SUSE Harvester

SUSE Harvester (officially rebranded as SUSE Virtualization starting with v1.4) is an open-source hyper-converged infrastructure (HCI) solution purpose-built for running virtual machines on Kubernetes. It combines KubeVirt for VM management, Longhorn for distributed storage, and an embedded RKE2 Kubernetes cluster into a single, turnkey platform. The latest release is v1.7 (January 2026), which adds ARM64 support and upgrades the base OS to SL Micro 6.1.

Harvester is positioned as a modern, cloud-native alternative to legacy virtualization platforms such as VMware vSphere and Proxmox VE. It is fully free and open source under the Apache 2.0 license, with optional commercial support from SUSE.

Harvester is installed as a bare-metal operating system image. Each node boots a customized SLE Micro-based OS (SL Micro 6.1 as of v1.7) with an embedded RKE2 Kubernetes cluster. All Harvester services run as Kubernetes workloads on top of this foundation.

Layer Breakdown

Harvester provides a full VM lifecycle management experience through its web UI, API, and standard Kubernetes tooling. Virtual machines are represented as VirtualMachine custom resources managed by KubeVirt.

Creating Virtual Machines

VMs can be created from the Harvester UI, via kubectl, or through the Rancher UI. Each VM is defined by a VirtualMachine CRD that specifies CPU, memory, disks, networks, and cloud-init configuration.

apiVersion: kubevirt.io/v1
kind: VirtualMachine
metadata:
  name: ubuntu-vm
  namespace: default
spec:
  running: true
  template:
    spec:
      domain:
        cpu:
          cores: 4
        memory:
          guest: 8Gi
        devices:
          disks:
            - name: rootdisk
              disk:
                bus: virtio
            - name: cloudinitdisk
              disk:
                bus: virtio
          interfaces:
            - name: default
              masquerade: {}
      networks:
        - name: default
          pod: {}
      volumes:
        - name: rootdisk
          persistentVolumeClaim:
            claimName: ubuntu-root-pvc
        - name: cloudinitdisk
          cloudInitNoCloud:
            userData: |
              #cloud-config
              password: changeme
              chpasswd: { expire: false }
              ssh_pwauth: true

VM Templates

Harvester supports VM templates that pre-define resource configurations, disk images, and cloud-init scripts. Templates can be versioned and shared across namespaces for standardized VM provisioning.

Key VM Operations

Live Migration

Live migration allows running VMs to move between Harvester nodes with zero downtime. This is essential for node maintenance, rolling upgrades, and workload rebalancing. Harvester triggers live migration automatically during upgrade operations.

Harvester uses Longhorn as its integrated distributed block storage engine. Every VM disk is backed by a Longhorn volume by default, providing replication, snapshots, and backup capabilities out of the box. Since v1.4, Harvester also supports third-party CSI drivers for VM data disks, enabling integration with external storage solutions for specific performance or infrastructure needs (note: VM backup currently works only with the Longhorn v1 data engine).

Storage Classes

Harvester ships with a default longhorn storage class. Custom storage classes can be created to control replica count, data locality, and reclaim policies.

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: harvester-longhorn-high-perf
provisioner: driver.longhorn.io
parameters:
  numberOfReplicas: "3"
  dataLocality: "best-effort"
  diskSelector: "nvme"
reclaimPolicy: Delete
volumeBindingMode: Immediate

Volume Management

Backup Targets

Longhorn supports exporting volume backups to external targets for disaster recovery. Harvester supports two backup target types:

• S3-compatible storage — AWS S3, MinIO, or any S3-compatible endpoint
• NFS — Network file system mount for on-premises backup

Harvester networking is powered by Multus and the bridge CNI plugin, enabling VMs to attach to multiple networks including VLAN-tagged physical networks. Canal (Calico + Flannel) provides the default cluster network for Kubernetes pods.

Network Types

Creating a VLAN Network

apiVersion: k8s.cni.cncf.io/v1
kind: NetworkAttachmentDefinition
metadata:
  name: vlan100
  namespace: default
  annotations:
    network.harvesterhci.io/route: '{"mode":"auto"}'
spec:
  config: |
    {
      "cniVersion": "0.3.1",
      "name": "vlan100",
      "type": "bridge",
      "bridge": "mgmt-br",
      "vlan": 100,
      "ipam": {}
    }

Network Isolation

Network isolation is achieved through VLAN segmentation at the physical switch level. Different VM workloads can be placed on separate VLANs to enforce tenant or security-zone boundaries. Harvester supports multiple VLAN networks per VM for multi-homed configurations.

Load Balancer Integration

For Kubernetes clusters running on Harvester VMs, the Harvester Cloud Provider can provision load balancers backed by Harvester's DHCP-allocated IPs or a configured IP pool, enabling LoadBalancer-type services without an external LB appliance.

Harvester integrates deeply with SUSE Rancher, providing a unified management plane for both virtualization infrastructure and Kubernetes clusters. Rancher can import Harvester clusters and use them as infrastructure providers for downstream K8s clusters.

Integration Capabilities

Harvester provides VM-level backup and restore capabilities, plus the ability to back up the Harvester cluster configuration itself for disaster recovery scenarios.

VM Backups

VM backups capture the complete state of a virtual machine, including all attached volumes and metadata. Backups are stored on a pre-configured backup target (S3 or NFS) and can be used to restore VMs on the same or a different Harvester cluster.

• On-demand backups — Triggered manually from the UI or API
• Scheduled backups — Cron-based backup schedules per VM
• Incremental — After the initial full backup, subsequent backups are incremental

Restore Procedures

Harvester Cluster Backup

Beyond VM-level backups, administrators should back up the Harvester cluster configuration including etcd snapshots, Kubernetes resource definitions, and Harvester settings. RKE2 provides automatic etcd snapshots that can be configured for external S3 storage.

Harvester supports rolling upgrades that update the entire stack — OS, Kubernetes, KubeVirt, Longhorn, and Harvester components — with minimal disruption to running VMs.

Upgrade Process

1. Pre-flight checks — Harvester validates cluster health, Longhorn volume replicas, and available disk space before starting
2. Download upgrade image — New version ISO is uploaded or pulled from a configured repository
3. Node-by-node rolling upgrade — Each node is cordoned, VMs are live migrated to other nodes, the node is upgraded, then uncordoned
4. Post-upgrade validation — Cluster health checks confirm all components are running the new version

Zero-Downtime Requirements

Upgrade Troubleshooting

• Stuck upgrades — Check the Upgrade custom resource status and Harvester controller logs
• Failed VM migration — Identify non-migratable VMs and shut them down manually before retrying
• Longhorn volume degraded — Wait for volume replica rebuilds to complete before proceeding to the next node
• Rollback — Harvester does not support in-place rollback; restore from a pre-upgrade etcd snapshot if necessary

How Harvester compares to other popular virtualization platforms:

Harvester is released under the Apache License 2.0, making it fully free and open source with no feature restrictions, no per-socket fees, and no node limits.

Key items to verify when planning, deploying, or auditing a Harvester environment: