Hyper-V 2012 R2 using Failover Clustering with CSV and SCVMM 2012

• 2+ identical Windows Server 2012 R2 hosts with hardware virtualization + latest BIOS/firmware

• Joined to the same AD domain, healthy DNS, accurate time (NTP)

• Shared storage reachable by all hosts (iSCSI, Fibre Channel, or SMB 3.0 for Hyper-V)

• At least 3 networks (recommended):

  • Management (host/cluster/AD)

  • Live Migration (and/or SMB Direct if you have RDMA)

  • Storage (iSCSI/SMB3) — never route this over the mgmt network

Service/RunAs account for VMM with local admin on hosts and AD rights for cluster object creation (details below)


1) Active Directory prep (cluster object permissions)

Failover clustering creates a Cluster Name Object (CNO) in AD.

• Put your Hyper-V hosts into an OU.

• Either:

  • Pre-stage the CNO (create a disabled computer object “HVCLU01”), and grant the host computer objects “Full Control” on it, or

  • Grant the hosts (or the VMM RunAs account) Create Computer Objects on that OU.

This avoids classic “logon type” / “cannot create computer object” errors during cluster creation.

Host OS features & basic config (on every host)

Run these on each node (elevated PowerShell):

# Hyper-V, Failover Clustering, MPIO, and tools
Install-WindowsFeature Hyper-V, Failover-Clustering, Multipath-IO -IncludeManagementTools

# Optional: enable MPIO for iSCSI
mpclaim.exe -r -i -a "VEN_MSFT&PROD_iSCSI"   # adjust to your array’s Vendor/Product as needed

# (If you’ll use SMB Direct/RDMA)
Enable-NetAdapterRDMA -Name "NIC-LM","NIC-Storage"

# Create a NIC Team (LBFO) for Management/VM traffic (optional, example)
New-NetLbfoTeam -Name "Team0" -TeamMembers "NIC1","NIC2" -TeamingMode SwitchIndependent -LoadBalancingAlgorithm Dynamic

# Create a Hyper-V vSwitch on the team
New-VMSwitch -Name "vSwitch-Prod" -NetAdapterName "Team0" -AllowManagementOS $true

Network tips (2012 R2):

• Keep Live Migration on its own high-bandwidth network (enable Compression or SMB LM later).

• Use Jumbo frames consistently on storage/LM if your network supports it.

• If using SMB 3.0 storage or SMB-based Live Migration, enable RDMA (RoCE/iWARP) where possible.

Connect shared storage to all hosts

iSCSI example

On each host:

# Start iSCSI service
Set-Service MSiSCSI -StartupType Automatic
Start-Service MSiSCSI

# Discover & connect targets
New-IscsiTargetPortal -TargetPortalAddress "10.10.40.10"
Get-IscsiTarget | Connect-IscsiTarget -IsPersistent $true -IsMultipathEnabled $true

Format & align disks (one host only)

1. Bring new LUNs online, initialize GPT, create volumes (NTFS for CSV on 2012 R2), no drive letters needed (Cluster will re-present as C:\ClusterStorage\...).

2. Don’t put anything on them yet; clustering will convert to CSVFS layer.

For SMB 3.0 (Scale-Out File Server): present a continuously available SMB share from SOFS to all Hyper-V hosts; CSV step is not used in that model.

Install Hyper-V & set default paths (each host)

# Optional: set default VM paths now or let VMM control later
Set-VMHost -VirtualMachinePath "D:\VMs" -VirtualHardDiskPath "D:\VMs\VHDs"

Validate and create the Windows Failover Cluster

On one host (or any admin workstation with the tools):

# Validation (must be clean)
Test-Cluster -Node HVHOST01,HVHOST02 -Include "Inventory","Network","Storage"

# Create the cluster (without grabbing storage yet)
New-Cluster -Name "HVCLU01" -Node HVHOST01,HVHOST02 -StaticAddress 10.10.10.50 -NoStorage

# Optional: set quorum to a file share witness or disk witness
Set-ClusterQuorum -FileShareWitness "\\fileserver\witness$"

Add disks into the cluster, then convert to CSV:

# See available cluster disks
Get-ClusterAvailableDisk | Add-ClusterDisk

# Convert selected disks to CSV (repeat per disk)
Add-ClusterSharedVolume -Name "Cluster Disk 1"

You’ll see them under C:\ClusterStorage\Volume1, Volume2, etc.

Cluster networking & Live Migration settings

Mark cluster networks for their intended roles:

# List cluster networks
Get-ClusterNetwork | ft Name, Address, Role

# Example: Storage network = ClusterOnly, LM network = ClusterAndClient
(Get-ClusterNetwork "StorageNet").Role = 1           # ClusterOnly
(Get-ClusterNetwork "LiveMigrationNet").Role = 3     # ClusterAndClient

# Live Migration: prefer LM network, enable compression (2012 R2 feature)
(Get-VMHost).VirtualMachineMigrationPerformanceOption = "Compression"
Set-VMHost -VirtualMachineMigrationEnabled $true -VirtualMachineMigrationAuthenticationType CredSSP -VirtualMachineMigrationMaximum 4

# Make sure LM network metrics prefer the LM VLAN/subnet
Set-ClusterNetwork -Name "LiveMigrationNet" -Metric 100
Set-ClusterNetwork -Name "ManagementNet" -Metric 1000

Test a basic clustered VM (before VMM)

On one host:

# Quick test VM on CSV
New-VM -Name "TestVM01" -MemoryStartupBytes 2GB -Generation 2 -Path "C:\ClusterStorage\Volume1\VMs" -SwitchName "vSwitch-Prod"
New-VHD -Path "C:\ClusterStorage\Volume1\VMs\TestVM01\TestVM01.vhdx" -SizeBytes 60GB -Dynamic
Add-VMHardDiskDrive -VMName "TestVM01" -Path "C:\ClusterStorage\Volume1\VMs\TestVM01\TestVM01.vhdx"

Start-VM TestVM01
Move-ClusterVirtualMachineRole -Name "TestVM01" -Node HVHOST02   # Live migrate test

Bring in SCVMM 2012 R2 (recommended)

Install VMM 2012 R2 + latest Update Rollup, SQL instance, and VMM Library.

