Ceph CSI Drivers
There are three CSI drivers integrated with Rook that will enable 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 to pods, or the exports can be mounted 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 with 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 supported Ceph CSI version is 3.3.0 or greater with Rook. Refer to ceph csi releases for more information.
Static Provisioning¶
Both drivers also support the creation of static PV and static PVC from existing RBD image/CephFS volume. Refer to static PVC 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 needs to be set in both the storageclass and snapshotclass.
Liveness Sidecar¶
All CSI pods are deployed with a sidecar container that provides a prometheus metric for tracking if the CSI plugin is alive and running. These metrics are meant to be collected by prometheus but can be accesses through a GET request to a specific node ip. for example curl -X get http://[pod ip]:[liveness-port][liveness-path] 2>/dev/null | grep csi
the expected output should be
Check the monitoring doc to see how to integrate CSI liveness and grpc metrics into ceph monitoring.
Dynamically Expand Volume¶
Prerequisite¶
- For filesystem resize to be supported for your Kubernetes cluster, the kubernetes version running in your cluster should be >= v1.15 and for block volume resize support the Kubernetes version should be >= v1.16. Also,
ExpandCSIVolumes
feature gate has to be enabled for the volume resize functionality to work.
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 storageclass. Now expand the PVC by editing the PVC pvc.spec.resource.requests.storage
to a higher values than the current size. Once PVC is expanded on backend and same is reflected size is reflected on application mountpoint, the status capacity pvc.status.capacity.storage
of PVC will be updated to new size.
RBD Mirroring¶
To support RBD Mirroring, the CSI-Addons sidecar will be started in the RBD provisioner pod. The CSI-Addons supports the VolumeReplication operation. The volume replication controller provides common and reusable APIs for storage disaster recovery. It is based on csi-addons/spec specification and can be used by any storage provider. It follows the controller pattern and provides extended APIs for storage disaster recovery. The extended APIs are provided via Custom Resource Definitions (CRDs).
Prerequisites¶
Kubernetes version 1.21 or greater is required.
Enable CSIAddons Sidecar¶
To enable the CSIAddons sidecar and deploy the controller, Please follow the steps below
Ephemeral volume support¶
The generic ephemeral volume feature adds support for specifying PVCs in the volumes
field to indicate a user would like to create a Volume as part of the pod spec. This feature requires the GenericEphemeralVolume feature gate to be enabled.
For example:
A volume claim template is defined inside the pod spec which refers to a volume provisioned and used by the pod with its lifecycle. The volumes are provisioned when pod get spawned and destroyed at time of pod delete.
Refer to ephemeral-doc for more info. Also, See the example manifests for an RBD ephemeral volume and a CephFS ephemeral volume.
CSI-Addons Controller¶
The CSI-Addons Controller handles the requests from users to initiate an operation. Users create a CR that the controller inspects, and forwards a request to one or more CSI-Addons side-cars for execution.
Deploying the controller¶
Users can deploy the controller by running the following commands:
This creates the required crds and configure 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 command in the cluster to enable the CSI-Addons sidecar:
- Update the
rook-ceph-operator-config
configmap and patch the following configurations
- After enabling
CSI_ENABLE_CSIADDONS
in the configmap, a new sidecar container with namecsi-addons
should now start automatically in the RBD CSI provisioner and nodeplugin pods.
Note
Make sure the version of ceph-csi used is v3.5.0+
.
CSI-ADDONS Operation¶
CSI-Addons supports the following operations:
- Reclaim Space
- Creating a ReclaimSpaceJob
- Creating a ReclaimSpaceCronJob
- Annotating PersistentVolumeClaims
- Network Fencing
- Creating a NetworkFence
- Volume Replication
- Creating VolumeReplicationClass
- Creating VolumeReplication CR
Enable RBD Encryption Support¶
Ceph-CSI supports encrypting individual RBD PersistentVolumeClaim with LUKS encryption. More details can be found here with full list of supported encryption configurations. A sample configmap can be found here.
Note
Rook also supports OSD encryption (see encryptedDevice
option here).
Using both RBD PVC encryption and OSD encryption together will lead to double encryption and may reduce read/write performance.
Unlike OSD encryption, existing ceph clusters can also enable Ceph-CSI RBD PVC encryption support and multiple kinds of encryption KMS can be used on the same ceph cluster using different storageclasses.
Following steps demonstrate how to enable support for encryption:
- 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:
- Update the
rook-ceph-operator-config
configmap and patch the following configurations
- Create necessary resources (secrets, configmaps etc) as required by the encryption type. In this case, create
storage-encryption-secret
secret in the namespace of pvc as shown:
- Create a new storageclass with additional parameters
encrypted: "true"
andencryptionKMSID: "<key used in configmap>"
. An example is show below:
- PVCs created using the new storageclass will be encrypted.