Windows Storage Spaces Says the Pool Failed. Don't Delete the Pool.

Written and maintained by the Desert Data Recovery lab team. · Updated September 16, 2026

We already wrote about Synology DSM's Volume Crashed / Repair, QNAP's Volume Not Active / Recover, the RAID controller's Rebuild button, Unraid Format and TrueNAS destroy-and-recreate, and Windows Initialize Disk. This is the Storage Spaces screen: a pool that will not come up, virtual disks Missing or Failed, and Delete Pool / Remove Disk / Reset sitting there like a diagnostic. The words are different. The write is the same class of mistake.

Physical disks, then a pool, then a virtual disk

A Storage Spaces volume is not one thing. The stack is: physical disks → a storage pool → a virtual disk (Simple, Mirror, or Parity) → an NTFS or ReFS volume. Control Panel (Storage Spaces), Server Manager (File and Storage Services → Storage Pools), and PowerShell (Get-StoragePool, Get-VirtualDisk, Get-PhysicalDisk) report statuses on those layers separately. They do not say which layer failed.

The resiliency type is the first fact that matters:

• Simple (no resiliency) is striping. One dead physical disk typically takes the virtual disk with it. People hear "Spaces" and assume Windows is protecting them. Simple means it is not. We already wrote that on RAID 0 — one drive dies. This post does not redo that page. If the virtual disk was Simple, start there.
• Two-way / Three-way Mirror keeps copies. One (or two) members can drop and the virtual disk can still be intact — if the remaining copies are healthy and the pool metadata still agrees.
• Parity / Dual parity is RAID-5 / RAID-6 shaped. One (or two) members can drop. A second weak disk, a column you removed, or a Repair you started without knowing why the pool failed is the usual overwrite.

This post is the Mirror or Parity pool Windows marked Failed, Offline, or Read-only — and the "delete the pool and start over" trap. Do not apply the Simple / RAID 0 playbook as a click-by-click script here. Redundancy may still exist on the disks. Delete Pool spends it.

The public case log does not have a named Storage Spaces job. We will not invent a case ID or an outcome. The closest published RAID write-ups are a Synology four-drive RAID 5 (DDR-2026-0003) and an HPE SmartArray RAID 6 + RAID 5 (21607). Those are mdadm/NAS and a hardware controller. Failed on Storage Spaces can also be pool metadata or an NTFS/ReFS volume with every member still present. Those are different failures that share one family of UI statuses.

Failed is not "the files were erased"

Healthy / Warning / Degraded (wording varies by Control Panel vs Server Manager vs PowerShell) can mean the pool is still there and a member is weak or missing. The virtual disk may still attach. That is the one situation where a replace-and-repair can be a normal maintenance step — if every remaining disk is healthy, the virtual disk is still assembled, and you are following Microsoft's procedure for a single known-dead member. If the volume already will not mount, you are not in a backup-then-repair job.

Read-only means Windows attached the pool or virtual disk and then refused writes. Shares or drive letters may still be visible. Do not take that as permission to Repair, Optimize, or chkdsk "so it goes read-write again." Read-only is often the last good state before someone writes.

Offline / Failed / Detached on the pool, and Missing / Failed / Incomplete on the virtual disk, mean the volume is not serving files. Storage Spaces will not assemble the pool, or it assembled and the virtual disk will not attach, or the NTFS/ReFS volume will not mount. Delete Pool here is not "check the disks and tell me what happened." It is Windows tearing down a layer it has not identified.

A physical disk marked Failed, Lost communication, or Retired is a Spaces membership flag. It is not a lab report that the platters or NAND are blank. Keep that disk. The member the UI retired may still hold a copy or a parity stripe the pool needs.

This is not Storage Spaces Direct (S2D) on a failover cluster. If you are looking at Cluster Shared Volumes and a clustered pool, stop even harder — do not treat it as a consumer Control Panel pool and do not Evict / Remove Disk to "test." The rule that transfers is the same: do not let Windows "fix" the only copy.

The buttons that make it worse

Delete Pool / Remove Storage Pool. In Control Panel this is Delete pool. In Server Manager it is Delete Storage Pool. In PowerShell it is Remove-StoragePool. Windows then treats the pool, the virtual disks, and the volumes as gone. The friendly copy says you can create a new pool on the same disks. That is a write to Spaces metadata. It is not a read-only report.

Remove Disk / Prepare for removal / Retire. Taking a member out of a Failed or Read-only pool writes pool metadata and can start a Repair you did not mean to start. The disk you removed is often still a required copy. Do not "drop the red one" so the pool looks happier.

Reset disk / Reset-PhysicalDisk. Reset wipes the Spaces metadata on that physical disk so Windows can reuse it as a primordial disk. That is the map a lab reads first. Same class of mistake as Initialize Disk on a member — and people often do both in one sitting (reset or initialize the Failed disk, then create a new pool).

Repair / Optimize / Rebalance when you do not know which layer failed. Repair-VirtualDisk reconstructs from mirror copies or parity. Optimize-StoragePool moves slabs. Both write. That is correct only when one member is dead, the rest are healthy, and the virtual disk is still a resiliency problem. On a Failed pool the break may be metadata or the filesystem. Repair then writes reconstructed (or wrong) data onto members that were not a missing-disk problem. Same overwrite we describe in RAID rebuild vs. data recovery.

