containerized-data-importer/doc/preallocation.md

Data Volume preallocation

Introduction

In order to improve write performance, QEMU provides the ability to preallocate image files, used as backing for DataVolumes.

There are several methods for preallocation. CDI uses the best one available, depending on the underlying file system and device type. It tries to use the OS's fallocate call if the filesystem supports it and falls back to "full" preallocation for block devices. Preallocation does not depend on the source of the DV, i.e. it can be used for import, upload or blank DVs.

See qemu-img documentation to learn more about preallocation. See also below for considerations regarding different datavolume types.

Preallocation for a DataVolume

To preallocate space for a DataVolume, use preallocation field in DataVolume's spec:

apiVersion: cdi.kubevirt.io/v1beta1
kind: DataVolume
metadata:
  name: preallocated-datavolume
spec:
  source:
    ...
  storage:
    ...
  preallocation: true

Enabling preallocation globally

Preallocation can be also turned on for all DataVolumes with an entry in the spec.config of the CDI resource:

kubectl patch cdis.cdi.kubevirt.io cdi -p '{"spec": { "config": { "preallocation": true }}}' -o json --type merge

If not specified, the preallocation option defaults to false.

Considerations

Preallocation can be used in the following cases:

• import (any source)
• upload
• cloning
• blank images
• blank block data images.

However, different DataVolume types use different preallocation methods:

• for blank block devices: the space is filled with zeros.
• for cloning volumes: handling sparse files is turned off, so the destination volume is filled in full, even if the source volume is not preallocated.
• blank images, upload and import volumes use qemu-img preallocation option, using falloc if available, and full otherwise.