Skip to content

How to Create a Storage Spaces Pool and Mirrored Virtual Disk Print

  • 0

Applies to: Windows Server 2019, Windows Server 2022 and Windows Server 2025, subject to the workload and edition requirements below.

Understand the pool, virtual disk and volume layers

Standalone Storage Spaces combines selected physical disks into a storage pool. A virtual disk created from that pool defines resiliency and allocation. Windows then partitions and formats that virtual disk to create a usable volume. Keeping these layers distinct helps avoid destructive mistakes: deleting a virtual disk is not the same as removing an empty drive letter, and a healthy file system cannot compensate for a failed underlying pool.

This guide creates a fixed-size two-copy mirrored virtual disk using two specifically approved blank disks on one server. It is not Storage Spaces Direct and does not create shared clustered storage. A mirror can tolerate the disk failure allowed by its layout, but it does not protect against deletion, ransomware, a failed server or loss of the whole storage location. Maintain independent backups and understand the enclosure, controller and power failure boundaries.

Choose supported blank disks and sufficient capacity

Use supported storage hardware that exposes individual disks correctly. RAID abstraction, hidden device identities and unsuitable controller settings can make a design unsupported or defeat its intended fault visibility. Microsoft documents additional restrictions for Storage Spaces over iSCSI or Fibre Channel; do not build this mirrored example on arbitrary SAN LUNs. In a VPS, confirm that the provider supports the required storage presentation and understand that two virtual disks may share the same physical failure domain.

A two-way mirror requires at least two eligible disks and stores two copies, so usable capacity is lower than raw capacity. Leave pool headroom for metadata, growth and repair operations. Do not size the virtual disk by simply adding the raw disk capacities. Three-copy mirroring and enclosure-aware designs have different disk-count and fault-domain requirements. Confirm the intended number of copies explicitly rather than assuming every object named Mirror has the same protection.

Only blank approved disks should be pooled. CanPool True identifies eligibility, not authorization to erase or consume a device. Record serial numbers and unique identifiers, verify them against the hardware or platform inventory, and retain an existing disk-layout record before changing storage.

Inventory and explicitly select the two devices

Use elevated Windows PowerShell and reliable console access. List every physical disk, including ones that are not poolable, so you understand the context. This example uses unique identifiers taken from the reviewed inventory. Replace both placeholder values and confirm that each resolves to exactly one eligible healthy disk before continuing.

Get-PhysicalDisk | Select-Object FriendlyName,SerialNumber,UniqueId,CanPool,HealthStatus,OperationalStatus,Size
$approvedIds = @('REPLACE-DISK-ONE-UNIQUEID','REPLACE-DISK-TWO-UNIQUEID')
$disks = @(Get-PhysicalDisk -CanPool $true | Where-Object { $_.UniqueId -in $approvedIds })
if ($disks.Count -ne 2) { throw 'The two approved eligible disks were not uniquely identified.' }
$disks | Format-Table FriendlyName,SerialNumber,UniqueId,Size,HealthStatus
Get-StorageSubSystem | Select-Object FriendlyName,UniqueId,HealthStatus

Review the printed identifiers manually. Do not replace the explicit filter with every CanPool device to bypass the safety check. If a disk unexpectedly cannot pool, identify its prior use and state; clearing it is a separate destructive operation requiring evidence that no needed data remains.

Create the pool and explicit two-copy mirror

Select the intended Windows storage subsystem using its actual identifier. A server may expose more than one subsystem, especially when external storage management is installed. The following resolves a single reviewed subsystem and creates a pool from only the selected devices. Existing names should be investigated before rerunning creation commands.

$subsystemId = 'REPLACE-WITH-APPROVED-SUBSYSTEM-UNIQUEID'
$sub = @(Get-StorageSubSystem | Where-Object UniqueId -eq $subsystemId)
if ($sub.Count -ne 1) { throw 'Select exactly one approved storage subsystem.' }
New-StoragePool -FriendlyName 'DataPool' `
  -StorageSubSystemUniqueId $sub[0].UniqueId -PhysicalDisks $disks
New-VirtualDisk -StoragePoolFriendlyName 'DataPool' `
  -FriendlyName 'DataMirror' -ResiliencySettingName Mirror `
  -NumberOfDataCopies 2 -ProvisioningType Fixed -Size 500GB

The 500 GB size is illustrative and must fit the reviewed resilient capacity. Fixed provisioning reserves capacity at creation, making the initial commitment more visible than thin allocation. It does not eliminate the need to monitor the pool. If creation fails for insufficient capacity or layout constraints, inspect the pool and supported sizes instead of reducing protection or selecting additional unapproved disks.

Initialize only the newly created virtual disk

Retrieve the Windows disk through the virtual-disk association, not through a guessed disk number. Confirm there is exactly one disk and that it is RAW. Review its size and identity before initializing it. If it already has a partition style, stop; the remaining commands are intended for a new empty virtual disk.

$disk = @(Get-VirtualDisk -FriendlyName 'DataMirror' | Get-Disk)
if ($disk.Count -ne 1 -or $disk[0].PartitionStyle -ne 'RAW') {
    throw 'Expected exactly one new RAW virtual disk.'
}
$disk[0] | Format-List Number,FriendlyName,UniqueId,PartitionStyle,Size
Initialize-Disk -Number $disk[0].Number -PartitionStyle GPT
if (Get-Volume -DriveLetter D -ErrorAction SilentlyContinue) {
    throw 'Drive letter D is already in use; choose an approved free letter.'
}
New-Partition -DiskNumber $disk[0].Number -UseMaximumSize -DriveLetter D |
  Format-Volume -FileSystem NTFS -NewFileSystemLabel 'Data'

Formatting creates a new file system and is destructive if applied to the wrong device. NTFS is chosen here for a general file-service example; ReFS may be appropriate for other supported workloads but has different feature compatibility. Agree on the file system, allocation-unit requirements and drive letter before executing the format.

Verify every layer and a real data workflow

Check the pool, virtual disk, member devices, Windows disk and formatted volume. Expect healthy states and the intended mirror properties, including two data copies. A volume appearing in Explorer is not enough: a degraded pool can still present a usable drive while lacking the expected redundancy.

Get-StoragePool -FriendlyName 'DataPool' | Format-List HealthStatus,OperationalStatus,Size,AllocatedSize
Get-VirtualDisk -FriendlyName 'DataMirror' | Format-List HealthStatus,OperationalStatus,ResiliencySettingName,NumberOfDataCopies,Size
Get-StoragePool -FriendlyName 'DataPool' | Get-PhysicalDisk
Get-Volume -DriveLetter D
Get-StorageJob

Write disposable test data, read it back and verify a backup and restore through the planned protection system. Establish monitoring for degraded health, free capacity, failed devices and repair jobs. If failure testing is required, use a dedicated lab or a vendor-supported controlled plan; do not pull a production disk merely because a mirror should survive one failure. Confirm that operators can reliably match a reported failed device to its physical slot or provider identifier.

Diagnose degraded state before changing membership

For a warning or degraded virtual disk, inspect physical disk health, controller and System events, connectivity and active storage jobs. A missing disk can be a cable, controller or path problem rather than a failed medium. Preserve serial-number evidence and follow the supported replacement procedure for the actual condition. Do not reset all disks, remove the pool or force-retire devices as a general repair.

If a repair cannot complete, verify available capacity and whether the layout has enough eligible devices to restore the intended copies. Adding capacity and expanding the virtual disk and file system are distinct operations. For poor performance, compare read and write latency, queueing, application load and controller behaviour; changing resiliency is not an in-place tuning shortcut. A mirror’s ability to remain online during failure is valuable time to recover protection, not permission to ignore degraded alerts.

Rollback preserves data before removing storage objects

If creation stops before a volume holds data, inventory the objects actually created and remove only those new objects through a reviewed cleanup. Once applications use D:, migration and backup come first. Remove application dependencies and shares, copy data to verified storage, validate restoration, and only then plan removal of partitions, virtual disk and pool in the appropriate order.

Do not include blanket disk-clearing commands in an ordinary rollback. Removing the pool destroys its storage arrangement and cannot be undone by recreating names. Keep the original disks and configuration evidence intact until migration acceptance. Record the selected devices, subsystem, resiliency, provisioning, volume layout and replacement process so future expansion or failure recovery continues from a documented design rather than assumptions about what the name DataMirror means.

Related guidance and official references

Present storage using iSCSI; Measure disk latency; Create a full server backup.


Was this answer helpful?

« Back