Skip to content

Harvester v1.2.0 Release
Compare
Choose a tag to compare
@guangbochen guangbochen released this 08 Sep 14:26
· 299 commits to master since this release
v1.2.0
a845350

Release Note

This release follows Harvester v1.2.0, the documentation is available at https://docs.harvesterhci.io/v1.2.

Rancher Support

If you are using Rancher for multi-cluster management of Harvester, you must upgrade the Rancher server to v2.7.6 first before upgrading the Harvester to v1.2.0.

For more details about the Harvester & Rancher support matrix, please visit our website here.

Downloads

ISO

💿 harvester-v1.2.0-amd64.iso

Additional Artifacts

📁 https://releases.rancher.com/harvester/v1.2.0/harvester-v1.2.0-vmlinuz-amd64
📁 https://releases.rancher.com/harvester/v1.2.0/harvester-v1.2.0-initrd-amd64
📁 https://releases.rancher.com/harvester/v1.2.0/harvester-v1.2.0-rootfs-amd64.squashfs
📁 https://releases.rancher.com/harvester/v1.2.0/harvester-v1.2.0-amd64.raw.zst
✅ https://releases.rancher.com/harvester/v1.2.0/harvester-v1.2.0-amd64.sha512
📝 https://releases.rancher.com/harvester/v1.2.0/version.yaml

Install

Harvester can be installed with the ISO image or via PXE boot.

Highlights

  • The new Branding setting allows an administrator to customize the product name, logos and color scheme. See #3396.
  • The system footprint of Harvester is reduced, better fitting Edge use cases. See #3262.
  • Automatic adjustment of ResourceQuota during migration. When migrating VMs in Harvester, the resource limit of ResourceQuota is increased based on the specifications of the target VM before migration; and automatically decreased or restored to its previous value when the migration is complete. See #3124.
  • Harvester supports adding an emulated Trusted Platform Module (TPM) to a VM. TPM is a cryptoprocessor that secures hardware using cryptographic keys and is a hard requirement of Windows 11. See #2910.
  • Harvester supports passing through SR-IOV capable network virtual functions (VFs) to workload VMs. See #2763.
  • Add rancher-vcluster(experimental) addon and bare-metal container support for the Harvester cluster. See #2679.
  • Harvester node driver(v0.6.5) supports multiple NICs and multiple disks. See #2426.
  • Harvester supports installing third-party Container Storage Interface (CSI) in your Harvester cluster to support and use external storage as the VM's non-root data disks. See #2405.
  • Enable harvester-seeder addon to configure Out-of-Band Access for a node and map the corresponding underlying hardware to a node. See #2318.
  • New installation mode supports installing Harvester to a machine without configuring the node. See #2198.
  • Introduce new Harvester cloud provider design in the guest RKE2 Kubernetes cluster and add VM load balancer support in the Harvester cluster. See #2134.
  • Harvester supports draining the node when entering maintenance mode. See #2723.
  • Harvester supports creating a VM template with data from an existing VM, including root and data disks. See #571.
  • Pod limit per node increased to 200. See #2707.

