OpenEBS 4.2 Release Notes

Release Summary

OpenEBS version 4.2 introduces several functional fixes and new features focused on improving high availability (HA), replica rebuilds, and overall stability. A key highlight is Mayastor's new support for RDMA as a transport protocol for volume targets over NVMe over Fabrics (NVMeoF). In addition, the release includes various usability and functional fixes for both the LocalPV ZFS and LocalPV Hostpath provisioners, along with documentation enhancements to help users and new contributors get started quickly.

Change Summary

Local Storage

• ZFS Enhancements:
  • Support for the zstd-fast compression algorithm in LocalPV ZFS.
  • Flexibility to choose between refquota or quota for ZFS volumes.
  • Fixes in volume provisioning, controller behavior, reservation, and deployment.
• LocalPV Hostpath Improvements:
  • Eviction tolerations added to the Hostpath provisioner deployment.
  • Helper pod now allowed to use the host network.
  • Added a global imageRegistry Helm parameter for specifying an image registry for all images.

LocalPV Release Notes

• LocalPV-ZFS:
  Refer to the LocalPV-ZFS v2.7.0 release for detailed changes.
• LocalPV Hostpath:
  Refer to the LocalPV Hostpath v4.2.0 release for detailed changes.

Known Issues for LocalPV

• Controller Pod Restart on Single Node Setups:
  After upgrading, single node setups may face issues where the ZFS-localpv/LVM-localpv controller pod does not enter the Running state due to changes in the controller manifest (now a Deployment) and missing affinity rules.
  • Workaround: Delete the old controller pod to allow the new pod to be scheduled correctly.

Replicated Storage (Mayastor)

• NVMe-oF RDMA Support:
  • Mayastor volume targets can now be shared over RDMA transport, providing high throughput and reduced latency.
  • Enabled via a Helm chart option that works in conjunction with specifying an RDMA-capable network interface.
• CSAL ftl bdev Support:
  • Integration of the SPDK ftl bdev (Cloud Storage Acceleration Layer) to support layered devices with a fast cache for buffering writes.
• Additional Enhancements:
  • Detection of the builtin nvme-tcp module on Talos-like platforms.
  • Stability fixes in volume target failover and application move scenarios.
  • Eviction tolerations added to the DSP operator deployment and CSI-controller for faster failover.

Upgrade and Backward Incompatibilities

• Kubernetes Requirement: Kubernetes 1.23 or higher is recommended.
• Engine Compatibility: Upgrades to OpenEBS 4.2.0 are supported only for the following engines:
  • Local PV Hostpath
  • Local PV LVM
  • Local PV ZFS
  • Mayastor (from earlier editions, 3.10.x or below)

Replicated PV Mayastor Release Notes

• Mayastor
  Refer to the Mayastor v2.8.0 release for detailed changes.

Known Behavioural Limitations

• The Mayastor IO engine fully utilizes allocated CPU cores regardless of I/O load, running a poller at full speed.
• A Mayastor DiskPool is limited to a single block device and cannot span multiple block devices.

Known Issues

Replicated PV (Mayastor)

• DiskPool Capacity Expansion:
  Mayastor does not support the capacity expansion of DiskPools as of v2.8.0.
• IO-Engine Pod Restarts:
  Under heavy I/O and during constant scaling up/down of volume replicas, the io-engine pod may restart occasionally.
• fsfreeze Operation Failure:
  If a pod-based workload is scheduled on a node that reboots and the pod lacks a controller (such as a Deployment or StatefulSet), the volume unpublish operation might not be triggered. This leads the control plane to assume the volume is still published, causing the fsfreeze operation to fail during snapshot creation.
  • Workaround: Recreate or reinstate the pod to ensure proper volume mounting.