Based on investigations between Veeam and Microsoft, two underlying causes have been identified.
Resilient Change Tracking
Resilient Change Tracking monitors data changes on VHDX disks. However, depending on the specific cluster configuration, such as the disk-to-cluster-volume ratio and the available performance, an I/O bottleneck may occur.
This underlying issue was resolved in the following ways:
- Server 2025 — Included out-of-the-box.
- Server 2022 — Included in the October 2023 Cumulative Update (KB5031364)
- Server 2019 — Included in the October 2023 Cumulative Update (KB5031361)
- Azure Stack HCI 22H2 - Included in the October 2023 Cumulative Update (KB5031364)
- Azure Stack HCI 21H2 – Included in the October 2023 Cumulative Update (KB5031364)
I/O Thread Handling
The I/O profile for backup operations utilizes buffered I/O, whereas the rest of the operating system employs unbuffered I/O.
- Server 2025 — Microsoft has advised Veeam that Server 2025 is not affected by this
- Server 2022 and Azure Stack HCI 22H2 — Microsoft included a fix to improve how I/O threads are handled in the February 11, 2025 Cumulative Update. Note, however, that although the fix is included, it is disabled by default and must be enabled on each Hyper-V host using one of the following methods:
- Create the following registry value:
Key Location: HKLM\SYSTEM\CurrentControlSet\Policies\Microsoft\FeatureManagement\Overrides
Value Name: 636159629
Value Type: DWORD (32-Bit) Value
Value Data: 1
- Install the fix enablement policy (msi) available from Microsoft:
Windows Server 2022 KB5051979 250131_08457 Feature Preview.msi
Enabling the fix using either method before installing the Cumulative Update will ensure that only one reboot is necessary. Typically, a reboot is required after installing the Cumulative Update and again after enabling the fix. By enabling the fix beforehand, it will remain dormant but ready to activate once the update is installed, eliminating the need for an additional reboot.