Storage Agnostic qcow2 Incremental Backup Support

KubeVirt Enhancement Proposal: Incremental Backup

Design Document: Virt Incremental qcow2 Datamover

Overview

This epic implements storage agnostic qcow2 incremental backup support for virtual machines in OADP/Velero environments. The feature introduces a lightweight alternative to traditional VolumeSnapshot-based backups by leveraging libvirt tools for incremental qcow2 backups, significantly improving backup performance and enabling more frequent backup operations.

Background and Motivation

Taking a VolumeSnapshot and using kopia to process the entire volume for incremental changes is a heavyweight process. Creating an incremental qcow2 backup for the same volume is generally much more lightweight. This enhancement aims to:

• Improve volume backup performance in KubeVirt environments
• Enable more frequent backups through reduced resource overhead
• Leverage existing libvirt tooling for VM-specific backup operations
• Integrate seamlessly with existing Velero backup/restore processes

Architecture Overview

Virt Backup Controller

• New controller component for coordinating libvirt backup/restore operations
• Integration with Velero via existing infrastructure (BackupItemActions, RestoreItemActions)
• Support for DataUpload/DataDownload CRs with custom DataMover specification

Integration Points

• Minimal changes to Velero core (leverages existing extensibility)
• Uses Velero's volume policy model with new "virt" action
• Coordinates with existing OADP infrastructure

Key Features

Incremental Backup Support

• First backup for a volume is full backup
• Subsequent backups are incremental, capturing only changes
• Significant performance improvement over traditional CSI snapshot methods
• Support for more frequent backup schedules

Storage Agnostic Design

• Works independently of underlying storage provider
• Leverages qcow2 format for efficient incremental operations
• Compatible with various backup storage locations (BSL)

VM-Aware Operations

• Integration with KubeVirt VirtualMachineInstance resources
• Support for QEMU backup capabilities (CBT - Changed Block Tracking)
• Automatic detection of backup-enabled disks

Technical Implementation

Upstream Velero Changes

• Update volume policy model to allow unrecognized actions (treat as "skip")
• Add SnapshotType support in DataUpload/DataDownload operations
• Enhance restore actions for virt datamover compatibility

Plugin Development

_ _VirtualMachineInstance BackupItemAction (BIA) Plugin:*

• Check for QEMU backup enablement (Spec.Domain.Devices.Disk[i].CBT = true)
• Validate volume policy configuration ("virt" action)
• Create temporary PVC for qcow2 image storage
• Coordinate with Virt Backup Controller

_ _RestoreItemAction (RIA) Plugin:*

• Handle restoration of qcow2-backed volumes
• Coordinate with controller for libvirt restore operations
• Manage temporary storage during restore process

Controller Implementation

_ _Virt Backup Controller:*

• Watch for DataUpload CRs with virt datamover specification
• Execute libvirt backup operations using appropriate tools
• Manage backup chain and incremental relationships
• Handle backup storage and retrieval operations

Volume Policy Configuration

Policy Requirements

• VM volumes must be configured with "virt" policy action
• No mixing of policy types within single VM (all "virt" or "skip")
• Automatic fallback for unsupported configurations

Prerequisites and Requirements

VM Configuration

• QEMU backup must be enabled on VM disks (CBT = true)
• VMs must be running KubeVirt with appropriate capabilities
• Sufficient temporary storage for qcow2 image processing

Infrastructure Requirements

• Virt Backup Controller deployment
• Access to libvirt tools and capabilities
• Network connectivity for backup storage operations

Use Cases

Primary Use Cases

• High-frequency backups for development environments
• Efficient backup of large VM workloads
• Storage-independent backup solutions
• Reduced backup window requirements

User Workflow

1. Configure VM with QEMU backup enabled (CBT = true)
2. Set volume policy to "virt" for target volumes
3. Create OADP backup with virt-enabled volumes
4. Controller performs incremental qcow2 backup
5. Restore operations use libvirt tooling for recovery

Benefits and Advantages

Performance Improvements

• Significantly faster than CSI snapshot + kopia processing
• Reduced resource consumption during backup operations
• Minimal impact on running VM performance
• Efficient incremental change detection

Operational Benefits

• More frequent backup schedules possible
• Storage provider independence
• Leverages proven libvirt backup technology
• Seamless integration with existing OADP workflows

Limitations and Considerations

Current Limitations

• Requires QEMU backup capability (CBT support)
• VM-specific solution (not applicable to non-VM workloads)
• Additional controller component required
• Temporary storage requirements for qcow2 processing

Future Enhancements

• Deep integration with Velero data mover pluggability
• Enhanced monitoring and observability
• Support for additional VM backup formats
• Performance optimization and tuning

Testing Strategy

• Unit tests for plugin components
• Integration tests with KubeVirt environments
• End-to-end backup and restore validation
• Performance benchmarking against traditional methods
• Multi-VM scenario testing
• Failure recovery and error handling validation

Target Release

• Implementation target: OADP 1.6.0
• Technology Preview phase initially
• Full support after validation and testing

Success Criteria

• Successful incremental backup of KubeVirt VMs
• Performance improvement over traditional CSI snapshots
• Seamless integration with existing OADP workflows
• Reliable restore operations
• Comprehensive documentation and user guidance