Chapter 2. Release notes

You can provision OpenShift Virtualization clusters that run Red Hat Enterprise Linux CoreOS (RHCOS) version 9.8 and version 10.2. RHCOS 10.2 is the default operating system. RHCOS 9.8 is also supported. Live migration of virtual machines (VMs) between RHCOS 9.x and RHCOS 10.x worker nodes is supported.

CNV-84749

The v1 API version of the HyperConverged custom resource (CR) is the default version. The v1beta1 version is deprecated and will be removed in a future release. OpenShift Virtualization adds new API fields only to the v1 version. Updates from earlier versions require no manual action.

CNV-79504

After you enable the enableMultiArchBootImageImport feature gate, OpenShift Virtualization creates architecture-specific boot sources for each supported architecture, such as rhel9-amd64 and rhel9-arm64. You can also specify the architecture for standalone data volumes and virtual machines. For more information, see Heterogeneous cluster support.

CNV-67900

The virt-platform-autopilot controller lowers deployment complexity by automatically enforcing configuration best practices across OpenShift Virtualization and adjacent Operators, such as the Kube Descheduler Operator and Cluster Observability Operator. It simplifies initial setup and maintains safe, optimal settings for swap readiness, load-aware balancing, and enhanced cluster observability to help ensure long-term stability across version upgrades.

CNV-81291

On clusters where the Kube Descheduler Operator is installed, OpenShift Virtualization automatically configures the descheduler to run in load-aware balancing mode, which accounts for both CPU and memory usage when it rebalances virtual machine (VM) workloads across nodes. The descheduler is an optional component that OpenShift Virtualization does not install for you. If the Kube Descheduler Operator is not installed, no rebalancing occurs. You can disable load-aware rebalancing in the OpenShift Container Platform web console. For more information, see How automated postinstallation configuration works.

CNV-80073

OpenShift Container Platform 5 replaces Extended Update Support (EUS) releases with Extended Life Cycle (ELC) releases. ELC releases occur every third minor version, starting with OpenShift Container Platform 5.2. OpenShift Container Platform 5.0 is not an ELC release.

Because ELC releases are three minor versions apart, a Control Plane Only update passes through two intermediate minor versions instead of one. For each intermediate version, update OpenShift Container Platform first, and then update OpenShift Virtualization to the latest z-stream release of that version before you move to the target ELC version.

For more information, see Control Plane Only updates.

CNV-82691

You can use the self-service Technical Supportability Review (TSR) on the Red Hat Customer Portal to validate your cluster configuration against Red Hat common practices. The self-service TSR uses AI to evaluate your cluster’s must-gather data and provides a prioritized executive summary that identifies your cluster’s top risks and recommends corrective actions. The TSR performs hundreds of checks across the OpenShift Container Platform platform, including OpenShift Virtualization, and coverage is continually expanding.

For more information, see Technical Supportability Review with AI tool and Red Hat Technical Supportability Review with AI: Proactive AI-Driven Cluster Assessments for OpenShift Container Platform.

On IBM Z® and IBM® LinuxONE, you can deploy IBM® Fusion Access for SAN. IBM® Fusion Access for SAN is a cloud-native solution that helps you manage existing Storage Area Network (SAN) infrastructure for workloads on OpenShift Virtualization. For more information, see Fusion Access for SAN.

CNV-98210

The template.kubevirt.io/v1beta1 API group and the VirtualMachineTemplate custom resource definition (CRD) are generally available. You can define a virtual machine (VM) blueprint as a cluster-native VirtualMachineTemplate object and then create VMs from that object. You can use the VirtualMachineTemplateRequest CRD to create a VirtualMachineTemplate from an existing VirtualMachine. This release updates the API group from v1alpha1 to v1beta1.

CNV-84116

You can configure a custom video device type when you create a virtual machine (VM). Configuring a custom device type overrides the default video configuration. You can specify different video devices, based on your guest operating system requirements and performance needs.

