit -attrib "CertificateTemplate:Workgroup Cluster Node" Node1.req Node1.cer

(Or use the Certification Authority MMC to approve the pending request.) This yields Node1.cer. Finally, import the issued certificate on Node1:

certreq -accept Node1.cer

This will automatically place the certificate in the Local Machine Personal store with the private key.

• Using Certificates MMC (if the CA web portal is available): On each node, open Certificates (Local Computer) MMC and under Personal > Certificates, initiate New Certificate Request. Use the Active Directory Enrollment Policy if the node can reach the CA’s web enrollment (even if not domain-joined, you can often authenticate with a domain user account for enrollment). Select the custom template and supply the DNS names. Complete the enrollment to obtain the certificate in the Personal store.
• On a domain-joined helper system: Alternatively, use a domain-joined machine to request on behalf of the node (using the “Enroll on behalf” feature with an Enrollment Agent certificate, or simply request and then export/import). This is more complex and usually not needed unless policy restricts direct enrollment.

After obtaining each certificate, verify on the node that it appears in Certificates (Local Computer) > Personal > Certificates. The Issued To should be the node’s FQDN, and on the Details tab you should see the required EKUs and SAN entries. Also import the CA’s Root CA certificate into Trusted Root Certification Authorities on each node (the certreq -accept step may do this automatically if the chain is provided; if not, manually import the CA root). A quick check using the Certificates MMC or PowerShell can confirm trust. For example, to check via PowerShell:

Get-ChildItem Cert:\LocalMachine\My | Where-Object {$_.Subject -like "*Node1*"} | Select-Object Subject, EnhancedKeyUsageList, NotAfter

Make sure the EnhancedKeyUsageList shows both Client and Server Authentication and that NotAfter (expiry) is a reasonable date. Also ensure no errors about untrusted issuer – the Certificate status should show “This certificate is OK”.

Option: Self-Signed Certificates for Testing

For a lab or proof-of-concept (where an enterprise CA is not available), you can use self-signed certificates. The key is to create a self-signed cert that includes the proper names and EKUs, and then trust that cert across all nodes. Use PowerShell New-SelfSignedCertificate with appropriate parameters. For example, on Node1:

$cert = New-SelfSignedCertificate -DnsName "Node1.mylocal.net", "Node1", "Cluster1.mylocal.net", "Cluster1" `
-CertStoreLocation Cert:\LocalMachine\My `
-KeyUsage DigitalSignature, KeyEncipherment `
-TextExtension @("2.5.29.37={text}1.3.6.1.5.5.7.3.1;1.3.6.1.5.5.7.3.2")

This creates a certificate for Node1 with the specified DNS names and both ServerAuth/ClientAuth EKUs. Repeat on Node2 (adjusting names accordingly). Alternatively, you can generate a temporary root CA certificate and then issue child certificates to each node (PowerShell’s -TestRoot switch simplifies this by generating a root and end-entity cert together).

If you created individual self-signed certs per node, export each node’s certificate (without the private key) and import it into the Trusted People or Trusted Root store of the other nodes. (Trusted People works for peer trust of specific certs; Trusted Root works if you created a root CA and issued from it). For example, if Node1 and Node2 each have self-signed certs, import Node1’s cert as a Trusted Root on Node2 and vice versa. This is required because self-signed certs are not automatically trusted.

Using CA-issued certs is strongly recommended for production. Self-signed certs should only be used in test environments, and if used, monitor and manually renew them before expiration (since there’s no CA to do it). A lot of problems have occurred in production systems because people used self signed certs and forgot that they expire. 

With certificates in place, we can configure Windows Remote Management (WinRM) to use them. WinRM is the service behind PowerShell Remoting and many remote management tools. By default, WinRM uses HTTP (port 5985) and authenticates via Kerberos or NTLM. In a workgroup scenario, NTLM over HTTP would be used – we want to avoid that. Instead, we will enable WinRM over HTTPS (port 5986) with our certificates, providing encryption and the ability to use certificate-based authentication for management sessions.

