Travis Nielsen
68b66ca937
The csi images in the operator helm chart previously were a single
image name including the repository and tag in a single string.
Now the repository and tag values are separated to allow for
greater customization of the repository from the version tags.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
(cherry picked from commit 64f71d9258)
1.5 KiB
1.5 KiB
v1.14 Pending Release Notes
Breaking Changes
- The minimum supported version of Kubernetes is v1.25. Upgrade to Kubernetes v1.25 or higher before upgrading Rook.
- The Rook operator config
CSI_ENABLE_READ_AFFINITYwas removed. v1.13 clusters that have modified this value to be"true"must set the option as desired in each CephCluster as documented here before upgrading to v1.14. - Rook is beginning the process of deprecating CSI network "holder" pods.
If there are pods named
csi-*plugin-holder-*in the Rook operator namespace, see the detailed documentation to disable them. This is optional for v1.14, but will be required in a future release.
Features
- Kubernetes versions v1.25 through v1.29 are supported.
- Ceph daemon pods using the
defaultservice account now use a newrook-ceph-defaultservice account. - Allow setting the Ceph
applicationon a pool - Create object stores with shared metadata and data pools. Isolation between object stores is enabled via RADOS namespaces.
- The feature support for VolumeSnapshotGroup has been added to the RBD and CephFS CSI driver.
- Support for virtual style hosting for s3 buckets in the CephObjectStore.
- Add option to specify prefix for the OBC provisioner.
- Support Azure Key Vault for storing OSD encryption keys.
- Separate image repository and tag values in the helm chart for the CSI images.