Change a VM and verify what took effect

Identify the VM, current cluster state and intended operation before making a change. A capacity request, a power action and a manifest inspection have different validation and permission paths.

Browser demo · No account · Simulated data

Start from the live resource

Creation submits a complete desired configuration. An existing-VM change starts from the current cluster object and constructs a specific mutation. An old database snapshot is not sufficient context when another process has changed the VM.

Check when the change becomes effective

CPU, memory and disk targets are evaluated against the modification context. Do not assume every change can take effect while the guest runs. The accepted design distinguishes desired-state changes from effects that still require a restart.

Separate power and inspection

Start, stop and restart use environment-aware approval policy. Manifest export is an administrator inspection capability in the current implementation. It is not a general user operation with the same approval path as a capacity request.

Walk through the task

  1. Identify the VM and environment

    Confirm its owning service, namespace and current state before deciding whether to change resources or perform a power operation.

  2. Review supported targets

    Inspect current resource values and the support matrix before submitting a capacity target and reason.

    Inspect VMModifyContext and VMModifyRequest
  3. Follow the relevant workflow

    For a power action, check the environment-based approval path. For a capacity request, review the exact change and any restart requirement.

    Read power-operation policy
  4. Inspect execution and escalate precisely

    Check live VM state after the decision. Keep the original cluster error if the operation fails; a platform administrator can inspect the live manifest.

    Read manifest authorization tests

Requirements and limits

  • Supported online changes depend on the live VM, KubeVirt capabilities and validation; no universal hotplug or zero-downtime guarantee is made.
  • The frontend demo cannot validate a real mutation or guest restart.
  • Current manifest export requires platform administrator access; VM read access alone is insufficient.

Documentation and implementation

Common questions

Does approval mean the guest has new resources?

No. Check execution and live state. An accepted desired-state change can still require a restart before its guest-visible effect.

Can every VM user export live YAML?

No. Current manifest export is restricted to platform administrators; handler tests cover that boundary.