Perform these steps on each cluster node:

1. Verify certificate for WinRM: WinRM requires a certificate in the Local Computer Personal store that has a Server Authentication EKU and whose Subject or SAN matches the hostname. We have already enrolled such a certificate for each node. Double-check that the certificate’s Issued To (CN or one of the SAN entries) exactly matches the hostname that clients will use (e.g. the FQDN). If you plan to manage via short name, ensure the short name is in SAN; if via FQDN, that’s covered by CN or SAN. The certificate must not be expired or revoked, and it should be issued by a CA that the clients trust (not self-signed unless the client trusts it).
2. Enable the HTTPS listener: Open an elevated PowerShell on the node and run:

winrm quickconfig -transport:https

This command creates a WinRM listener on TCP 5986 bound to the certificate. If it says no certificate was found, you may need to specify the certificate manually. You can do so with:

# Find the certificate thumbprint (assuming only one with Server Auth)
$thumb = (Get-ChildItem Cert:\LocalMachine\My | Where-Object {$_.EnhancedKeyUsageList -match "Server Authentication"} | Select-Object -First 1 -ExpandProperty Thumbprint)
New-Item -Path WSMan:\LocalHost\Listener -Transport HTTPS -Address * -CertificateThumbprint $thumb -Force

Verify listeners with:

winrm enumerate winrm/config/listener

You should see an HTTPS listener with hostname, listening on 5986, and the certificate’s thumbprint. WinRM will automatically choose a certificate that meets the criteria (if multiple are present, it picks the one with CN matching machine name, so ideally use a unique cert to avoid ambiguity).

Disable unencrypted/HTTP access (optional but recommended): Since we want all remote management encrypted and to eliminate NTLM, you can disable the HTTP listener. Run:

Remove-WSManInstance -ResourceURI winrm/config/Listener -SelectorSet @{Address="*", Transport="HTTP"}

This ensures WinRM is only listening on HTTPS. Also, you may configure the WinRM service to reject unencrypted traffic and disallow Basic authentication to prevent any fallback to insecure methods:

winrm set winrm/config/service '@{AllowUnencrypted="false"}'
winrm set winrm/config/service/auth '@{Basic="false"}'

(By default, AllowUnencrypted is false anyway when HTTPS is used, and Basic is false unless explicitly enabled.)

TrustedHosts (if needed): In a workgroup, WinRM won’t automatically trust hostnames for authentication. However, when using certificate authentication, the usual TrustedHosts requirement may not apply in the same way as for NTLM/Negotiate. If you plan to authenticate with username/password over HTTPS (e.g. using Basic or default CredSSP), you will need to add the other nodes (or management station) to the TrustedHosts list on each node. This isn’t needed for the cluster’s internal communication (which uses certificates via clustering, not WinRM), but it might be needed for your remote PowerShell sessions depending on method. To allow all (not recommended for security), you could do:

Set-Item WSMan:\localhost\Client\TrustedHosts -Value "*"

Or specify each host:

Set-Item WSMan:\localhost\Client\TrustedHosts -Value "Node1,Node2,Cluster1"

This setting allows the local WinRM client to talk to those remote names without Kerberos. If you will use certificate-based authentication for WinRM (where the client presents a cert instead of username/password), TrustedHosts is not required – certificate auth doesn’t rely on host trust in the same way.

(Optional) Configure certificate authentication for admin access: One of the benefits of HTTPS listener is you can use certificate mapping to log in without a password. For advanced users, you can issue a client certificate for yourself (with Client Authentication EKU), then configure each server to map that cert to a user (for example, map to the local Administrator account). This involves creating a mapping entry in winrm/config/service/certmapping. For instance:

# Example: map a client cert by its subject to a local account
winrm create winrm/config/service/certmapping @{CertificateIssuer= "CN=YourCA"; Subject="CN=AdminUserCert"; Username="Administrator"; Password="<adminPassword>"; Enabled="true"}

Then from your management machine, you can use that certificate to authenticate. While powerful, this goes beyond the core cluster setup, so we won’t detail it further. Without this, you can still connect to the nodes using Enter-PSSession -ComputerName Node1 -UseSSL -Credential Node1\Administrator (which will prompt for the password but send it safely over the encrypted channel).

At this point, we have each node prepared with a trusted certificate and WinRM listening securely. Test the connectivity: from one node, try to start a PowerShell remote session to the other using HTTPS. For example, on Node1 run:

Test-WsMan Node2 -UseSSL
Enter-PSSession -ComputerName Node2 -UseSSL -Credential Node2\Administrator

You should connect without credential errors or warnings (you may get a certificate trust prompt if the client machine doesn’t trust the server cert — make sure the CA root is in the client’s trust store as well). Once you can manage nodes remotely over HTTPS, you’re ready to create the cluster.

