Keep Your VM Dancing With Volume Migration
summary
Declarative API triggers a full online copy of a VM disk while the guest runs, using updateVolumeStrategy=migration to create a VolumeMigration object and a second vmlauncher pod that performs a qemu block-copy between PVCs; migration can take hours for large disks, supports cancel by reverting the volumeSet, and works with local or shared storage but not for RWO-mounted local live-migrations. Current limits: cross-node only, no automatic guest filesystem resize, and PVC creation/planing is left to higher-level tooling.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 1:53Kubernetes lacks volume migration
- 3:43Declarative API for migration
- 7:19How migration works
- 9:07Canceling migration
- 10:10Limitations and future plans
- 12:20Live demo
- 15:54Q&A
related talks
28mFrom Pre-Population To Disasters: Manage And Protect the State Of VMs
31mMigrating from VMware to Kubernetes
27mCase Study: Volume Populators for Virtual Disks?
22mDirty Quota-Based VM Live Migration Auto-Converge
32mYour Pods can be stateless, your VMs cannot! How to manage state in k8s
42mVirtual Machine Orchestration in Kubernetes Utilizing KubeVirt
27mZero copy migration from VMware to Kubernetes
24mCapacity-aware Dynamic Volume Provisioning For LVM-based Local Storage