Skip to content

Est. 2011ยทMicrosoft Partner ID 7033487ยทDelivery under 3 minยทSupport 7 days a week

Your vault is empty.

License Error Event ID 5120

Event ID 5120 and 5142: Cluster Shared Volumes drop into paused or offline

11 min read Updated October 5, 2026 Windows Server: RDS, Hyper-V & Clustering

Fix it now

Event ID 5120 means a Cluster Shared Volume has entered a paused state and all I/O is queued until a path is re-established; Event ID 5142 means the volume is no longer accessible from this node at all. The status code inside the event is the diagnosis, so read that before changing anything.

Run these on the affected node, in an elevated PowerShell session

Get-ClusterSharedVolume
Get-ClusterSharedVolumeState | Format-Table Node,VolumeName,StateInfo,VolumeFriendlyName
mpclaim -s -d
Get-ClusterLog -Destination C:\temp\ -TimeSpan 15 -UseLocalTime
  1. Open the 5120 and read the status it carries. Microsoft publishes about twenty of them with separate causes, and they point at genuinely different places.
  2. If the only status you have is STATUS_CLUSTER_CSV_AUTO_PAUSE_ERROR (c0130021), stop. Microsoft states that individual occurrences of that one are false positives and can safely be ignored.
  3. If it is STATUS_BAD_NETWORK_PATH (c00000be), turn on path verification: Set-MPIOSetting -NewPathVerificationState Enabled then Set-MPIOSetting -NewPathVerificationPeriod 30
  4. For every status, Microsoft’s first three steps are the same: current network and storage drivers, current firmware, and all applicable updates and hotfixes applied.
  5. If the node shows redirected rather than direct access, compare its path count against a healthy node before you touch the cluster.

STATUS_CONNECTION_RESET (c000020d) is the one that points at your antivirus exclusions rather than at the fabric. It is worth knowing because the symptom is indistinguishable from a network fault.

If the volume is back in direct access and the virtual machines have resumed, you can stop here. The next section explains what CSV does when a node loses its disk, and why that hides some faults for weeks.

Why it happens

A Cluster Shared Volume is a volume every node mounts at once under a numbered folder in the ClusterStorage path on the system drive, with the same path on every node. One node owns it as coordinator and handles metadata; the others write straight to the disk over their own storage paths. That is direct I/O, and it is what makes CSV fast.

When a node loses its own path, CSV does not fail. It redirects that node’s I/O across a cluster network to the coordinator. Microsoft distinguishes two kinds: file system redirection, which is per volume, and block redirection at the file-block level, which is what happens on storage connectivity loss and is the faster of the two. The transport is SMB 3.0, which is why the cluster network adapters must keep Client for Microsoft Networks and File and Printer Sharing for Microsoft Networks enabled, and why the Server and Workstation services must be running and set to start automatically on every node.

Event 5121 is the warning that this has happened: the volume is no longer directly accessible from this node and I/O is being redirected over the network to the node that owns it. It often arrives weeks before the 5120 that anyone notices, and it is the single most useful event on this page.

A storage path has dropped and the node is quietly redirected

You have this one if 5121 on one node first, then 5120 under load. Other nodes are unaffected.

  1. Compare mpclaim -s -d against a healthy node and count paths per disk.
  2. Check the HBA or NIC for link errors and the switch port for CRC or discard counters.
  3. Enable path verification with Set-MPIOSetting -NewPathVerificationState Enabled and set the period to the vendor’s value.
  4. Once paths are back, move the coordinator role or take the CSV offline and online in a window so the state is re-evaluated.

SMB redirection has been hardened away, so there is no fallback

You have this one if 5120 or 5142 straight away with no 5121 first, and the node fails the instant its direct path goes.

  1. On the cluster network adapters, confirm Client for Microsoft Networks and File and Printer Sharing for Microsoft Networks are both enabled.
  2. Confirm the Server and Workstation services are running and set to start automatically on every node.
  3. Test node to node: Test-NetConnection <other-node> -Port 445
  4. Confirm the cluster networks are still allowed to carry cluster communication in Failover Cluster Manager.