All cluster nodes need the Hyper-V role (for running VMs) and the Failover Clustering feature. We will use PowerShell to install these simultaneously on each server. On each node: Open an elevated PowerShell (locally or via your new WinRM setup) and run:

Install-WindowsFeature -Name Failover-Clustering, Hyper-V -IncludeManagementTools -Restart

This installs the Hyper-V hypervisor, the clustering feature, and management tools (including the Failover Cluster Manager and Hyper-V Manager GUI, and PowerShell modules). The server will restart if Hyper-V was not previously enabled (we include -Restart for convenience). After reboot, run the command on the next node (if doing it remotely, do one at a time). Alternatively, use the Server Manager GUI or Install-WindowsFeature without -Restart and reboot manually. After all nodes are back up, verify the features:

Get-WindowsFeature -Name Hyper-V, Failover-Clustering

It should show both as Installed. Also confirm the Failover Clustering PowerShell module is available (Get-Module -ListAvailable FailoverClusters) and the Cluster service is installed (though not yet configured).

Cluster service account: Windows Server 2016+ automatically creates a local account called CLIUSR used by the cluster service for internal communication. Ensure this account was created (Computer Management > Users). We won’t interact with it directly, but be aware it exists. Do not delete or disable CLIUSR – the cluster uses it alongside certificates for bootstrapping. (All cluster node communications will now use either Kerberos or certificate auth; NTLM is not needed in WS2019+ clusters.)

Now that you’ve backflipped and shenaniganed with all the certificates, you can actually get around to building the cluster.

Here we will create the cluster and add nodes to it using PowerShell. The cluster will use a DNS name for its administrative access point (since there is no Active Directory for a traditional cluster computer object). The basic steps are:

• Validate the configuration (optional but recommended).
• Create the cluster (initially with one node to avoid cross-node authentication issues).
• Join additional node(s) to the cluster.
• Configure cluster networking, quorum, and storage (CSV).

Validate the Configuration (Cluster Validation)

It’s good practice to run the cluster validation tests to catch any misconfiguration or hardware issues before creating the cluster. Microsoft supports a cluster only if it passes validation or if any errors are acknowledged as non-critical.

Run the following from one of the nodes (this will reach out to all nodes):

Test-Cluster -Node Node1.mylocal.net, Node2.mylocal.net

Replace with your actual node names (include all 2 or 4 nodes). The cmdlet will run a series of tests (network, storage, system settings). Ensure that all tests either pass or only have warnings that you understand. For example, warnings about “no storage is shared among all nodes” are expected if you haven’t yet configured iSCSI or if using SMB (you can skip storage tests with -Skip Storage if needed). If critical tests fail, resolve those issues (networking, disk visibility, etc.) before proceeding.

Create the Cluster (with the First Node)

On one node (say Node1), use the New-Cluster cmdlet to create the cluster with that node as the first member. By doing it with a single node initially, we avoid remote authentication at cluster creation time (no need for Node1 to authenticate to Node2 yet):

New-Cluster -Name "Cluster1" -Node Node1 -StaticAddress "10.0.0.100" -AdministrativeAccessPoint DNS

Here:

• -Name is the intended cluster name (this will be the name clients use to connect to the cluster, e.g. for management or as a CSV namespace prefix). We use “Cluster1” as an example.
• -Node Node1 specifies which server to include initially (Node1’s name).
• -StaticAddress sets the cluster’s IP address (choose one in the same subnet that is not in use; this IP will be brought online as the “Cluster Name” resource). In this example 10.0.0.100 is the cluster IP.
• -AdministrativeAccessPoint DNS indicates we’re creating a DNS-only cluster (no AD computer object). This is the default in workgroup clusters, but we specify it explicitly for clarity.

The command will proceed to create the cluster service, register the cluster name in DNS (if DNS is configured and dynamic updates allowed), and bring the core cluster resources online. It will also create a cluster-specific certificate (self-signed) for internal use if needed, but since we have our CA-issued certs in place, the cluster may use those for node authentication.

Note: If New-Cluster fails to register the cluster name in DNS (common in workgroup setups), you might need to create a manual DNS A record for “Cluster1” pointing to 10.0.0.100 in whatever DNS server the nodes use. Alternatively, add “Cluster1” to each node’s hosts file (as we did in prerequisites). This ensures that the cluster name is resolvable. The cluster will function without AD, but it still relies on DNS for name resolution of the cluster name and node names.

