OpenShift Virtualization and Oracle Cloud Infrastructure known issues and limitations

Updated

Oracle Cloud Infratructure support is currently available (GA) starting with OpenShift Virtualization 4.19.6. There are known issues and limitations to be aware and understand before starting a deployment.

Storage known issues and limitations

50Gi Minimum Block Volume Size

OCI enforces a minimum volume size of 50Gi for block storage, which is significantly larger than defaults on other provisioners. This significantly affects performance. General concerns:

  • Increased I/O time and slower execution for large volume operations.
  • Timeouts in workflows involving data import/export or verification steps.
  • Cost increases for users as 50Gi volumes are enforced even when requesting 10Mi PVCs for backend storage.

These are the areas that are affected the most:

Export Operations:


The Kubevirt Export API ([Export API Documentation](https://kubevirt.io/user-guide/storage/export_api/)) may perform poorly when exporting 50Gi block-mode volumes:
  • Downloads are significantly slower.
  • Export server may hang or timeout when processing large raw block data.
  • The issue appears related to internal processing overhead, not throttling or network speed.

Import Operations:


CDI ([CDI Documentation](https://kubevirt.io/user-guide/storage/containerized_data_importer/)) requires more time to import images due to slower provisioning of 50Gi volumes.

General Volume Operations:

  • Operations that scan the entire volume (e.g., md5sum) may time out or perform poorly.
  • These issues stem from the size of the underlying volume, not specific bugs in CNV or KubeVirt.
  • Volumes with hardcoded small capacity (such as vTPMs or EFIs) that only request about 10Mi, will still receive a 50Gi volume.

Recommendations:

  • Increase timeouts for performance-sensitive operations.
  • Avoid large-volume data integrity tasks in critical paths.

Volume Expansion Not Reflected in Guest (Filesystem Volumes)

When expanding filesystem-mode PVCs, the volume size is correctly updated in Kubernetes but may not appear updated inside the guest OS without manual resizing.

Impact:

  • The guest OS may not see or utilize the expanded capacity until a filesystem resize is triggered.

Notes:

  • Block-mode PVCs are not affected.
  • This is a known issue in Kubevirt and not OCI-specific, but replicates more easily with OCI due to the large block volumes.
  • This bug will be fixed in OpenShift Virtualization 4.19.2 and will be backported to earlier version later

Tracking Bug:


[CNV-58023](https://issues.redhat.com/browse/CNV-58023)

Compute known issues and limitations

  • Virtual Machine Volumes with hard-coded small capacity (such as vTPMs or EFIs) that only request about 10Mi, will still receive a 50Gi volume.
SBR
Product(s)
Category
Components
Tags
Article Type