Volume Creation Failure Due to Image Format Mismatch

Problem

The volume creation from an image fails with the following error:

Volume creation error
    
 
cinder.exception.ImageUnacceptable:Image <ID> is unacceptable: The image format was claimed to be 'raw' but the image data appears to be in a different format.
Copy

This leads to:

Volume stuck in error or downloading state.
Orphaned files on Tintri/NFS backends.
VM boot attempts failing with libvirt format errors.

Environment

Private Cloud Director Virtualization - v2025.6 and Higher
Self-Hosted Private Cloud Director Virtualization - v2025.6 and Higher
Block Storage using Tintri VMstore (NFS backend)
Image stored in local file store
Example versions:
- cinder.volume.drivers.vmstore.nfs
- Image Service metadata (disk_format=raw, container_format=bare)

Cause

A image was incorrectly labeled as raw, but its actual data content was qcow2 (or another format). Block Storage’s image_utils.check_image_format() layer detected:

Metadata says raw
Data is actually non-raw

As a result:

Image conversion was aborted
Volume remained partially created
Back-end cleanup was incomplete

Diagnostics

Check image metadata

Command
    
xxxxxxxxxx
 
openstack image show <image-id>
Copy

Verify actual image format - Must run on Image Service host

Command
    
xxxxxxxxxx
 
qemu-img info /var/lib/glance/images/glance/<IMAGE_ID>
Copy

Example Output
    
xxxxxxxxxx
 
file format: qcow2
Copy

Check Tintri logs for refresh failure

Cinder-volume-base logs
    
xxxxxxxxxx
 
ERR-API-8001: Failed to find requested resource
Copy

Check orphaned files on the block storage backend

Block Storage Host
    
xxxxxxxxxx
 
ls -lh /tintri/<TENANT>/cinder/volume-<VOLUME_UUID>
Copy

Resolution

If the desired format of the image is actually qcow2, update Image service metadata:

Command
    
xxxxxxxxxx
 
openstack image set --disk-format qcow2 <IMAGE_ID>
Copy

Clean up stale backend file - Only if the volume was stuck/errored when checking the openstack command, and the openstack volume delete <VOLUME_UUID> does not remove the volume

Block Storage Host
    
xxxxxxxxxx
 
ssh <TINTRI_MOUNT_HOST>rm /tintri/<TENANT>/cinder/volume-<VOLUME_UUID>
Copy

Ensure to delete only the stale volume which is in stuck/errored state due to disk_format mismatch, else if a valid volume is deleted it will result on data loss.

Validation

Re-upload or fix image metadata, the disk_format should now reflect actual format.

Command
    
xxxxxxxxxx
 
openstack image show <IMAGE_ID> | grep disk_format
Copy

Create test bootable volume

Command
    
xxxxxxxxxx
 
openstack volume create --image <IMAGE_ID> --size 10 test-vol
Copy

Check volume status

Command
    
xxxxxxxxxx
 
openstack volume show test-vol | grep status
Copy

Expected status:

Command Output
    
xxxxxxxxxx
 
status: available
Copy

Confirm no orphaned data: Cleanup confirmed on Tintri backend.

Additional Information

To prevent this issue in the future, you can adopt any of the following options:

Option A

Enforce qcow2 labelling by always uploading with:

Command Option
    
xxxxxxxxxx
 
--disk-format qcow2
Copy

Option B

Validate before import on image upload:

Command
    
xxxxxxxxxx
 
qemu-img info <FILE>
Copy

Option C

Enable image conversion - Use Image Service import workflow with:

Command
    
xxxxxxxxxx
 
glance-directweb-downloadcopy-image
Copy

Last updated on

Was this page helpful?