SUSE Virtualization (formerly Harvester)

SUSE Virtualization (formerly Harvester) is an open-source hyperconverged infrastructure (HCI) platform built on Kubernetes. It runs on bare metal servers and combines KVM-based virtual machine management, distributed block storage, and networking in a single cluster. It manages virtual machines (VMs) and containers together, and SUSE positions it as an open-source alternative for operators moving off traditional virtualization stacks.

Key Capabilities

• Virtualization Layer: KubeVirt provides VM management using KVM on top of Kubernetes. The platform runs on SUSE Linux Micro, an immutable Linux distribution, and RKE2 (Rancher Kubernetes Engine 2). Both are bundled, so no separate operating system install is required.
• VM Lifecycle Management: VMs can be created, edited, cloned, and deleted through the web UI, the Kubernetes API (VirtualMachine objects), or the Terraform provider. Options include SSH key injection, cloud-init, graphical and serial consoles, node and workload affinity rules, and VM templates. Documented guest operating systems include Windows.
• Live Migration: VMs move between hosts with no downtime when prerequisites are met. VMs with host device passthrough (PCI, vGPU) or single-node scheduling constraints are non-migratable. Live-migratable VMs are migrated automatically before a node is upgraded.
• Distributed Storage: SUSE Storage, built on the Longhorn project, provides block volumes for VMs and pods. It creates a dedicated storage controller per volume and replicates volumes synchronously across nodes. Third-party storage works through Container Storage Interface (CSI) drivers, with a vendor certification program. Volumes can be created from qcow2, raw, or ISO images, and a dedicated storage network can isolate replication traffic.
• Backup, Snapshot, and Restore: VM backups go to an NFS or S3-compatible backup target and can restore a new VM or replace an existing one. Backups cover SUSE Storage volumes only, not volumes on external storage.
• Networking: A built-in management cluster network (mgmt) carries intra-cluster traffic, with bonded NICs recommended for production. Custom cluster networks isolate VM traffic, and VMs attach to VLAN or untagged networks. A virtual IP (VIP) provides access to the cluster.
• Rancher Integration: SUSE Rancher Prime imports and manages multiple SUSE Virtualization clusters through its Virtualization Management feature, with Rancher authentication and role-based access control (RBAC) for multi-tenancy. Rancher can provision RKE2 guest Kubernetes clusters on VMs through the Harvester node driver. An experimental feature supports container workloads directly on the cluster.
• Add-ons: Optional add-ons keep the base installation small. They include rancher-monitoring (Prometheus and Grafana), rancher-logging, pcidevices-controller, nvidia-driver-toolkit, and vm-import-controller. Experimental add-ons include kubeovn-operator, csi-driver-lvm, and rancher-vcluster. A gated SUSE VM Migration Toolkit add-on migrates VMs from VMware.
• Upgrades: Upgrades run on a four-month minor release and two-month patch cadence, in connected and air-gapped environments. Downgrades are not supported.

Audience & Use Cases

• Audience: Infrastructure and virtualization administrators, platform engineers, and cloud-native operators.
• Use Cases: Replacing or consolidating traditional virtualized infrastructure, running VMs alongside containers on one platform, migrating VMs from VMware, and deploying from core data centers to edge locations. SUSE also documents it as a host for guest Kubernetes clusters.

Technical Specifications

• Hardware (production): AMD64 or ARM64 with hardware-assisted virtualization, 16 cores minimum, 64 GB memory minimum, and 500 GB disk minimum with 1 TB or more recommended. Disks need 5,000+ random IOPS (SSD or NVMe), and only local disks and hardware RAID are supported. A 10 Gbps Ethernet network is the minimum for production.
• Cluster Topology: A three-node cluster is required for high availability (HA). The first three nodes become management nodes with an etcd quorum. Single-node clusters are supported but lack HA, multiple replicas, and live migration. Witness nodes are supported. Mixed-architecture clusters are not supported.
• Boot and Installation: Installation is from an ISO or by iPXE script. New installations from v1.8.0 require UEFI. Nested virtualization is not supported.
• Current Version Components (v1.8): KubeVirt v1.7, SUSE Storage v1.11, SUSE Rancher Prime v2.14, RKE2 v1.35, SUSE Linux Micro 6.2.