The KubeVirt project provides extensions to Kubernetes via custom resources.

These resources are a collection a API objects that defines a virtual machine within Kubernetes.

I think it’s important to point out the two great resources that I used to compile information for this post:

• user-guide
• api-reference

With that let’s take a look at the objects that are available.

KubeVirt top-level objects

Below is a list of the top level API objects and descriptions that KubeVirt provides.

• VirtualMachine (vm[s]) - represents a virtual machine in the runtime environment of Kubernetes.

• OfflineVirtualMachine (ovm[s]) - handles the virtual machines that are not running or are in a stopped state.

• VirtualMachinePreset (vmpreset[s]) - is an extension to general VirtualMachine configuration behaving much like PodPresets from Kubernetes. When a VirtualMachine is created, any applicable VirtualMachinePresets will be applied to the existing spec for the VirtualMachine. This allows for re-use of common settings that should apply to multiple VirtualMachines.

• VirtualMachineReplicaSet (vmrs[s]) - tries to ensures that a specified number of VirtualMachine replicas are running at any time.

DomainSpec is listed as a top-level object but is only used within all of the objects above. Currently the DomainSpec is a subset of what is configurable via libvirt domain XML.

VirtualMachine

VirtualMachine is mortal object just like a Pod within Kubernetes. It only runs once and cannot be resurrected. This might seem problematic especially to an administrator coming from a traditional virtualization background. Fortunately later we will discuss OfflineVirtualMachines which will address this.

First let’s use kubectl to retrieve a list of VirtualMachine objects.

$ kubectl get vms -n nodejs-ex
NAME      AGE
mongodb   5d
nodejs    5d

We can also use kubectl describe

$ kubectl describe vms -n test
Name:         testvm
Namespace:    test
Labels:       guest=testvm
              kubevirt.io/nodeName=kn2.virtomation.com
              kubevirt.io/size=small
...output...
Events:
  Type    Reason              Age                From                               Message
  ----    ------              ----               ----                               -------
  Normal  SuccessfulCreate    59m                virtualmachine-controller          Created virtual machine pod virt-launcher-testvm-8h927
  Normal  SuccessfulHandOver  59m                virtualmachine-controller          Pod owner ship transfered to the node virt-launcher-testvm-8h927
  Normal  Created             59m (x2 over 59m)  virt-handler, kn2.virtomation.com  VM defined.
  Normal  Started             59m                virt-handler, kn2.virtomation.com  VM started.

And just in case if you want to return the yaml definition of a VirtualMachine object here is an example.

$ kubectl -o yaml get vms mongodb -n nodejs-ex
apiVersion: kubevirt.io/v1alpha1
kind: VirtualMachine
...output...

The first object we will annotate is VirtualMachine. The important sections .spec for VirtualMachineSpec and .spec.domain for DomainSpec will be annotated only in this section then referred to in the other object sections.

apiVersion: kubevirt.io/v1alpha1
kind: VirtualMachine
metadata:
  annotations: {}
  labels: {}
  name: string
  namespace: string
spec: {}

Node Placement

Kubernetes has the ability to schedule a pod to specific nodes based on affinity and anti-affinity rules.

Node affinity is also possible with KubeVirt. To constrain a virtual machine to run on a node define a matching expressions using node labels.

affinity:
  nodeAffinity:
    preferredDuringSchedulingIgnoredDuringExecution:
      - preference:
          matchExpressions:
            - key: string
              operator: string
              values:
                - string
        weight: 0
    requiredDuringSchedulingIgnoredDuringExecution:
      nodeSelectorTerms:
        - matchExpressions:
            - key: string
              operator: string
              values:
                - string

A virtual machine can also more easily be constrained by using nodeSelector which is defined by node’s label and value. Here is an example

nodeSelector:
  kubernetes.io/hostname: kn1.virtomation.com

Clocks and Timers

Configures the virtualize hardware clock provided by QEMU.

domain:
  clock:
    timezone: string
    utc:
      offsetSeconds: 0

The timer defines the type and policy attribute that determines what action is take when QEMU misses a deadline for injecting a tick to the guest.

domain:
  clock:
    timer:
      hpet:
        present: true
        tickPolicy: string
      hyperv:
        present: true
      kvm:
        present: true
      pit:
        present: true
        tickPolicy: string
      rtc:
        present: true
        tickPolicy: string
        track: string

CPU and Memory

The number of CPU cores a virtual machine will be assigned. .spec.domain.cpu.cores will not be used for scheduling use .spec.domain.resources.requests.cpu instead.