A 5120 carrying STATUS_MEDIA_WRITE_PROTECTED (c00000a2) is Microsoft’s documented signal that SMB is not enabled at all.

A filter driver in the stack is forcing redirection

You have this one if 5124 or 5125 naming drivers, usually after an antivirus or backup agent upgrade.

  1. Read which one you have: 5124 is file system filter drivers, 5125 is volume drivers. Both force I/O over the network and both name the drivers in the event.
  2. List what is attached with fltmc filters and fltmc instances on that node.
  3. Take the driver names to the vendor and ask about interoperability with Cluster Shared Volumes, which is what the event text itself instructs.
  4. Reboot after removing or updating a filter driver, because filters attach at mount time.

The volume is redirected by design and nobody realised

You have this one if One CSV is permanently in redirected access, everything is healthy, and no path is missing.

  1. Check how the volume is formatted. A CSV pre-formatted with ReFS on SAN-attached storage will not use direct I/O regardless of everything else being correct.
  2. Format SAN volumes NTFS before adding them to CSV if you want direct I/O; keep ReFS for Storage Spaces Direct, where it buys data integrity and block cloning.
  3. Check whether redirected access was turned on deliberately for the volume, which the cluster records separately from a failure.

Full reference

The status codes inside a 5120, and where each sends you

This is the part of the event most people skip. Microsoft publishes the list, and the codes are not interchangeable.

Status Documented cause
STATUS_CLUSTER_CSV_AUTO_PAUSE_ERROR (c0130021) A false positive on its own. Investigate only if other codes appear alongside it
STATUS_BAD_NETWORK_PATH (c00000be) Storage configuration or MPIO settings; enable MPIO path verification
STATUS_CONNECTION_RESET (c000020d) Network, HBA or disk problems – or antivirus exclusions
STATUS_IO_TIMEOUT (c00000b5) A redirected file system I/O exceeded its timeout (two minutes synchronous, four asynchronous)
STATUS_MEDIA_WRITE_PROTECTED (c00000a2) SMB is not enabled
STATUS_DEVICE_BUSY (80000011) The persistent reservation on the volume was lost
STATUS_NO_SUCH_DEVICE (c000000e) Connectivity between the node and the CSV was lost; check network and firewall cluster exclusions
STATUS_BAD_IMPERSONATION_LEVEL (c00000a5) An operation was performed on the coordinator node without administrator permissions

The other events in the family

Event Published meaning
5120 The volume has entered a paused state because of the status shown; all I/O is queued until a path is re-established
5121 The volume is no longer directly accessible from this node; I/O is redirected over the network to the owning node
5124 Active file system filter drivers were found that could interfere with CSV; I/O is redirected and performance may degrade
5125 Active volume drivers were found that could interfere with CSV; same consequence
5136 Redirected access was turned on for the volume, so every node’s access to it is redirected
5142 The volume is no longer accessible from this node because of the error shown; troubleshoot storage and network connectivity
5150 A cluster physical disk resource failed and the CSV was put into a failed state with the error shown

Direct I/O, and what quietly removes it

Three things decide whether a node uses its own storage path. Whether it has one, which MPIO answers. Whether the CSV stack is clean, which 5124 and 5125 answer. And whether the file system supports it at all: a CSV formatted ReFS on top of a SAN operates in redirected I/O mode permanently, whatever else is configured. Format SAN-attached volumes NTFS before converting them if direct I/O matters; ReFS belongs on Storage Spaces Direct, where the integrity and block-cloning behaviour is the point.

Why one node’s storage fault looks like a cluster outage

Redirection is per node, but the coordinator is not. Every metadata operation on a volume goes through the node that owns it, so a single CSV holding every virtual machine concentrates all of that on one node, and a redirected node pushes all its data traffic over the cluster network on top. That combination turns a fault that affects one host into slow virtual machines everywhere. More, smaller CSVs spread both the metadata and the blast radius.