At this point, the cluster exists with one node (Node1). You can verify by running cluster cmdlets on Node1, for example: Get-Cluster (should list “Cluster1”) and Get-ClusterNode (should list Node1 as up). In Failover Cluster Manager, you could also connect to “Cluster1” (or to Node1) and see the cluster.

Add Additional Nodes to the Cluster

Now we will add the remaining node(s) to the cluster:

On each additional node, run the following (replace “Node2” with the name of that node and adjust cluster name accordingly):

Add-ClusterNode -Cluster Cluster1 -Name Node2

Run this on Node2 itself (locally). This instructs Node2 to join the cluster named Cluster1. Because Node2 can authenticate the cluster (Node1) via the cluster’s certificate and vice versa, the join should succeed without prompting for credentials. Under the hood, the cluster service on Node2 will use the certificate (and CLIUSR account) to establish trust with Node1’s cluster service.

Repeat the Add-ClusterNode command on each additional node (Node3, Node4, etc. one at a time). After each join, verify by running Get-ClusterNode on any cluster member – the new node should show up and status “Up”.

If for some reason you prefer a single command from Node1 to add others, you could use:

# Run on Node1:
Add-ClusterNode -Name Node2, Node3 -Cluster Cluster1

This would attempt to add Node2 and Node3 from Node1. It may prompt for credentials or require TrustedHosts if no common auth is present. Using the local Add-ClusterNode on each node avoids those issues by performing the action locally. Either way, at the end all nodes should be members of Cluster1.

Quorum configuration is critical, especially with an even number of nodes. The cluster will already default to Node Majority (no witness) or may try to assign a witness if it finds eligible storage. 

Use a witness to avoid a split-brain scenario. If you have a small shared disk (LUN) visible to both nodes, that can be a Disk Witness. Alternatively, use a Cloud Witness (Azure). To configure a disk witness, first make sure the disk is seen as Available Storage in the cluster, then run:

Get-ClusterAvailableDisk | Add-ClusterDisk
Set-ClusterQuorum -Cluster Cluster1 -NodeAndDiskMajority 0 /disk:<DiskResourceName>

(Replace <DiskResourceName> with the name or number of the disk from Get-ClusterResource). Using Failover Cluster Manager, you can run the Configure Cluster Quorum wizard and select “Add a disk witness”. If no shared disk is available, the Cloud Witness is an easy option (requires an Azure Storage account and key). For cloud witness:

Set-ClusterQuorum -Cluster Cluster1 -CloudWitness -AccountName "<StorageAccount>" -AccessKey "<Key>"

Do not use a File Share witness – as noted earlier, file share witnesses are not supported in workgroup clusters because the cluster cannot authenticate to a remote share without AD.

A 4-node cluster can sustain two node failures if properly configured. It’s recommended to also configure a witness for even-number clusters to avoid a tie (2–2) during a dual-node failure scenario. A disk or cloud witness is recommended (same process as above). With 4 nodes, you would typically use Node Majority + Witness. The cluster quorum wizard can automatically choose the best quorum config (typically it will pick Node Majority + Witness if you run the wizard and have a witness available).

You can verify the quorum configuration with Get-ClusterQuorum. Make sure it lists the witness you configured (if any) and that the cluster core resources show the witness online.

Add Cluster Shared Volumes (CSV) or Configure VM Storage

Next, prepare storage for Hyper-V VMs. If using a shared disk (Block storage like iSCSI/SAN), after adding the disks to the cluster (they should appear in Storage > Disks in Failover Cluster Manager), you can enable Cluster Shared Volumes (CSV). CSV allows all nodes to concurrently access the NTFS/ReFS volume, simplifying VM placement and live migration. To add available cluster disks as CSV volumes:

Get-ClusterDisk | Where-Object IsClustered -eq $true | Add-ClusterSharedVolume

This will take each clustered disk and mount it as a CSV under C:\ClusterStorage\ on all nodes. Alternatively, right-click the disk in Failover Cluster Manager and choose Add to Cluster Shared Volumes. Once done, format the volume (if not already formatted) with NTFS or ReFS via any node (it will be accessible as C:\ClusterStorage\Volume1\ etc. on all nodes). Now this shared volume can store all VM files, and any node can run any VM using that storage.