cpu:
  cores: 1

There are two supported resource limits and requests: cpu and memory. A .spec.domain.resources.requests.memory should be defined to determine the allocation of memory provided to the virtual machine. These values will be used to in scheduling decisions.

resources:
  limits: {}
  requests: {}

Watchdog Devices

.spec.domain.watchdog automatically triggers an action via Libvirt and QEMU when the virtual machine operating system hangs or crashes.

watchdog:
  i6300esb:
    action: string
  name: string

Features

.spec.domain.features are hypervisor cpu or machine features that can be enabled. After reviewing both Linux and Microsoft QEMU virtual machines managed by Libvirt both acpi and apic should be enabled. The hyperv features should be enabled only for Windows-based virtual machines. For additional information regarding features please visit the virtual hardware configuration in the kubevirt user guide.

features:
  acpi:
    enabled: true
  apic:
    enabled: true
    endOfInterrupt: true
  hyperv:
    relaxed:
      enabled: true
    reset:
      enabled: true
    runtime:
      enabled: true
    spinlocks:
      enabled: true
      spinlocks: 0
    synic:
      enabled: true
    synictimer:
      enabled: true
    vapic:
      enabled: true
    vendorid:
      enabled: true
      vendorid: string
    vpindex:
      enabled: true

QEMU Machine Type

.spec.domain.machine.type is the emulated machine architecture provided by QEMU.

machine:
  type: string

Here is an example how to retrieve the supported QEMU machine types.

    $ qemu-system-x86_64 --machine help
    Supported machines are:
    ...output...
    pc                   Standard PC (i440FX + PIIX, 1996) (alias of pc-i440fx-2.10)
    pc-i440fx-2.10       Standard PC (i440FX + PIIX, 1996) (default)
    ...output...
    q35                  Standard PC (Q35 + ICH9, 2009) (alias of pc-q35-2.10)
    pc-q35-2.10          Standard PC (Q35 + ICH9, 2009)

Disks and Volumes

.spec.domain.devices.disks configures a QEMU type of disk to the virtual machine and assigns a specific volume and its type to that disk via the volumeName.

devices:
  disks:
    - cdrom:
        bus: string
        readonly: true
        tray: string
      disk:
        bus: string
        readonly: true
      floppy:
        readonly: true
        tray: string
      lun:
        bus: string
        readonly: true
      name: string
      volumeName: string

cloudInitNoCloud injects scripts and configuration into a virtual machine operating system. There are three different parameters that can be used to provide the cloud-init coniguration: secretRef, userData or userDataBase64.

See the user-guide for examples of how to use .spec.volumes.cloudInitNoCloud.

volumes:
  - cloudInitNoCloud:
      secretRef:
        name: string
      userData: string
      userDataBase64: string

An emptyDisk volume creates an extra qcow2 disk that is created with the virtual machine. It will be removed if the VirtualMachine object is deleted.

emptyDisk:
  capacity: string

Ephemeral volume creates a temporary local copy on write image storage that will be discarded when the VirtualMachine is removed.

ephemeral:
  persistentVolumeClaim:
    claimName: string
    readOnly: true
name: string

persistentVolumeClaim volume persists after the VirtualMachine is deleted.

persistentVolumeClaim:
  claimName: string
  readOnly: true

registryDisk volume type uses a virtual machine disk that is stored in a container image registry.

registryDisk:
  image: string
  imagePullSecret: string

Virtual Machine Status

Once the VirtualMachine object has been created the VirtualMachineStatus will be available. VirtualMachineStatus can be used in automation tools such as Ansible to confirm running state, determine where a VirtualMachine is running via nodeName or the ipAddress of the virtual machine operating system.

kubectl -o yaml get vm mongodb -n nodejs-ex

# ...output...
status:
  interfaces:
  - ipAddress: 10.244.2.7
  nodeName: kn2.virtomation.com
  phase: Running

Example using --template to retrieve the .status.phase of the VirtualMachine.

kubectl get vm mongodb --template {{.status.phase}} -n nodejs-ex
Running

Examples

• https://kubevirt.io/user-guide/virtual_machines/virtual_machine_instances/#virtualmachineinstance-api

OfflineVirtualMachine

An OfflineVirtualMachine is an immortal object within KubeVirt. The VirtualMachine described within the spec will be recreated with a start power operation, host issue or simply a accidental deletion of the VirtualMachine object. For a traditional virtual administrator this object might be appropriate for most use-cases.

Just like VirtualMachine we can retrieve the OfflineVirtualMachine objects.

$ kubectl get ovms -n nodejs-ex
NAME      AGE
mongodb   5d
nodejs    5d

And display the object in yaml.