When the common causes are all ruled out

  • Check for Event ID 1135 in the same window. A node being removed from cluster membership produces CSV events as a consequence, and chasing the storage then is wasted effort.
  • Collect the cluster log from both affected nodes for a narrow window rather than one node for a wide one.
  • Confirm the coordinator has not moved repeatedly during the incident; repeated ownership changes are a symptom of something upstream, not of the volume.
  • Check the physical disk resource itself. A 5150 says the disk resource failed and took the CSV with it, which is a different investigation from a path loss.

When a licence is the actual fix

Nothing here is fixed by buying anything, and any suggestion that it is should be treated with suspicion. There is one licensing question worth settling while the cluster is open in front of you. If the storage under these volumes is Storage Spaces Direct rather than a SAN, that feature exists only in the Datacenter edition and every node in the cluster needs it. The same edition choice decides virtual machine density: Windows Server Standard covers two virtual machines plus one Hyper-V host per licence, Datacenter covers unlimited virtual machines plus one Hyper-V host per licence. Arco supplies Windows Server 2025 Datacenter licensing and can work out the core count per node before you order.

Every code this article covers

Code What it points at Source
Event ID 5120 The Cluster Shared Volume entered a paused state because of the status shown in the event; all I/O is queued until a path to the volume is re-established Microsoft Learn
Event ID 5142 The Cluster Shared Volume is no longer accessible from this node because of the error shown; the guidance is to troubleshoot this node’s storage and network connectivity Microsoft Learn
Event ID 5121 The volume is no longer directly accessible from this node, so I/O is redirected over the network to the node that owns the volume Microsoft Learn
Event ID 5125 Active volume drivers were found on the device stack that could interfere with CSV, so I/O is redirected over the network; the file system filter equivalent is event 5124 Microsoft Learn
Event ID 5150 A cluster physical disk resource failed and the Cluster Shared Volume was put into a failed state with the error shown Microsoft Learn

Confirm the fix worked

  1. Get-ClusterSharedVolumeState reports direct access on every node, not redirected.
  2. mpclaim -s -d shows the full expected path count on the node that failed.
  3. A live migration to the previously failing node and back completes without a pause.
  4. No new 5120, 5121 or 5142 appears in the FailoverClustering log through a full backup window.
  5. If a 5120 does recur, it carries only STATUS_CLUSTER_CSV_AUTO_PAUSE_ERROR and nothing else.

Questions people ask about this

Do the virtual machines lose data when a 5120 fires?

Not normally. The published behaviour is that I/O is queued rather than failed, and a paused virtual machine usually resumes when a path returns. The risk arrives when the volume goes fully inaccessible mid-write, so check the volume for NTFS corruption events afterwards.

Is redirected access a fault?

No, it is a designed fallback, and on ReFS over a SAN it is permanent by design. It is still worth investigating when it is unexpected, because every byte that node writes then crosses the cluster network.

Should I reboot the node that reported the 5120?

Not first. A reboot clears transient path problems and destroys the evidence. Collect the cluster log, read the status code in the event and check path counts before restarting anything.

Can I ignore a 5120?

Only one: Microsoft states that a 5120 carrying STATUS_CLUSTER_CSV_AUTO_PAUSE_ERROR alone is a false positive. Any other status, or that status with company, is worth reading.

Would a different Windows Server edition make CSV more reliable?

No. CSV behaves identically in Standard and Datacenter. Edition decides virtualisation rights and whether Storage Spaces Direct is available, not how the volume handles a lost path.

Related error codes

Was this article helpful?

Your feedback helps us improve our documentation.

Related articles

Free Fix HTTP 403.14 and 404.3: IIS will not serve the files that are clearly there Free Fix HTTP 401.1, 401.2 and 401.3: IIS keeps rejecting perfectly valid logons License Error Event ID 5002 and 5011: the WsusPool application pool keeps dying Free Fix Event ID 3040 and 3080: Hyper-V cannot open the VHDX and the VM stays off
โ† Back to Knowledge Base