RATATOSKRATATOSK
Sign in

Rook

v1.21.0Storage & Data
Oct 6, 2026

CHECK 4OTHER 6

A breaking release changes Helm chart and Ceph defaults and removes automatic setup for some resources and images. It also adds storage and object-management capabilities, so chart references, kernel compatibility, new NFS installs, and custom CSI image settings may need review.

Check if affected (4)

  • breakingHelm OCI chart tag format

    Applies if you reference Helm OCI chart tags in scripts or tooling.

    Helm OCI chart tags no longer include the v prefix, so 1.21.0 replaces v1.21.0. Scripts and tooling that reference charts by tag need to use the new format.

  • breakingCeph msgrv2 default requirement

    Applies if you cannot use msgrv2 because of an older kernel.

    Ceph msgrv2 is required by default and requires kernel 5.11. The CephCluster setting network.connections.requireMsgr2: false, or the same value under cephClusterSpec in Helm values, disables the protocol.

  • breakingAutomatic .nfs pool creation

    Applies if you are creating a new NFS install.

    New NFS installs must define a CephBlockPool for the .nfs pool because Rook no longer creates it automatically. A sample specification is available in deploy/examples/nfs.yaml; upgraded clusters are not affected.

  • breakingDefault CSI driver image configuration

    Applies if you use custom CSI images.

    Rook no longer sets default CSI driver container images. The rook-csi-operator-image-set-configmap now has empty values so the ceph-csi-operator uses its built-in defaults; custom image overrides must remain present in the configmap after an upgrade.

All 6 other recorded changesadditions 6

additions (6)

  • - RBD Quality of Service (QoS) support is configured via the VolumeAttributesClass using the krbd mounter with cgroup v2 io.max enforcement. See the [RBD QoS documentation](Documentation/Storage-Configuration/Block-Storage-RBD/rbd-qos.md) for details.
  • - Automated OSD replacement: An OSD deployment can be annotated to mark it for replacement. Rook will drain and destroy it by preserving its CRUSH position to later reuse it when new device will be available on the same node. All types of OSDs supported for host-based clusters, including OSDs sharing a metadata device. PVC-based OSDs are not supported for this new replacement design. See [OSD replacement design document](./design/ceph/osd-replacement.md) for details.
  • - The rook-ceph-cluster Helm chart can create CephObjectStoreUser resources via the new cephObjectStoreUsers value.
  • - The toolbox deployments from the Helm chart and the example manifests now reload the keyring and ceph.conf automatically after CephX key rotation, mon failover, or a config override change.
  • - CephCluster dashboard TLS certificates can now be configured from a same-namespace Kubernetes TLS Secret with spec.dashboard.sslCertificateRef when the dashboard SSL is enabled. Rook reconciles updates to the referenced Secret and restores the default self-signed certificate when the reference is removed.
  • - CephObjectStoreUser gained settings spec.defaultPlacement and spec.defaultStorageClass, which set the RGW user's default bucket placement target and default storage class. Both are optional and validated by RGW. The effective values are reported in status.info. Removing either field stops Rook from managing it and leaves the last applied value in place on the RGW user, except that changing defaultPlacement without a defaultStorageClass resets the storage class to the new placement target's default.
Add Rook to your stack

A weekly email arrives when a release needs action. Like the breaking changes in this release.

Add to stack →
Official release note ↗All Rook releases →