Skip to content

Ceph CSI Drivers

There are three CSI drivers integrated with Rook that are used in different scenarios:

  • RBD: This block storage driver is optimized for RWO pod access where only one pod may access the storage. More information.
  • CephFS: This file storage driver allows for RWX with one or more pods accessing the same storage. More information.
  • NFS (experimental): This file storage driver allows creating NFS exports that can be mounted on pods, or directly via an NFS client from inside or outside the Kubernetes cluster. More information

The Ceph Filesystem (CephFS) and RADOS Block Device (RBD) drivers are enabled automatically by the Rook operator. The NFS driver is disabled by default. All drivers will be started in the same namespace as the operator when the first CephCluster CR is created.

Supported Versions

The two most recent Ceph CSI version are supported with Rook. Refer to ceph csi releases for more information.

Static Provisioning

The RBD and CephFS drivers support the creation of static PVs and static PVCs from an existing RBD image or CephFS volume/subvolume. Refer to the static PVC documentation for more information.

Configure CSI Drivers in non-default namespace

If you've deployed the Rook operator in a namespace other than rook-ceph, change the prefix in the provisioner to match the namespace you used. For example, if the Rook operator is running in the namespace my-namespace the provisioner value should be my-namespace.rbd.csi.ceph.com. The same provisioner name must be set in both the storageclass and snapshotclass.

To find the provisioner name in the example storageclasses and volumesnapshotclass, search for: # csi-provisioner-name

Configure custom Driver name prefix for CSI Drivers

To use a custom prefix for the CSI drivers instead of the namespace prefix, set the CSI_DRIVER_NAME_PREFIX environment variable in the operator configmap. For instance, to use the prefix my-prefix for the CSI drivers, set the following in the operator configmap:

kubectl patch cm rook-ceph-operator-config -n rook-ceph -p $'data:\n "CSI_DRIVER_NAME_PREFIX": "my-prefix"'

Once the configmap is updated, the CSI drivers will be deployed with the my-prefix prefix. The same prefix must be set in both the storageclass and snapshotclass. For example, to use the prefix my-prefix for the CSI drivers, update the provisioner in the storageclass:

1
2
3
4
5
6
apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: rook-ceph-block-sc
provisioner: my-prefix.rbd.csi.ceph.com
...

The same prefix must be set in the volumesnapshotclass as well:

1
2
3
4
5
6
apiVersion: snapshot.storage.k8s.io/v1
kind: VolumeSnapshotClass
metadata:
  name: rook-ceph-block-vsc
driver: my-prefix.rbd.csi.ceph.com
...

When the prefix is set, the driver names will be:

  • RBD: my-prefix.rbd.csi.ceph.com
  • CephFS: my-prefix.cephfs.csi.ceph.com
  • NFS: my-prefix.nfs.csi.ceph.com

Note

Please be careful when setting the CSI_DRIVER_NAME_PREFIX environment variable. It should be done only in fresh deployments because changing the prefix in an existing cluster will result in unexpected behavior.

To find the provisioner name in the example storageclasses and volumesnapshotclass, search for: # csi-provisioner-name

Liveness Sidecar

All CSI pods are deployed with a sidecar container that provides a Prometheus metric for tracking whether the CSI plugin is alive and running.

Check the monitoring documentation to see how to integrate CSI liveness and GRPC metrics into Ceph monitoring.

Dynamically Expand Volume

Prerequisites

To expand the PVC the controlling StorageClass must have allowVolumeExpansion set to true. csi.storage.k8s.io/controller-expand-secret-name and csi.storage.k8s.io/controller-expand-secret-namespace values set in the storageclass. Now expand the PVC by editing the PVC's pvc.spec.resource.requests.storage to a higher values than the current size. Once the PVC is expanded on the back end and the new size is reflected on the application mountpoint, the status capacity pvc.status.capacity.storage of the PVC will be updated to the new size.

RBD Mirroring

To support RBD Mirroring, the CSI-Addons sidecar will be started in the RBD provisioner pod. CSI-Addons support the VolumeReplication operation. The volume replication controller provides common and reusable APIs for storage disaster recovery. It is based on the csi-addons specification. It follows the controller pattern and provides extended APIs for storage disaster recovery. The extended APIs are provided via Custom Resource Definitions (CRDs).

Enable CSIAddons Sidecar

To enable the CSIAddons sidecar and deploy the controller, follow the steps below

Ephemeral volume support

The generic ephemeral volume feature adds support for specifying PVCs in the volumes field to create a Volume as part of the pod spec. This feature requires the GenericEphemeralVolume feature gate to be enabled.

For example:

kind: Pod
apiVersion: v1
...
  volumes:
    - name: mypvc
      ephemeral:
        volumeClaimTemplate:
          spec:
            accessModes: ["ReadWriteOnce"]
            storageClassName: "rook-ceph-block"
            resources:
              requests:
                storage: 1Gi

A volume claim template is defined inside the pod spec, and defines a volume to be provisioned and used by the pod within its lifecycle. Volumes are provisioned when a pod is spawned and destroyed when the pod is deleted.

