Technically: Controlling KubeVirt VMs from a pod
Logically: Ability to leverge Kubernetes workload controllers with KubeVirt VMs
Let a pod control a VM - not directly (as in creating a qemu process), but indirectly by talking to KubeVirt.
The vmctl pod works with Virtual Machines (previousl offline virtual
machines).
- The pod only acts as a "remote control" to start and stop virtual machine instances
- The pod can be simple
- All VM configurations are done on the corresponding virtual machine
- Deploy KubeVirt demo on minikube
- Deploy
deployment.yaml - Scale the deployment up
Follow this guide to setup the base demo:
https://github.com/kubevirt/demo
$ kubectl apply \
-f https://raw.githubusercontent.com/fabiand/vmctl/master/manifests/deployment.yaml
$ kubectl get deployments
NAME DESIRED CURRENT UP-TO-DATE AVAILABLE AGE
vmctl 0 0 0 0 3s
Now you are ready to scale the deployment, and indirectly the number of VMs:
$ kubectl scale --replicas=1 deployment/vmctl
deployment.extensions "vmctl" scaledThis is just like scaling any other deployment.
You can now check that the scaling really happened:
$ kubectl get deployments
NAME DESIRED CURRENT UP-TO-DATE AVAILABLE AGE
vmctl 1 1 1 1 3mAnd you can also look at the VM instances in to see that really VMs were spawned:
$ kubectl get pods,vms
NAME READY STATUS RESTARTS AGE
virt-launcher-testvm-j7f8k 2/2 Running 0 39m
vmctl-58ff778cc4-wskgs 1/1 Running 0 23s
NAME AGE
testvm 39m
testvm-vmctl-58ff778cc4-wskgs 20sScaling down just works as expected:
$ kubectl scale --replicas=0 deployment/vmctl
deployment.extensions "vmctl" scaled
$ kubectl get deployments
NAME DESIRED CURRENT UP-TO-DATE AVAILABLE AGE
vmctl 0 0 0 0 3h
$ kubectl get pods,vms
NAME READY STATUS RESTARTS AGE
virt-launcher-testvm-j7f8k 2/2 Running 0 41m
NAME AGE
testvm 41mAnd the instance is gone again.
Note: One VM will always be defined, as it act's as the prototype for the VMs any
vmctlpod is creating.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent