Fix ERROR_CLUSTER_PARTIAL_WRITE (0x1722) on Windows Server

1. Disk Timeout Between Nodes

The culprit here is almost always a disk timeout. When a cluster node writes to shared storage (like a CSV or SMB share) and doesn't get an acknowledgment within a few seconds, you get 0x1722. This happens most often with iSCSI volumes or older SAN arrays that can't keep up with the write load.

I've seen this exact error on Windows Server 2016 and 2019 clusters connecting to Dell EqualLogic or HP 3PAR arrays. The fix is dead simple: increase the disk timeout value.

reg add HKLM\System\CurrentControlSet\Services\Disk /v TimeOutValue /t REG_DWORD /d 120 /f

Set that to 120 seconds (default is 60). Reboot the node. If you can't reboot right away, you can set it on the active node, then move resources, and reboot each node during maintenance. Don't bother with changing the MPIO timeout — that rarely helps unless you're seeing MPIO-specific errors.

After the reboot, the error should stop. If it doesn't, move to step 2.

2. iSCSI Initiator or Target Misconfiguration

iSCSI is the second most common trigger. The initiator on your cluster nodes might have mismatched settings compared to the target. Specifically, check the CRC / digest settings. If the initiator uses header digest but the target doesn't support it (or vice versa), you'll get partial writes that look like 0x1722.

Here's what I do every time:

1. Open iSCSI Initiator on each node.
2. Go to Configuration tab.
3. Click Change next to Header Digest and Data Digest.
4. Set both to Disabled.
5. Repeat for all connections.

Now check the target side. If you're using a software target like Windows iSCSI Target Server, make sure the target's digest settings match. A mismatch here will cause intermittent write failures that only show up under load.

Also verify the iSCSI link speed. I've had cases where a 1 GbE connection was running at 100 Mbps because of a bad cable or switch port. Run the following on each node:

Get-NetAdapter | Where Status -eq Up | Select Name, LinkSpeed

If you see anything below 1 Gbps, that's your bottleneck. Replace cable, re-terminate, or swap switch ports. Don't waste time running iperf — just fix the physical link.

3. Storage Queue Depth Saturation

Less common but nasty when it hits. The storage queue depth gets maxed out because the SAN or NAS can't process writes fast enough. This happens with NVMe over Fabrics or high-speed iSCSI when you have too many concurrent writers on a CSV volume.

I saw this on a Hyper-V cluster with 8 nodes hammering a single CSV. The fix is to limit the per-node queue depth, not increase it. Counterintuitive, I know. But when the storage is overwhelmed, throttling it down prevents partial writes.

Set this on each node:

reg add HKLM\System\CurrentControlSet\Services\Disk\TimeOutValue /v QueueDepth /t REG_DWORD /d 16 /f

Default is 32 or 64 depending on the driver. Drop it to 16. Reboot again. If the error persists after this, you're probably hitting a storage controller firmware bug. Check with your vendor for a firmware update.

Quick Reference Table

If none of these work, check the event log for additional clues. Look for Event ID 5127 (disk error) or 1069 (cluster resource failure). Sometimes the error is a red herring — a failing NIC or a flaky HBA can cause partial writes too. Replace hardware if you've gone through all three steps and still see the error.