Managed KubeVirt on your own servers: a VMware alternative
Run virtual machines on your own bare metal with RunOS and KubeVirt. Compare host, storage, migration, and recovery requirements before moving from VMware.
RunOS uses KubeVirt to run virtual machines on the servers you own. RunOS installs KubeVirt on your Kubernetes cluster and gives you the Console, CLI, and MCP to operate VMs. This is an option if you want to move suitable workloads off VMware vSphere and keep the hardware.
How do I turn a server into a VM host?
Join a supported Ubuntu bare-metal server to a RunOS cluster and opt the server in as a VM host. The cluster needs a supported Kubernetes version and a LINSTOR storage group backed by a thin-provisioned pool. Install virtualization in RunOS, then create a VM group and launch a guest from an image.
You can start with one eligible server. To move a running VM between hosts, add another eligible server and choose a storage profile that supports live migration.
What can I manage in RunOS?
| What you need | What RunOS provides |
|---|---|
| Provision and operate VMs | Create VMs in the Console or with runos vms. Use RunOS MCP to run permitted VM operations from an assistant. |
| Separate workloads | Assign every VM to a group. Groups isolate traffic by default and let you set quotas for CPU, memory, disk, and VM count. |
| Access and reuse guests | Connect over SSH or use a serial or VNC console. Capture snapshots and images to reuse guest configurations. |
| Move workloads during maintenance | Live-migrate eligible VMs with runos vms migrate. The VM keeps its address. |
See the RunOS KubeVirt release notes for a CLI example and more on VM operations.
What should I test before moving from VMware?
Start with one representative workload. Check the guest image, network access, disk performance, and the way you back up and restore its data. Test live migration on your intended storage profile and rehearse recovery before moving production workloads.
- Host failure: RunOS does not automatically restart a VM on another host. Confirm that the failed host has powered off before you use
runos vms recover. - GPU passthrough: A VM with a GPU attached stays on its host and cannot live-migrate.
- Feature fit: Treat this as an alternative for the workloads you validate. Check any vSphere workflows you depend on, including image conversion and backup integration.
Talk to us about a VM pilot if you want to check your host, storage, and guest requirements before moving a workload.
Related use cases
- Data center operators Discover, provision, and recover physical servers through their management controllers. Allocate machines to tenant accounts on hardware you own.
- GPU capacity Run on-demand GPU VMs, interruptible spot VMs, and inference pools on your fleet. Set priorities and preview what RunOS would reclaim.