If using an SMB 3 share (NAS or file server), you won’t add this to cluster storage; instead, each Hyper-V host will connect to the SMB share directly. Ensure each node has access credentials for the share. In a workgroup, that typically means the NAS is also in a workgroup and you’ve created a local user on the NAS that each node uses (via stored credentials) – this is outside the cluster’s control. Each node should be able to New-SmbMapping or simply access the UNC path. Test access from each node (e.g. Dir \\NAS\HyperVShare). In Hyper-V settings, you might set the Default Virtual Hard Disk Path to the UNC or just specify the UNC when creating VMs. Note: Hyper-V supports storing VMs on SMB 3.0 shares with Kerberos or certificate-based authentication, but in a workgroup you’ll likely rely on a username/password for the share (which is a form of local account usage at the NAS). This doesn’t affect cluster node-to-node auth, but it’s a consideration for securing the NAS.

At this stage, run some quick checks to ensure the cluster is healthy:

• Get-Cluster – should show the cluster name, IP, and core resources online.
• Get-ClusterNode – all nodes should be Up.
• Get-ClusterResource – should list resources (Cluster Name, IP Address, any witness, any disks) and their state (Online). The Cluster Name resource will be of type “Distributed Network Name” since this is a DNS-only cluster.
• Use Failover Cluster Manager (you can launch it on one of the nodes or from RSAT on a client) to connect to “Cluster1”. Ensure you can see all nodes and storage. When prompted to connect, use <clustername> or <clusterIP> – with our certificate setup, it may be best to connect by cluster name (make sure DNS/hosts is resolving it to the cluster IP). If a certificate trust warning appears, it might be because the management station doesn’t trust the cluster node’s cert or you connected with a name not in the SAN. As a workaround, connect directly to a node in cluster manager (e.g. Node1), which then enumerates the cluster.

Now you have a functioning cluster ready for Hyper-V workloads, with secure authentication between nodes. Next, we configure Hyper-V specific settings like Live Migration.

One major benefit introduced in Windows Server 2025 is support for Live Migration in workgroup clusters (previously, live migration required Kerberos and thus a domain). In WS2025, cluster nodes use certificates to mutually authenticate for live migration traffic. This allows VMs to move between hosts with no downtime even in the absence of AD. We will enable and tune live migration for our cluster.

By default, the Hyper-V role might have live migration disabled (for non-clustered hosts). In a cluster, it may be auto-enabled when the Failover Clustering and Hyper-V roles are both present, but to ensure it it, run:

Enable-VMMigration

This enables the host to send/receive live migrations. In PowerShell, no output means success. (In Hyper-V Manager UI, this corresponds to ticking “Enable incoming and outgoing live migrations” in the Live Migrations settings.)

In a workgroup, the only choice in UI would be CredSSP (since Kerberos requires domain). CredSSP means you must initiate the migration from a session where you are logged onto the source host so your credentials can be delegated. We cannot use Kerberos here, but the cluster’s internal PKU2U certificate mechanism will handle node-to-node auth for us when orchestrated via Failover Cluster Manager. No explicit setting is needed for cluster-internal certificate usage & Windows will use it automatically for the actual live migration operation.  If you were to use PowerShell, the default MigrationAuthenticationType is CredSSP for workgroup. You can confirm (or set explicitly, though not strictly required):

Set-VMHost -VirtualMachineMigrationAuthenticationType CredSSP

(This can be done on each node; it just ensures the Hyper-V service knows to use CredSSP which aligns with our need to initiate migrations from an authenticated context.)

If your cluster nodes were domain-joined, Windows Server 2025 enables Credential Guard which blocks CredSSP by default. In our case (workgroup), Credential Guard is not enabled by default, so CredSSP will function. Just be aware if you ever join these servers to a domain (or they were once joined to a domain before being demoted to a workgroup), you’d need to configure Kerberos constrained delegation or disable Credential Guard to use live migration.

