Goal

Since this feature did not see adoption in its Tech Preview phase we are going with a Pre-GA level of support. This means that we will fully support usage in PoCs but will require customers to file a support exception in order to be supported in production.
Provide GA-level support for migrating running and stopped VMs to a different storage class.

User Stories

• As a VM owner I want to move my VM storage (disks and TPM) from one storage class to another. If the VM is running I want it to remain running during the migration operation.

Non-Requirements

• Some storage cannot be migrated (due to technical limitations or lack of implementation). See here for more information. For the following cases, the storage can be migrated only if the VM is powered off and manual changes to the VM spec post migration may be necessary.
  • Shared disks
  • virtiofs disks
  • hotplugged disks (hope to remove this limitation in this epic)
  • Interface change (lun <-> disk)
  • lun disks (no support for the `slice` option libvirt RFE)
  • SCSI persistent reservation disks (VMs using these do not support live migration)
  • Local storage on the same node (In order to complete the storage live migration the VM must live migrate and this requires moving it to a different node)

Notes

• The user must be able to change the Volume Mode of the disk (in order to migrate from HPP to ceph for example)

Risks

• Graduation of VolumeUpdateStrategy and VolumeMigration
  • Needed in kubevirt:main in order to safely declare GA in CNV 4.18
  • PR is ready but on-hold by kubevirt maintainers