CNV-71195

The virt-platform-autopilot controller replaces the wasp-agent component for configuring higher VM workload density. When you enable memory overcommitment in the HyperConverged custom resource, the controller automatically deploys the necessary node-level configurations, which simplifies the setup process. You must manually provision swap storage on worker nodes by creating a MachineConfig object. For more information, see Configuring higher VM workload density.

CNV-83420

With this release, KubeVirt Redfish is generally available. KubeVirt Redfish exposes OpenShift Virtualization virtual machines through the standard Redfish API. Using KubeVirt Redfish, administrators can manage VM power states, boot configuration, and virtual media attachments. This update also adds bcrypt hashed passwords for more secure authentication to the KubeVirt Redfish API.

For more information, see Configuring KubeVirt Redfish for VM management.

You can pause the guest agent probe for a virtual machine (VM) by adding the kubevirt.io/pause-guest-agent-probes=true annotation to the VM. While the probe is paused, the kubelet does not restart the VM if the guest agent probe fails, for example during a guest operating system update. After the update completes, remove the annotation to resume the probe.

CNV-82132

On IBM Z® and IBM® LinuxONE, you can configure PCI passthrough for Network Express RoCE adapters and IBM® Internal Shared Memory (ISM) virtual PCI devices.

For more information, see PCI passthrough on IBM Z.

CNV-96643

You can deploy confidential computing workloads by using IBM® Confidential Computing Container Runtime for Red Hat Virtualization Solutions (CCRV). This integration enables CCRV workloads to use OpenShift Virtualization functionalities while maintaining confidential computing capabilities.

CNV-96593

On IBM Z® and IBM® LinuxONE clusters, OpenShift Virtualization only supports VMs on bare metal LPAR nodes. You can restrict both OpenShift Virtualization infrastructure components and VM workloads to bare metal LPAR nodes by applying a node placement rule.

For more information, see Configuring node placement for OpenShift Virtualization components on IBM Z.

CNV-95482

You can upload an IBM® Secure Execution bootable image and associate it with a predefined preference. Virtual machines (VMs) that you create from the image automatically enable the launchSecurity setting.

For more information, see Upload and use an IBM Secure Execution bootable image on IBM Z and IBM LinuxONE.

CNV-97316

If you have cluster administrator permissions, you can configure a default network for a project so that new virtual machines (VMs) in that project use it automatically. The default can be the pod network or a network attachment definition (NAD) that is available in the project. This setting applies only to VMs that you create after you configure it, and it does not change existing VMs.

For more information, see Configuring project-level default networks.

CNV-92410

You can change the network that a running virtual machine (VM) interface uses. In the OpenShift Container Platform web console, open the Configuration > Network tab, edit the existing interface, and select a different network.

CNV-82742

The DeclarativeHotplugVolumes feature gate is enabled by default in the HyperConverged custom resource (CR). You can insert and eject CD-ROM storage in a running virtual machine (VM) without restarting the VM and without changing the default configuration. In earlier releases, you enabled the feature gate manually because it was incompatible with ephemeral hot plug volumes.

CNV-81481

Administrators can bulk migrate storage classes of stopped virtual machines (VMs) in the OpenShift Container Platform web console. This functionality supports storage class migration for VMs in any state. You can select both running and stopped VMs in a single migration plan.

CNV-73500

You can clear the option to keep source persistent volume claims (PVCs) after a storage migration. If you clear this option in the web console, OpenShift Virtualization cleans up the original PVCs after the storage migration completes. By default, OpenShift Virtualization retains the source PVCs after a storage migration, and you clean up the PVCs manually.

CNV-73509

You can inject Kubernetes service account tokens into a virtual machine (VM) guest on Red Hat OpenShift Service on AWS (ROSA) clusters. The virt-launcher pod holds short-lived tokens used for AWS IAM Roles for Service Accounts (IRSA). OpenShift Virtualization uses containerPath volumes and virtiofs filesystem devices to expose these tokens to the VM guest, so that applications inside the VM can authenticate with AWS APIs without managing static credentials.