For security and performance, do not use the management network for VM migration if you have other NICs. We will designate the dedicated network (e.g. “LMNet” or a specific subnet) for migrations. You can configure this via PowerShell or Failover Cluster Manager. Using PowerShell, run the following on each node:

# Example: allow LM only on 10.0.1.0/24 network (where 10.0.1.5 is this node's IP on that network)
Set-VMMigrationNetwork 10.0.1.5
Set-VMHost -UseAnyNetworkForMigration $false

The Set-VMMigrationNetwork cmdlet adds the network associated with the given IP to the allowed list for migrations. The second cmdlet ensures only those designated networks are used. Alternatively, if you have the network name or interface name, you might use Hyper-V Manager UI: under each host’s Hyper-V Settings > Live Migrations > Advanced Features, select Use these IP addresses for Live Migration and add the IP of the LM network interface. In a cluster, these settings are typically per-host. It’s a good idea to configure it identically on all nodes.

Verify the network selection by running: Get-VMHost | Select -ExpandProperty MigrationNetworks. It should list the subnet or network you allowed, and UseAnyNetworkForMigration should be False.

Windows can either send VM memory over TCP, compress it, or use SMB Direct (if RDMA is available) for live migration. By default in newer Windows versions, compression is used as it offers a balance of speed without special hardware. If you have a very fast dedicated network (10 Gbps+ or RDMA), you might choose SMB to leverage SMB Multichannel/RDMA for highest throughput. To set this:

# Options: TCPIP, Compression, SMB
Set-VMHost -VirtualMachineMigrationPerformanceOption Compression

(Do this on each node; “Compression” is usually default on 2022/2025 Hyper-V.) If you select SMB, ensure your cluster network is configured to allow SMB traffic and consider enabling SMB encryption if security is a concern (SMB encryption will encrypt the live migration data stream). Note that if you enable SMB encryption or cluster-level encryption, it could disable RDMA on that traffic, so only enable it if needed, or rely on the network isolation as primary protection.

Depending on your hardware, you may allow multiple VMs to migrate at once. The default is usually 2 simultaneous live migrations. You can increase this if you have capacity:

Set-VMHost -MaximumVirtualMachineMigrations 4 -MaximumStorageMigrations 2

Adjust numbers as appropriate (and consider that cluster-level property (Get-Cluster).MaximumParallelMigrations might override host setting in a cluster). This setting can also be found in Hyper-V Settings UI under Live Migrations.

With these configured, live migration is enabled.

Test a live migration:

Create a test VM (or if you have VMs, pick one) and attempt to move it from one node to another using Failover Cluster Manager or PowerShell:

• In Failover Cluster Manager, under Roles, right-click a virtual machine, choose Live Migrate > Select Node… and pick another node. The VM should migrate with zero downtime. If it fails, check for error messages regarding authentication. Ensure you initiated the move from a node where you’re an admin (or via cluster manager connected to the cluster with appropriate credentials). The cluster will handle the mutual auth using the certificates (this is transparent – behind the scenes, the nodes use the self-created PKU2U cert or our installed certs to establish a secure connection for VM memory transfer).
• Alternatively, use PowerShell:

Move-ClusterVirtualMachineRole -Name "<VM resource name>" -Node <TargetNode>

This cmdlet triggers a cluster-coordinated live migration (the cluster’s Move operation will use the appropriate auth). If the migration succeeds, congratulations – you have a fully functional Hyper-V cluster without AD! 

Security Best Practices Recap and Additional Hardening

Additional best practices for securing a workgroup Hyper-V cluster include:

