What's New in Harvester v1.4.0

major features and improvements

Harvester v1.4.0 introduces several significant features focused on storage, security, and virtual machine management. Here's what's new:

Storage Enhancements

Longhorn V2 Data Engine (Experimental)

• Leverages Storage Performance Development Kit (SPDK)

• Significantly reduces I/O latency while improving IOPS and throughput

• Currently supports new volumes as extra disks

• Migration and cloning features planned for future releases

Volume Encryption

• Support for encrypting both boot volumes and data volumes

• Uses Linux dm_crypt and cryptsetup

• Requires Kubernetes secret for passphrase

• Dedicated StorageClass with encryption fields

Local Storage Support

• Improved performance and latency using local host storage

• Implemented through LVM for logical volume management

• Requires harvester-csi-driver-lvm add-on installation

RWX Volume Support

• Read-write-many (RWX) volumes now available for guest cluster workloads

• Requires RWX StorageClass configuration on both host and guest clusters

Virtual Machine Management

Enhanced Backup Capabilities

• Scheduled VM backups and snapshots with runtime controls

• Automatic VM image syncing during restoration

• Space usage limits for backups/snapshots at namespace and VM levels

• Each backup now includes corresponding VM image

Node Maintenance Improvements

• Configurable VM behavior during node maintenance

• Options for VM shutdown/restart instead of migration

• Support for pre-shutdown command execution

• Better control over VM placement post-maintenance

Hardware Support

USB Passthrough

• Direct VM access to USB devices

• Individual device passthrough rather than controller passthrough

• Allows sharing USB controller between node and VMs

• Managed through pcidevices-controller add-on

CPU Pinning

• Dedicated physical CPU cores for VMs

• Requires Kubernetes CPU Manager enablement

• Improves performance predictability for demanding workloads

Third-Party Storage Support

• Support for diskless server installations

• Compatible with external iSCSI devices and SAN storage arrays

• Requires pre-configured iSCSI target

Critical System Bugs Fixed

Storage

• Fixed single-replica volume breaking maintenance mode

• Resolved volume degradation status after VM restart

• Fixed disk cleanup during reinstallation

• Addressed storage quota and orphaned volume issues

Virtual Machine Operations

• Fixed VM migration reliability issues

• Resolved VM restore functionality for backups between versions

• Fixed VM startup location with PCI passthrough devices

• Corrected CPU manager startup problems

Network

• Fixed load balancer IP release during IPAM changes

• Resolved network interface renaming during reboots

• Fixed storage network configuration through UI

• Corrected VLAN ID validation issues

Backup & Restore

• Fixed backup loss during NFS service disconnection

• Resolved backup restoration on new clusters without networks

• Fixed image sync issues during cluster restoration

• Addressed backup scheduling conflicts

Known Issues to Watch

Three significant issues remain in 1.4:

1. VMs may lose connectivity to Harvester VIP/Node IP after RKE2 restart (#7041)

2. Backup restoration compatibility issues between v1.3.2 and v1.4.0 (#6721)

3. PCI Passthrough devices can cause VMs to start on incorrect hosts (#6648)

Users should consult documentation for workarounds and updates on these issues.

Impact on Upgrades

Key upgrade-related fixes:

• Resolved upgrade failures in witness node clusters

• Fixed image cleanup after v1.3.2 to v1.4.0 upgrades

• Addressed upgrade webhook error messaging

• Fixed upgrade initiation during ongoing upgrades