CNV-95907

The Containerized Data Importer (CDI) provisions the scratch space persistent volume claim (PVC) from the same storage class as the target volume instead of falling back to the cluster default storage class. This behavior applies to all clusters and requires no feature gate or HyperConverged custom resource (CR) configuration. To override it, specify a storage class in the spec.storage.scratchSpaceStorageClass field of the HyperConverged CR.

CNV-72238

You can assign a category to a virtual machine (VM) template and then filter the template catalog by category. You set the category when you create a template from an existing VM, and you can change it later. A category is optional, and a template can have only one category.

CNV-92485

On the Configuration > Network tab of a virtual machine (VM), each network name is a link. The link opens the detail page of the underlying network attachment definition (NAD), user-defined network (UDN), or cluster user-defined network (CUDN). Entries for Pod networking and for deleted networks remain plain text.

CNV-89211

You can build and save complex virtual machine (VM) queries directly in the search field, without opening the Advanced search dialog. You can combine multiple parameters and their multiple values in the search.

For more information, see Searching for virtual machines by using the web console.

CNV-74170

You can download table data in CSV format from OpenShift Virtualization list views in the OpenShift Container Platform web console, including Virtual Machines, Templates, DataSources, Bootable volumes, Migration policies, and Storage migrations. The downloaded CSV file has the filtered and sorted rows exactly as shown in the table and excludes hidden content and non-data columns.

CNV-89012

You can filter virtual machines by group in the OpenShift Container Platform web console. The Group filter is available in the advanced search dialog and as a drop-down filter in the main virtual machine list view.

CNV-94093

The left navigation menu now collapses when you go to OpenShift Virtualization pages, such as the virtual machine (VM) list and VM detail pages. This gives you more space to view and manage your VMs. If you expand the navigation menu, it stays open while you remain on virtualization pages. To turn off this behavior, go to Settings > User > General and clear the Auto-hide navigation menu toggle.

CNV-86441

OpenShift Virtualization exposes metrics that identify the GPUs attached to a virtual machine instance (VMI). You can correlate these metrics with the GPU metrics that the NVIDIA Data Center GPU Manager (DCGM) exporter provides to track GPU allocation and usage. Recording rules simplify the use of these metrics in OpenShift Virtualization dashboards.

CNV-51516

OpenShift Virtualization provides Perses-based dashboards that show virtual machine (VM) inventory, resource usage, service levels, time in status, top resource consumers, and node memory. OpenShift Virtualization deploys the dashboards automatically as PersesDashboard custom resources in the openshift-cnv namespace. To view the dashboards, click Observe > Dashboards(Perses) in the OpenShift Container Platform web console. The dashboards require Cluster Observability Operator (COO) with the monitoring stack enabled. For more information, see Virtualization dashboards overview.

CNV-42730

The following alerts for the OpenShift Virtualization Operator are included in the OpenShift Virtualization runbooks:

The HotplugVolume feature gate, which allows you to add storage without restarting your VM, is deprecated and will be removed in a future release. This feature gate will be replaced by DeclarativeHotplugVolumes.

CNV-73301

These machine types have been deprecated since OpenShift Virtualization 4.13 and will be removed in a future release as part of the move to Red Hat Enterprise Linux (RHEL) 10 user space. Update the machine type of your virtual machines (VMs) to pc-q35-rhel9.8.0. Updating the machine type requires a VM restart, so plan for downtime. For more information about updating multiple VMs at the same time, see Updating multiple virtual machines.

CNV-98500

The wasp-agent component, which configured higher VM workload density, is removed and is no longer supported. You configure higher VM workload density by using the virt-platform-autopilot controller. For more information, see Configuring higher VM workload density.

CNV-83420

Telemetry for ephemeral hot plug volumes is removed and is no longer supported. The DeclarativeHotplugVolumes feature gate replaces the deprecated HotplugVolume feature gate.

CNV-81481

Virtual machine (VM) owners can create, filter, and delete a user-generated template. You can create a template from an existing VM in the same project as the running VM or a different project from the OpenShift Virtualization user interface. To make sure that the data is consistent, stop the VM before you create a template. This capability is a Technology Preview feature.