• Certificate Security: The private keys of your node certificates are powerful – protect them. They are stored in the machine store (and likely marked non-exportable). Only admins can access them; ensure no unauthorized users are in the local Administrators group. Plan a process for certificate renewal before expiration. If using an enterprise CA, you might issue certificates with a template that allows auto-renewal via scripts or at least track their expiry to re-issue and install new certs on each node in time. The Failover Cluster service auto-generates its own certificates (for CLIUSR/PKU2U) and auto-renews them, but since we provided our own, we must manage those. Stagger renewals to avoid all nodes swapping at once (the cluster should still trust old vs new if the CA is the same). It may be wise to overlap: install new certs on all nodes and only then remove the old, so that at no point a node is presenting a cert the others don’t accept (if you change CA or template).
• Trusted Root and Revocation: All nodes trust the CA – maintain the security of that CA. Do not include unnecessary trust (e.g., avoid having nodes trust public CAs that they don’t need). If possible, use an internal CA that is only used for these infrastructure certs. Keep CRLs (Certificate Revocation Lists) accessible if your cluster nodes need to check revocation for each other’s certs (though cluster auth might not strictly require online revocation checking if the certificates are directly trusted). It’s another reason to have a reasonably long-lived internal CA or offline root.
• Disable NTLM: Since clustering no longer needs NTLM as of Windows 2019+, you can consider disabling NTLM fallback on these servers entirely for added security (via Group Policy “Network Security: Restrict NTLM: Deny on this server” etc.). However, be cautious: some processes (including cluster formation in older versions, or other services) might break. In our configuration, cluster communications should use Kerberos or cert. If these servers have no need for NTLM (no legacy apps), disabling it eliminates a whole class of attacks. Monitor event logs (Security log events for NTLM usage) if you attempt this. The conversation in the Microsoft tech community indicates by WS2022, cluster should function with NTLM disabled, though a user observed issues when CLIUSR password rotated if NTLM was blocked. WS2025 should further reduce any NTLM dependency.
• PKU2U policy: The cluster uses the PKU2U security provider for peer authentication with certificates. There is a local security policy “Network security: Allow PKU2U authentication requests to this computer to use online identities” – this must be enabled (which it is by default) for clustering to function properly. Some security guides recommend disabling PKU2U; do not disable it on cluster nodes (or if your organization’s baseline GPO disables it, create an exception for these servers). Disabling PKU2U will break the certificate-based node authentication and cause cluster communication failures.
• Firewall: We opened WinRM over 5986. Ensure Windows Firewall has the Windows Remote Management (HTTPS-In) rule enabled. The Failover Clustering feature should have added rules for cluster heartbeats (UDP 3343, etc.) and SMB (445) if needed. Double-check that on each node the Failover Cluster group of firewall rules is enabled for the relevant profiles (if your network is Public, you might need to enable the rules for Public profile manually, or set network as Private). Also, for live migration, if using SMB transport, enable SMB-in rules. If you enabled SMB encryption, it uses the same port 445 but encrypts payloads.
• Secure Live Migration Network: Ideally, the network carrying live migration is isolated (not routed outside of the cluster environment). If you want belt-and-suspenders security, you could implement IPsec encryption on live migration traffic. For example, require IPsec (with certificates) between the cluster nodes on the LM subnet. However, this can be complex and might conflict with SMB Direct/RDMA. Another simpler approach: since we can rely on our certificate mutual auth to prevent unauthorized node communication, focus on isolating that traffic so even if someone tapped it, you can optionally turn on SMB encryption for LM (when using SMB transport) which will encrypt the VM memory stream. At minimum, treat the LM network as sensitive, as it carries VM memory contents in clear text if not otherwise encrypted.
• Secure WinRM/management access: We configured WinRM for HTTPS. Make sure to limit who can log in via WinRM. By default, members of the Administrators group have access. Do not add unnecessary users to Administrators. You can also use Local Group Policy to restrict WinRM service to only allow certain users or certificate mappings. Since this is a workgroup, there’s no central AD group; you might create a local group for “Remote Management Users” and configure WSMan to allow members of that group (and only put specific admin accounts in it). Also consider enabling PowerShell Just Enough Administration (JEA) if you want to delegate specific tasks without full admin rights, though that’s advanced.
• Hyper-V host security: Apply standard Hyper-V best practices: enable Secure Boot for Gen2 VMs, keep the host OS minimal (consider using Windows Server Core for fewer attack surface, if feasible), and ensure only trusted administrators can create or manage VMs. Since this cluster is not in a domain, you won’t have AD group-based access control; consider using Authentication Policies like LAPS for unique local admin passwords per node.
• Monitor cluster events: Monitor the System event log for any cluster-related errors (clustering will log events if authentication fails or if there are connectivity issues). Also monitor the FailoverClustering event log channel. Any errors about “unable to authenticate” or “No logon servers” etc., would indicate certificate or connectivity problems.
• Test failover and failback: After configuration, test that VMs can failover properly. Shut down one node and ensure VMs move to other node automatically. When the node comes back, you can live migrate them back. This will give confidence that the cluster’s certificate-based auth holds up under real failover conditions.
• Consider Management Tools: Tools like Windows Admin Center (WAC) can manage Hyper-V clusters. WAC can be configured to use the certificate for connecting to the nodes (it will prompt to trust the certificate if self-signed). Using WAC or Failover Cluster Manager with our setup might require launching the console from a machine that trusts the cluster’s cert and using the cluster DNS name. Always ensure management traffic is also encrypted (WAC uses HTTPS and our WinRM is HTTPS so it is).