Forum Discussion

Krisztian's avatar
Krisztian
Copper Contributor
Nov 18, 2026

Hyper-V recovery: is guest execution-state rollback excluded when checkpoints are disabled?

Hello everyone,

I am seeking authoritative technical clarification regarding Hyper-V recovery behavior, specifically whether automatic recovery following a VMMS or VM worker process failure can ever restore an earlier guest execution state.

Environment and security requirement

The target environment is Windows 11 Pro with a Generation 2 Hyper-V virtual machine.

Our security architecture requires that previously executed operations cannot be unintentionally replayed following a recovery event.

We are considering a configuration in which:

  • Standard and production checkpoints are not used.
  • Automatic checkpoints are disabled.
  • Intentional saved-state creation and restoration are prohibited.
  • Execution authorization is maintained outside the virtual machine's recovery boundary.

The remaining question concerns unexpected termination and subsequent recovery of VMMS or vmwp.exe.

Specific technical questions

  1. Can any supported automatic Hyper-V recovery mechanism restore an earlier guest memory or CPU instruction-execution state under these conditions?
  2. Is such restoration explicitly excluded by the Hyper-V implementation or documented product guarantees, including exceptional failure and recovery paths?
  3. Are there relevant differences between VMMS service recovery, VM worker process recovery, host recovery, and virtual machine restart?
  4. Is there official Microsoft engineering documentation, implementation guidance, or testing evidence establishing the applicable behavior for specific Windows and Hyper-V versions?
  5. If a formal guarantee cannot be provided, what additional architectural safeguards are recommended for systems that must prevent the replay of previously executed operations?

Previous Microsoft Support investigation

This question was investigated with Microsoft Support, including a Level 2 escalation, on November 18, 2026.

Support case: #7111939474

Microsoft Support explicitly stated that it could not provide a formal guarantee excluding restoration of earlier guest execution state across all relevant failure and recovery scenarios.

The support team was also unaware of any official Microsoft engineering statement or publicly documented product guarantee establishing this exclusion as a supported security property.

Importantly, this does not mean that such restoration has been observed or demonstrated. It means that its exclusion could not be formally guaranteed through the available support documentation.

Microsoft Support recommended raising the question in the Microsoft Tech Community to seek further input from Hyper-V subject matter experts.

What we are looking for

We are not reporting an observed Hyper-V failure and are not asking how to restore a checkpoint.

We are trying to distinguish expected recovery behavior from a security guarantee that can safely be relied upon in system architecture.

A technically precise explanation, official documentation reference, or clarification from a Microsoft Hyper-V engineer would be extremely valuable.

Although our immediate environment is Windows 11 Pro, we would also welcome clarification of any relevant differences in Windows Server Hyper-V.

Thank you for your time and expertise.

2 Replies

  • msexpert44's avatar
    msexpert44
    Tin Contributor

    Disabling checkpoints does not, by itself, establish a formal guarantee against execution-state rollback.

     

    The important distinctions are:

     

    - VMMS failure: Service recovery does not inherently mean restoring an earlier guest memory state.

    - vmwp.exe failure: The VM worker process manages VM execution. Its failure may result in a VM crash or restart rather than transparent continuation.

    - Host failure or reboot: Recovery behavior depends on the VM's saved-state and startup configuration.

    - Checkpoints disabled: This prevents checkpoint-based rollback, but does not establish a universal guarantee covering every recovery mechanism.

     

    My recommendation for your security architecture:

     

    1. Do not treat disabling checkpoints as a security guarantee against replay.

    2. Maintain operation authorization and transaction state outside the VM.

    3. Use externally maintained monotonic counters, unique operation IDs and durable transaction records.

    4. Require external authorization before potentially replayable operations are executed again.

     

    The key conclusion: If preventing replay is a security requirement, design for replay resistance independently of Hyper-V's recovery behavior.

     

    Microsoft would need to provide an explicit, version-specific engineering guarantee before Hyper-V could safely be treated as enforcing this property.

  • Your security concern is valid. The key distinction is between restarting a VM and restoring its previous execution state.

     

    - VMMS failure: VMMS service recovery does not inherently imply restoration of an earlier guest memory state.

    - vmwp.exe failure: The VM worker process manages VM execution. Its termination can cause the VM to stop or enter a failed state.

    - Host restart: A VM may restart from its virtual disks, but that is different from restoring its previous CPU and RAM state.

    - Checkpoints disabled: Removes checkpoint-based rollback, but does not establish a universal guarantee covering every recovery scenario.

     

    Security recommendation

     

    If replay prevention is a strict security requirement:

     

    1. Keep checkpoints and automatic checkpoints disabled.

    2. Prohibit saved-state restoration.

    3. Maintain execution authorization and transaction state outside the VM.

    4. Use external, durable transaction IDs and anti-replay counters.

    5. Require fresh authorization after unexpected VM or host termination.

     

    Most importantly: Given Microsoft's Support response, do not treat Hyper-V as providing a formally guaranteed anti-replay property across all failure scenarios.

     

    You need an architectural safeguard outside Hyper-V rather than relying solely on its recovery behavior.