In VMM:

1. Create a Run As Account (e.g., VMM-SVC) that is:

   • Local admin on hosts

   • Has AD rights to create the cluster CNO (or you pre-staged the CNO)

2. Add your hosts:
   Fabric ➜ Servers ➜ Add Resources ➜ Hyper-V Hosts and Clusters ➜ Add by Windows Server Computers in a trusted AD domain ➜ choose RunAs.

3. (Optional but strongly recommended) Model networking in VMM:

   • Logical Network(s) (e.g., Prod, LiveMigration, Storage)

   • IP Pools for each as needed

   • Port Classifications (e.g., “Host Mgmt”, “VM Prod”)

   • Uplink Port Profiles and a Logical Switch (SR-IOV, RSS, etc. per your NICs)

   • Apply the Logical Switch to each host’s NIC team

4. Storage integration (optional):

   • Add your storage provider (SMI-S, SMB SOFS) to VMM (Fabric ➜ Storage).

   • Classify storage (e.g., Gold/Silver) and assign LUNs to the cluster via VMM.

5. Create (or import) the cluster in VMM:

   • If the cluster already exists, VMM will discover it when you add both hosts in the same Host Group.

   • Or use Create Host Cluster wizard in VMM to build it end-to-end.

6. Place VMs via VMM onto CSV paths under C:\ClusterStorage\….
   Use Availability Sets for anti-affinity, and set Preferred Owners/VM priority if needed.

CSV & storage best practices (2012 R2)

• CSV on NTFS for 2012 R2 (ReFS v1 back then lacked needed Hyper-V features).

• One CSV per tenant/tier is common; don’t overload a single CSV.

• ODX, UNMAP, and MPIO: enable where supported by your array.

• Keep host pagefile and temp files off CSV.

• Antivirus: exclude C:\ClusterStorage, VM process/hypervisor files (per MS AV guidance).

Patching & maintenance mode

• Use Cluster-Aware Updating (CAU) or VMM Servicing to live-migrate VMs and patch nodes one at a time.

• In VMM: Drain a host (maintenance mode), patch, reboot, return to service.

Enable monitoring and backups

• VMM + SCOM (optional) for performance/health.

• Backup: VSS-aware Hyper-V solution targeting host-level backups, not in-guest only.

• Make sure backup vendors support CSV redirection during snapshot operations on 2012 R2.

Common gotchas (and quick fixes)

• Cluster creation fails with computer object/CNO errors
  → Pre-stage the CNO disabled, or grant Create Computer Objects on the hosts’ OU to the hosts or VMM RunAs.

• Logon failure / “allow log on locally” type issues for service accounts
  → The VMM RunAs account needs to be local admin on hosts; ensure domain policy isn’t blocking interactive/remote logon required by agents.

• Disks won’t release / stuck as “clustered” after a failed attempt
  → Use Get-ClusterResource/Remove-ClusterResource or Clear-ClusterDiskReservation on each node; worst case, offline LUN to all nodes then re-present cleanly.

• Live Migration slow
  → Switch to Compression or SMB-based LM and ensure a dedicated high-bandwidth network; verify Jumbo/RDMA end-to-end.

Minimal “golden path” commands (copy/paste friendly)

# On each host
Install-WindowsFeature Hyper-V, Failover-Clustering, Multipath-IO -IncludeManagementTools
New-NetLbfoTeam -Name Team0 -TeamMembers NIC1,NIC2 -TeamingMode SwitchIndependent -LoadBalancingAlgorithm Dynamic
New-VMSwitch -Name vSwitch-Prod -NetAdapterName Team0 -AllowManagementOS $true

# iSCSI (each host)
Set-Service MSiSCSI -StartupType Automatic; Start-Service MSiSCSI
New-IscsiTargetPortal -TargetPortalAddress 10.10.40.10
Get-IscsiTarget | Connect-IscsiTarget -IsPersistent $true -IsMultipathEnabled $true

# Create the cluster (one time)
Test-Cluster -Node HVHOST01,HVHOST02 -Include "Inventory","Network","Storage"
New-Cluster -Name HVCLU01 -Node HVHOST01,HVHOST02 -StaticAddress 10.10.10.50 -NoStorage
Set-ClusterQuorum -FileShareWitness \\fileserver\witness$

# Add disks & convert to CSV
Get-ClusterAvailableDisk | Add-ClusterDisk
Add-ClusterSharedVolume -Name "Cluster Disk 1"

# Live Migration tuning
(Get-VMHost).VirtualMachineMigrationPerformanceOption = "Compression"
Set-VMHost -VirtualMachineMigrationEnabled $true -VirtualMachineMigrationAuthentic