[Question] Using fence_kubevirt in a k8s+Kubevirt environment.

[HideoYamauchi]

Hi All,

I am trying out pacemaker in a k8s+Kubevirt environment.
Previously, I configured STONITH using fence_kubevirt in an OCP-V environment.
In an OCP-V environment, fence_kubevirt can be configured using the steps in the following RHEL solution information.

• https://access.redhat.com/solutions/7068530

However, the same steps cannot be used with k8s+Kubevirt.

Can fence_kubevirt be used in environments other than OCP-V?
I have looked into this myself, but I could not find any information on using fence_kubevirt in a pacemaker cluster of k8s+Kubevirt VMs.

Best Regards,
Hideo Yamauchi.

[oalbrigt]

I have asked the dev, and will come back to you.

[HideoYamauchi]

@oalbrigt

Thank you.
I feel like I could use k8s+Kubevirt, but there's not much information available.

Many thanks,
Hideo Yamauchi.

[nirarg]

Hi @oalbrigt and @HideoYamauchi ,
I haven’t worked on this project for a long time, but it was developed with the upstream KubeVirt project.
To the best of my knowledge, it should be able to run as a standalone solution for Kubevirt on Kubernetes.

[HideoYamauchi]

Hi @nirarg

Thanks for your comment.

I'd be happy to use it, but I can't find any information on how to configure kubevirt on k8s+Kubevirt.
I've searched quite a bit, but I can't find anything other than information on RHEL (OCP-V).
Naturally, the same procedures as for RHEL cannot be used.

Best Regards,
Hideo Yamauchi.

[razo7]

Hi @HideoYamauchi,

IIUC, you are looking for a way to run Kubernetes with Kubevirt and the fence_kubevirt agent. If a pacemaker is not a requirement from your end, and you aim for a high-availability solution in your cluster, then I can suggest my team's project Medik8s. This project includes Kubernetes operators for these kinds of scenarios, and I believe Fence Agents Remediation (FAR) Operator would be a great fit for your needs.

This operator uses some of the well-known fence-agents from ClusterLabs, so after you install FAR, you will be able to execute the fence_kubevirt agent (or other supported agents). FAR adds a taint on the unhealthy node, executes the fence agent for this node, and then deletes the workloads (usually with the out-of-service taint).

Moreover, triggering and using the operator can be manual, and for an automatic solution with detection of unhealthy nodes (and more features like escalation between remediators), we recommend using the Node Healthcheck Operator.

I hope it suits your needs, and if you have more questions, then don't hesitate to reach out for help

[HideoYamauchi]

Hi @razo7

Thanks for your comment.

I'm familiar with Medik8s.

I've already tried it in an OCP-V environment, so I understand how useful it is.

The first thing we want to do is enable support between Pacemaker VM nodes, which requires fence_kubevirt.

The next step is Medik8s for worker node failures.

First, we need to check fence_kubevirt.

Best Regards,
Hideo Yamauchi.

[HideoYamauchi]

Hi @oalbrigt

I have asked the dev, and will come back to you.

Did you get any information after that?

Best Regards,
Hideo Yamauchi.