Upgrade

Upgrade from v1.1.2 to v1.2.0

Major Bug Fixes

  • The containerd registry file become empty after upgrading from Harvester v1.1.2. See #4156.
  • Booting in EFI mode checkbox is not working correctly with Terraform Harvester Provider. See #4145.
  • When you try to upgrade from Harvester v1.1.2 to v1.2.0, the upgrade is stuck on Waiting for RKE2 to be upgraded. fleet-agent was scaled down but failed to scale up. See #4119.
  • When you try to upgrade from Harvester v1.1.2 to v1.2.0, the upgrade is stuck on the post-drain job, and the VM is unresponsive. See #4095.
  • Projects/Namespaces page is not available. See #4094.
  • When accessing an imported Harvester in Rancher, Grafana metrics don't display. See #4087.
  • A VM needs at least 10 minutes to recover after a host failure automatically. See #4049.
  • The harvester_config of terraform-provider-rancher2 v1.25.0 and v3.0.0 are incompatible. See #3997.
  • When creating an IP Pool, the Project drop-down list is missing in the MCM (rancher-manager-support) view. See #3984.
  • When you upgrade from Harvester v1.1.2 to v1.2.0 and the pcidevices-controller addon is Enabled, no PCI Devices are displayed. See #3957.
  • When you upgrade from Harvester v1.1.2 to v1.2.0, you are unable to restore an existing VM snapshot. See #3943.
  • When viewing Harvester cluster from Virtualization Management page in Rancher, project members cannot see a newly created volume from the Volumes page. See #3873.
  • When you upgrade from Harvester v1.1.2 to v1.2.0, a change in Rancher causes upgrading RKE2 to fail because the local-rke-state secret is the wrong type. See #3858.
  • When editing a Host configuration, you cannot view or edit Disk tags. See #3849.
  • When creating an RKE2 cluster, the cluster's volumes and networks differ from the final configuration. See #3786.
  • When a node has a memory size of more than 1000 GB, the node resources on the Host page are not displayed correctly. See #3607.
  • When creating Output or Cluster Output, the output provider OpenSearch is missing. See #3589.
  • VM with hotplug volume gives IO error after deleing hotplug volume pods twice. See #3586.
  • The setting controller does not reprocess the value when there is an error. See #3519.
  • Fixed invalid havester-csi-ccm-versions setting. See #3404.
  • Can not control qemu-guest-agent cloud-init when creating guest RKE1/RKE2 cluster. See #3392.
  • Cluster Members cannot see Cluster Networks to choose from when creating a VM Network. See #3300.
  • Volume created by the Harvester CSI driver gets deleted after reprovisioning RKE2 nodes. See #3272.
  • During an upgrade, multiple nodes are cordoned off, causing the upgrade to be stuck waiting for volumes to become healthy. See #3216.
  • Time sync config and time sync monitoring missing. See #3133.
  • server_url with https:// but without the port :443 causes node join to fail. See #2737.
  • No resource limits set for the Kubevirt pods in harvester-system namespace. See #2729.
  • An upgrade is stuck with a node in "Pre-drained" state since the node's volume has too many system snapshots. See #2640.
  • After configuring the MTU attribute of the Network Config , the associated bridge interface still uses the default value of 1500. See #4353.
  • Fix uploading large image from local file fails with Context Canceled. See #3450.

Others

  • When installing a new Harvester node using v1.2.0 ISO image, make the default persistent partition size configurable and increase its default size. See #3842.
  • Bump Harvester dependency components in v1.2.0. See #3747.
  • Add priorityClass to the Longhorn. See #3756.
  • Bump baseOS to SLE Micro for Rancher 5.3. See #3279.
  • The rancher/harvester-cluster-repo image is now published on dockerhub. See #2670.
  • The pci-devices addon is now becoming GA feature. See #3524.
  • Support bundle kit enhancement in Harvester v1.2.0. See #3494, #3702.
  • Upgrade: speed up image preloading time and space utilization. See #4208.
  • Retain the Grafana PV when rancher-monitoring addon is disabled. See #4118.
  • Harvester single node cluster support. See #3889.
  • Add more pre-checks via Harvester webhook before upgrade. See #2935.
  • Display more detailed info of volumes in the volumes page. See #2856.
  • Support specify both global and custom termination grace period for stopping a virtual machine. See #2466.
  • Support advanced scheduling of VMs with affinity rules(Harvester & RKE2 only). See #2317.
  • Take fsfreeze before taking a VM backup or snapshot. See #1723.

Known Issues

  • When multiple rke2 clusters exist in the same namespace, an re-sync topology error: nodes not found in cloud-provider occurs where the nodes cannot be found in the cloud-provider controller. See #4459.
  • Fail to scale down RKE2 Harvester node driver cluster. See #4358.
  • Failed to pass guest cluster name into Harvester cloud provider. See #4232.
  • Unable to access Node IP addresses over HTTPS on VM router in the Harvester cluster. See #3960 [Workaround].
  • Unable to select Harvester Cloud Provider in Rancher 2.7.x. See #3750 [Workaround].
  • Harvester upgrade may stuck in another operation (install/upgrade/rollback) by fleet. See #3675 [Workaround].
  • Upgrade issues:

Dependency Component Versions

Component Version
Longhorn v1.4.3
KubeVirt v0.54.0
Embedded Rancher v2.7.5
RKE2 v1.25.9+rke2r1

Contributors

@albinsun
@bk201
@chrisho
@connorkuehl
@DaiYuzeng
@FrankYang0529
@futuretea
@guangbochen
@ibrokethecloud
@irishgordo
@khushboo-rancher
@lanfon72
@LucasSaintarbor
@mantis-toboggan-md
@markhillgit
@masteryyh
@n313893254
@noahgildersleeve
@rebeccazzzz
@starbops
@TachunLin
@Vicente-Cheng
@vickyhella
@w13915984028
@WebberHuang1118
@WuJun2016
@yaocw2020
@yasker

Contributors

Vicente-Cheng, bk201, and 26 other contributors
Assets 6