How to manage the VM in older version Kubernetes

[dyoung23]

Hi @fengye87 .I am using virtink v0.13.0 and Kubernetes v1.20.9 which have no --subresource option.
So how I can use it to manage the VM because I can't use kubectl patch vm $VM_NAME --subresource=status --type=merge -p "{\"status\":{\"powerAction\":\"$POWER_ACTION\"}}"

[dyoung23]

And how can I access a VM from a node out of Kubernetes cluster?

[fengye87]

Hi @dyoung23 , could you try with newer versions of kubectl, for example v1.21.x? There's a chance that this is a limitation of v1.20.x's kubectl, but the API is OK. But that's just my guess. Otherwise, I'm afraid there's no easy way to do this for now.

To access a VM outside the Kubernetes cluster, it's pretty much same as accessing a Pod. You can forward SSH port using kubectl, or using NodePort or LB service to expose SSH port.

[dyoung23]

@fengye87 Thanks for you reply. I updated my Kubernetes to v1.24.0,so I can manage the vm like Readme and I am looking if the API can manage the VM.
But I try to manage the VM using kubectl patch ... and find that when I Shutdown or Reset the vm,nothing happend,I can still ssh to the vm just get some new events in virtualmachine CRD like that.

Events:
  Type    Reason        Age                  From             Message
  ----    ------        ----                 ----             -------
  Normal  CreatedVMPod  25m                  virt-controller  Created VM Pod "vm-ubuntu-container-rootfs-9rmwk"
  Normal  Paused        8m43s                virt-daemon      Paused VM
  Normal  Resumed       7m53s                virt-daemon      Resumed VM
  Normal  Rebooted      7m9s                 virt-daemon      Rebooted VM
  Normal  PoweredOff    5m32s                virt-daemon      Powered off VM
  Normal  DeletedVMPod  5m32s                virt-controller  Deleted VM Pod "vm-ubuntu-container-rootfs-9rmwk"
  Normal  CreatedVMPod  5m18s                virt-controller  Created VM Pod "vm-ubuntu-container-rootfs-jbpmd"
  Normal  Shutdown      67s (x2 over 2m18s)  virt-daemon      Shutdown VM
  Normal  Reset         14s                  virt-daemon      Reset VM

[fengye87]

@dyoung23 Shutdown is ACPI shutdown, it requires OS's support to work. Reset should work, are you sure nothing happened?

[dyoung23]

@fengye87 I create a file in my vm and Reset.After that I can still ssh to this vm and see the file.

[fengye87]

@dyoung23 Reset is like force reboot, it will not reset VM's hard disk.

[dyoung23]

@fengye87 I see.Maybe it be reset really fast,so I can ssh to the vm all the time.

[fengye87]

@dyoung23 If you reset the VM while SSH logged in, it should log you out. Otherwise, there's something wrong with the reset action.

[dyoung23]

@fengye87 Yes,It really log out

[fengye87]

@dyoung23 Then it means your VM really got reset.

[dyoung23]

@fengye87 Yes.Thank you.

I thought how to manage VM using API,do we provide an official API or client to access Virtink?

[fengye87]

@dyoung23 Yes, there's a plan to add direct API support for VM power actions. It should arrive in the next release.

[dyoung23]

Hope the next release coming soon with a officail dashboard better.

[dyoung23]

@fengye87 I doubt how can I get the vm status using kubectl except from Events which will clear after a long time.

[fengye87]

@dyoung23 The VM's status have a phase field, you can get VM status (like if the VM is running) from this field.

[dyoung23]

@fengye87 But this phase field is really different with the power action.For example,I pause a vm,I only can get Running status,but I can't know it's paused from the status.

[fengye87]

@dyoung23 Good point. We shall improve this in the future.