Create a new pool on the same disks. A new storage pool stamps new Spaces metadata over the old pool. Do not "start clean" so the server has a drive letter again.

chkdsk /f, Repair-Volume, or Error checking on the virtual disk. That is a filesystem write to NTFS or ReFS. If the failure is the pool or the virtual disk, you just wrote to members that were not a volume problem. If it is ReFS, see the next section.

Initialize Disk on the members. Disk Management will offer Initialize on a physical disk it no longer sees as a Spaces member — or on every disk after someone already deleted the pool. Confirming it writes a new partition table. That is the Initialize Disk post. Do not do it to "see the drives as normal disks."

Yanking disks to "test in another PC." Storage Spaces cares about member identity and pool metadata on every disk. Pulling members to a spare desktop, swapping SATA ports, or letting a second Windows box Attach / Reset / Import the pool can change which disks Windows tries to repair. Label first. A disk goes back only in the slot or port it came from. Do not shuffle.

PowerShell that looks like a test. Remove-StoragePool, Reset-PhysicalDisk, Repair-VirtualDisk, Optimize-StoragePool, Set-StoragePool -IsReadOnly $false so you can write again — a prompt does not make those a diagnostic. If the data matters, that is a lab job on images.

ReFS vs NTFS — repair still writes

Storage Spaces volumes are often ReFS on Server, and NTFS on a lot of Windows 10/11 desktop pools. Neither filesystem is a free undo for a Failed pool.

ReFS integrity streams checksum file data. That is useful when a mirror still has a good copy: ReFS can prefer the copy that matches. It is not a backup of the pool. It does not reconstruct virtual-disk metadata you deleted. It does not roll back a Repair or an Optimize. And when ReFS is asked to repair — Error checking, Repair-Volume, chkdsk — it writes. Salvage can drop extents ReFS no longer trusts. Integrity streams are a detector, not an undelete.

NTFS has no integrity-stream story. chkdsk /f (or /r) on a virtual disk that is already Failed or Read-only is the usual Windows "make the letter come back" write. If the pool never attached, you may be running it against an incomplete virtual disk. If the pool did attach and the volume is dirty, you are still writing the only copy.

If the volume was BitLocker-encrypted — or Windows Device Encryption was on the machine — the key still matters after imaging. We cannot brute-force BitLocker. Bring the 48-digit recovery key if you have it. See Windows Device Encryption / BitLocker.

What to do right now

Do not delete the pool. Do not remove a disk. Do not reset, repair, optimize, initialize, or chkdsk to "see if it comes back." Do not create a new pool on the same disks.

1. Stop. Leave the machine alone. If a drive is clicking, grinding, or Storage Spaces / SMART errors are climbing, power it down. Extra uptime on a dying member is more heads-on-media you do not get back. See what a clicking drive means.
2. Photograph the UI. Control Panel Storage Spaces, Server Manager Storage Pools, or the PowerShell output for pool / virtual disk / physical disk: resiliency type (Simple / Mirror / Parity), statuses (Failed, Offline, Read-only, Missing), drive list with serials if shown, any exact error text. Screenshots help. They are not a diagnosis.
3. Label every disk before anything is unplugged. Slot, bay, or SATA/SAS port, matching the photos. Member identity is part of the configuration. If you already pulled a disk, write down which port it left.
4. Keep every member, including ones marked Failed, Lost communication, or Retired, the spare you already inserted, and any SSD that was in the pool next to HDDs. The disk you "replaced" may still be a required copy or parity member.
5. Write down what you remember: Simple vs Two-way / Three-way Mirror vs Parity / Dual parity; NTFS vs ReFS; column count if you ever saw it; BitLocker or Device Encryption keys if the volume was encrypted. Encryption without the key is still encryption.

If a disk is mechanical, pack it like any other failed drive. Do not open a drive to "check" it. Do not leave Repair or Optimize running overnight so the pool "finishes."

How a lab handles it

We do not delete your pool in Windows. We do not run Repair-VirtualDisk, Optimize-StoragePool, Reset-PhysicalDisk, or chkdsk against the patient disks.

Each member is imaged write-blocked, sector by sector, including disks that only read in pieces and disks Storage Spaces already marked Failed. The pool metadata and virtual-disk layout are reconstructed from those images — Simple, Mirror, or Parity — then the NTFS or ReFS volume is extracted offline. A wrong guess about member order, column count, or which slab map to walk costs a clone, not the original. That is the same method we use on hardware RAID, Synology, QNAP, Unraid, and TrueNAS. Imaging uses the lab tools already named on this site — including the ACE Lab PC-3000 and DeepSpar Disk Imager — then file work happens on the copies.

We will tell you if the data is not there. A Delete Pool that finished, a new pool on the same disks, or a disk whose platters are gone is not a puzzle we pretend we can solve. There is no Storage Spaces case in the public case log, and we will not invent one. We don't advertise a success rate. The free evaluation is how you find out which failure you have before anyone bills you.

If the virtual disk was Simple, start with RAID 0 — one drive dies. If the screen is Synology DSM rather than Storage Spaces, see don't click Repair. If it is QNAP, see don't click Recover. If it is Unraid or TrueNAS, see what not to click on those boxes. If it is a controller Rebuild, see RAID rebuild vs. data recovery. Disk Management offering Initialize on a member is Initialize Disk. For how these cases run in the lab, see our RAID data recovery service.