CNV-81577

Virtual machine (VM) owners can create VMs from the OpenShift Virtualization cluster native template custom resource. The VM template tracks a golden image that is updated periodically, reducing errors and ensuring uniformity in the virtualized environment. You can host the template in all namespaces that you can control.

CNV-73392

The label for organizing and grouping virtual machines (VMs) in the web console is Groups, which more accurately reflects the functionality. This capability is a Technology Preview feature. For more information, see Organize virtual machines by using the web console.

CNV-91926

You can use the Model Context Protocol (MCP) server to manage OpenShift Virtualization resources through high-level workflows. This release adds tools for persistent storage, virtual machine (VM) templates, and HyperConverged Operator status and configuration. The kubevirt toolset is not enabled by default. To use these tools, add kubevirt to the toolsets setting in the OpenShift Container Platform MCP server configuration file. This capability is a Technology Preview feature.

CNV-77676

Cluster administrators can set the .spec.roleAggregationStrategy field in the HyperConverged custom resource (CR) to Manual to stop aggregating the virtualization cluster roles admin, edit, and view to users. After an administrator selects Manual, regular users cannot interact with virtualization resources such as virtual machines (VMs) in their projects, either from the command-line interface (CLI) or the web console. To grant a user these permissions, the administrator creates a RoleBinding object. This capability is a Technology Preview feature.

CNV-87048

Before this update, when you detached a disk from a virtual machine (VM), the DataVolume object retained a stale owner reference to the original VM. As a consequence, if you deleted the original VM, garbage collection deleted the DataVolume object and its underlying persistent volume claim (PVC). This caused data loss even when the disk was attached to a different VM. With this release, OpenShift Virtualization updates the owner reference when you detach a disk, so the DataVolume object no longer references the original VM. As a result, a detached disk and its PVC are preserved when you delete the VM that the disk was originally attached to.

CNV-71659

Uninstalling OpenShift Virtualization does not remove the feature.node.kubevirt.io node labels that OpenShift Virtualization creates. As a consequence, nodes can still appear as if they are configured for virtualization workloads.

To work around this problem, manually remove the feature.node.kubevirt.io labels from affected nodes after you uninstall OpenShift Virtualization.

CNV-38543

Live migration fails if a virtual machine name exceeds 47 characters. As a consequence, you cannot live migrate VMs with longer names.

To work around this problem, use VM names that are 47 characters or fewer when you create VMs that you plan to live migrate.

CNV-61066

If you enable the spec.featureGates.enableMultiArchBootImageImport feature gate after boot sources are imported, OpenShift Virtualization recreates the boot source import resources by using architecture-suffixed names. The names depend on the architectures that the cluster supports, for example fedora-amd64, fedora-arm64, and fedora-s390x. The earlier resources, which have no architecture suffix, remain in the openshift-virtualization-os-images namespace. As a consequence, duplicate boot sources are displayed in the web console and in the CLI. Storage usage can also increase, because the extra persistent volume claims (PVCs) and volume snapshots are still present.

To work around this problem, delete the leftover boot source resources. First, identify the active DataSource objects, which are the objects that resolve to the PVC or volume snapshot that you want to keep. Then, delete the stale DataSource objects and the PVCs and volume snapshots that they reference. As a result, only the architecture-suffixed boot sources remain.

CNV-68996

A MultiNamespaceVirtualMachineStorageMigration object can intermittently report the Completed phase within milliseconds of the migration starting, before the migration copies any data. By default, a migration plan that you create in the OpenShift Container Platform web console decommissions the source volumes after a successful migration.

As a consequence, the controller can delete the source persistent volume claims (PVCs) even though the migration never copied their contents to the target data volumes and PVCs. This can result in data loss.

To work around this problem, select the Keep original volumes at source after successful migration checkbox when you create the storage migration. Verify that the target volumes contain the migrated data, and then delete the source volumes manually. As a result, the source data remains available if a migration reports completion without copying data.

CNV-98287