$ kubectl -o yaml get ovms mongodb -n nodejs-ex
apiVersion: kubevirt.io/v1alpha1
kind: OfflineVirtualMachine
metadata:
...output...

We continue by annotating OfflineVirtualMachine object.

apiVersion: kubevirt.io/v1alpha1
kind: OfflineVirtualMachine
metadata:
  annotations: {}
  labels: {}
  name: string
  namespace: string
spec:

What is Running in OfflineVirtualMachine?

.spec.runStrategy controls whether and when the associated VirtualMachineInstance object is created. In other words this controls the power status of the virtual machine.

  runStrategy: Always

This will create a VirtualMachineInstance object which will instantiate and power on a virtual machine.

kubectl patch offlinevirtualmachine mongodb --type merge -p '{"spec":{"runStrategy": "Always"}}' -n nodejs-ex

This will delete the VirtualMachineInstance object which will power off the virtual machine.

kubectl patch offlinevirtualmachine mongodb --type merge -p '{"spec":{"runStrategy": "Halted"}}' -n nodejs-ex

And if you would rather not have to remember the kubectl patch command above the KubeVirt team has provided a cli tool virtctl that can start and stop a guest.

./virtctl start mongodb -n nodejs-ex
./virtctl stop mongodb -n nodejs-ex

Offline Virtual Machine Status

Once the OfflineVirtualMachine object has been created the OfflineVirtualMachineStatus will be available. Like VirtualMachineStatus OfflineVirtualMachineStatus can be used for automation tools such as Ansible.

kubectl -o yaml get ovms mongodb -n nodejs-ex

# ...output...
status:
  created: true
  ready: true

Example using --template to retrieve the .status.conditions[0].type of OfflineVirtualMachine.

kubectl get ovm mongodb --template "{{.status.ready}}" -n nodejs-ex
true

VirtualMachineReplicaSet

VirtualMachineReplicaSet is great when you want to run multiple identical virtual machines.

Just like the other top-level objects we can retrieve VirtualMachineReplicaSet.

$ kubectl get vmrs -n nodejs-ex
NAME      AGE
replica   1m

With the replicas parameter set to 2 the command below displays the two VirtualMachine objects that were created.

$ kubectl get vms -n nodejs-ex
NAME           AGE
replicanmgjl   7m
replicarjhdz   7m

Pause rollout

The .spec.paused parameter if true pauses the deployment of the VirtualMachineReplicaSet.

  paused: true

Replica quantity

The .spec.replicas number of VirtualMachine objects that should be created.

  replicas: 0

The selector must be defined and match labels defined in the template. It is used by the controller to keep track of managed virtual machines.

selector:
  matchExpressions:
    - key: string
      operator: string
      values:
        - string
  matchLabels: {}

Virtual Machine Template Spec

The VMTemplateSpec is the definition of a VirtualMachine objects that will be created.

In the VirtualMachine section the .spec VirtualMachineSpec describes the available parameters for that object.

template:
  metadata:
    annotations: {}
    labels: {}
    name: string
    namespace: string
  spec: {}

Replica Status

Like the other objects we already have discussed VMReplicaSetStatus is an important object to use for automation.

status:
  readyReplicas: 0
  replicas: 0

Example using --template to retrieve the .status.readyReplicas and .status.replicas of VirtualMachineReplicaSet.

$ kubectl get vmrs replica --template "{{.status.readyReplicas}}" -n nodejs-ex
2
$ kubectl get vmrs replica --template "{{.status.replicas}}" -n nodejs-ex
2

Examples

• https://kubevirt.io/user-guide/virtual_machines/replicaset/#example

VirtualMachinePreset

This is used to define a DomainSpec that can be used for multiple virtual machines.

To configure a DomainSpec for multiple VirtualMachine objects the selector defines which VirtualMachine the VirtualMachinePreset should be applied to.

$ kubectl get vmpreset -n nodejs-ex
NAME       AGE
m1.small   17s

Domain Spec

See the VirtualMachine section above for annotated details of the DomainSpec object.

spec:
  domain: {}

Preset Selector

The selector is optional but if not defined will be applied to all VirtualMachine objects; which is probably not the intended purpose so I recommend always including a selector.

selector:
  matchExpressions:
    - key: string
      operator: string
      values:
        - string
  matchLabels: {}

Examples

• https://kubevirt.io/user-guide/virtual_machines/presets/#examples

We provided an annotated view into the KubeVirt objects - VirtualMachine, OfflineVirtualMachine, VirtualMachineReplicaSet and VirtualMachinePreset. Hopefully this will help a user of KubeVirt to understand the options and parameters that are currently available when creating a virtual machine on Kubernetes.