Faster Backups, Less Storage: Unlocking Incremental Backups in KubeVirt
summary
KubeVirt adds storage-agnostic incremental backups that use QEMU-native Changed Block Tracking to send only changed blocks after an initial full snapshot, cutting typical transfer sizes from full-VM to megabyte-scale deltas. The implementation places a lightweight QCOW2 overlay in a VM state PVC (requires VM restart) and exposes a controller/CRD pipeline—VirtualMachineBackup and VirtualMachineBackupTracker—that runs in push mode, attaches a target PVC to virt-launcher, and writes sparse QCOW2 outputs via libvirt; pull mode and offline VM handling are planned.
AI-written summary · about
chapters (10)
- 0:00Incremental backups in KubeVirt
- 1:28Current backup limits
- 4:26Why incremental and storage agnostic
- 6:55Change Block Tracking and checkpoints
- 9:24Enabling CBT with QCOW2
- 11:13Push vs pull backup modes
- 12:38New backup CRDs
- 15:50Backup pipeline workflow
- 17:17Backup output and restoration
- 19:04Progress and future work
related talks
14mIncremental snapshots/backups in KVM plugin of Apache CloudStack
06mVM: When Compute Moves but Data Stays
23mImporting KVM instances from external KVM hosts or QCOW2
38mImproving your Kubernetes Workload Security
18mAgility Unleashed: Migrating VMs to next generation open source platforms
32mKubernetes Backup Legitimized: CSI Changed Block Tracking Has Arrived
42mEnhancing Hypervisor and Cloud Solutions Using Embedded Linux
23mLive migration from VMware to OpenStack