Refer to the ephemeral-doc for more info. See example manifests for an RBD ephemeral volume and a CephFS ephemeral volume.

CSI-Addons Controller

The CSI-Addons Controller handles requests from users. Users create a CR that the controller inspects and forwards to one or more CSI-Addons sidecars for execution.

Deploying the controller

Deploy the controller by running the following commands:

1
2
3
kubectl create -f https://github.com/csi-addons/kubernetes-csi-addons/releases/download/v0.9.1/crds.yaml
kubectl create -f https://github.com/csi-addons/kubernetes-csi-addons/releases/download/v0.9.1/rbac.yaml
kubectl create -f https://github.com/csi-addons/kubernetes-csi-addons/releases/download/v0.9.1/setup-controller.yaml

This creates the required CRDs and configures permissions.

Enable the CSI-Addons Sidecar

To use the features provided by the CSI-Addons, the csi-addons containers need to be deployed in the RBD provisioner and nodeplugin pods, which are not enabled by default.

Execute the following to enable the CSI-Addons sidecars:

  • Update the rook-ceph-operator-config configmap and patch the following configuration:

    kubectl patch cm rook-ceph-operator-config -nrook-ceph -p $'data:\n "CSI_ENABLE_CSIADDONS": "true"'
    
  • After enabling CSI_ENABLE_CSIADDONS in the configmap, a new sidecar container named csi-addons will start automatically in the RBD CSI provisioner and nodeplugin pods.

CSI-Addons Operations

CSI-Addons supports the following operations:

Enable RBD and CephFS Encryption Support

Ceph-CSI supports encrypting PersistentVolumeClaims (PVCs) for both RBD and CephFS. This can be achieved using LUKS for RBD and fscrypt for CephFS. More details on encrypting RBD PVCs can be found here, which includes a full list of supported encryption configurations. More details on encrypting CephFS PVCs can be found here. A sample KMS configmap can be found here.

Note

Not all KMS are compatible with fscrypt. Generally, KMS that either store secrets to use directly (like Vault) or allow access to the plain password (like Kubernetes Secrets) are compatible.

Note

Rook also supports OSD-level encryption (see encryptedDevice option here).

Using both RBD PVC encryption and OSD encryption at the same time will lead to double encryption and may reduce read/write performance.

Existing Ceph clusters can also enable Ceph-CSI PVC encryption support and multiple kinds of encryption KMS can be used on the same Ceph cluster using different storageclasses.

The following steps demonstrate the common process for enabling encryption support for both RBD and CephFS:

  • Create the rook-ceph-csi-kms-config configmap with required encryption configuration in the same namespace where the Rook operator is deployed. An example is shown below:
apiVersion: v1
kind: ConfigMap
metadata:
  name: rook-ceph-csi-kms-config
  namespace: rook-ceph
data:
  config.json: |-
    {
      "user-secret-metadata": {
        "encryptionKMSType": "metadata",
        "secretName": "storage-encryption-secret"
      }
    }
  • Update the rook-ceph-operator-config configmap and patch the following configurations
kubectl patch cm rook-ceph-operator-config -nrook-ceph -p $'data:\n "CSI_ENABLE_ENCRYPTION": "true"'
  • Create the resources (secrets, configmaps etc) required by the encryption type. In this case, create storage-encryption-secret secret in the namespace of the PVC as follows:
1
2
3
4
5
6
7
apiVersion: v1
kind: Secret
metadata:
  name: storage-encryption-secret
  namespace: rook-ceph
stringData:
  encryptionPassphrase: test-encryption
1
2
3
4
5
6
7
8
9
apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: rook-ceph-block-encrypted
parameters:
  # additional parameters required for encryption
  encrypted: "true"
  encryptionKMSID: "user-secret-metadata"
# ...
  • PVCs created using the new storageclass will be encrypted.

Note

CephFS encryption requires fscrypt support in Linux kernel, kernel version 6.6 or higher.

Enable Read affinity for RBD and CephFS volumes

Ceph CSI supports mapping RBD volumes with KRBD options and mounting CephFS volumes with ceph mount options to allow serving reads from an OSD closest to the client, according to OSD locations defined in the CRUSH map and topology labels on nodes.

Refer to the krbd-options document for more details.

Execute the following step to enable read affinity for a specific ceph cluster:

  • Patch the ceph cluster CR to enable read affinity:
kubectl patch cephclusters.ceph.rook.io <cluster-name> -n rook-ceph -p '{"spec":{"csi":{"readAffinity":{"enabled": true}}}}'
1
2
3
  csi:
    readAffinity:
      enabled: true
  • Add topology labels to the Kubernetes nodes. The same labels may be used as mentioned in the OSD topology topic.

Ceph CSI will extract the CRUSH location from the topology labels found on the node and pass it though krbd options during mapping RBD volumes.

Note

This requires Linux kernel version 5.8 or higher.