Using Container Storage Interface (CSI) - Configuring Persistent Storage | Configuring Clusters

Overview
Architecture
- External CSI Controllers
- CSI Driver DaemonSet
Example deployment
Dynamic Provisioning
Usage

Overview

Container Storage Interface (CSI) allows OpenShift Container Platform to consume storage from storage backends that implement the CSI interface as persistent storage.

CSI volumes are currently in Technology Preview and not for production workloads. CSI volumes may change in a future release of OpenShift Container Platform. Technology Preview features are not supported with Red Hat production service level agreements (SLAs), might not be functionally complete, and Red Hat does not recommend to use them for production. These features provide early access to upcoming product features, enabling customers to test functionality and provide feedback during the development process.

See the Red Hat Technology Preview features support scope for more information.

OpenShift Container Platform does not ship with any CSI drivers. It is recommended to use the CSI drivers provided by community or storage vendors.

OpenShift Container Platform 3.10 supports version 0.2.0 of the CSI specification.

Architecture

CSI drivers are typically shipped as container images. These containers are not aware of OpenShift Container Platform where they run. To use CSI-compatible storage backend in OpenShift Container Platform, the cluster administrator must deploy several components that serve as a bridge between OpenShift Container Platform and the storage driver.

The following diagram provides a high-level overview about the components running in pods in the OpenShift Container Platform cluster.

It is possible to run multiple CSI drivers for different storage backends. Each driver needs its own external controllers' deployment and DaemonSet with the driver and CSI registrar.

External CSI Controllers

External CSI Controllers is a deployment that deploys one or more pods with three containers:

External CSI attacher container that translates attach and detach calls from OpenShift Container Platform to respective ControllerPublish and ControllerUnpublish calls to CSI driver
External CSI provisioner container that translates provision and delete calls from OpenShift Container Platform to respective CreateVolume and DeleteVolume calls to CSI driver
CSI driver container

The CSI attacher and CSI provisioner containers talk to the CSI driver container using UNIX Domain Sockets, ensuring that no CSI communication leaves the pod. The CSI driver is not accessible from outside of the pod.

attach, detach, provision, and delete operations typically require the CSI driver to use credentials to the storage backend. Run the CSI controller pods on infrastructure nodes so the credentials never leak to user processes, even in the event of a catastrophic security breach on a compute node.

The external attacher must also run for CSI drivers that do not support third-party attach/detach operations. The external attacher will not issue any ControllerPublish or ControllerUnpublish operations to the CSI driver. However, it still must run to implement the necessary OpenShift Container Platform attachment API.

CSI Driver DaemonSet

Finally, the CSI driver DaemonSet runs a pod on every node that allows OpenShift Container Platform to mount storage provided by the CSI driver to the node and use it in user workloads (pods) as persistent volumes (PVs). The pod with the CSI driver installed contains the following containers:

CSI driver registrar, which registers the CSI driver into the openshift-node service running on the node. The openshift-node process running on the node then directly connects with the CSI driver using the UNIX Domain Socket available on the node.
CSI driver.

The CSI driver deployed on the node should have as few credentials to the storage backend as possible. OpenShift Container Platform will only use the node plug-in set of CSI calls such as NodePublish/NodeUnpublish and NodeStage/NodeUnstage (if implemented).

Example deployment

Since OpenShift Container Platform does not ship with any CSI driver installed, this example shows how to deploy a community driver for OpenStack Cinder in OpenShift Container Platform.

Create a new project where the CSI components will run and a new service account that will run the components. Explicit node selector is used to run the Daemonset with the CSI driver also on master nodes.

# oc adm new-project csi --node-selector=""
Now using project "csi" on server "https://example.com:8443".

# oc create serviceaccount cinder-csi
serviceaccount "cinder-csi" created

# oc adm policy add-scc-to-user privileged system:serviceaccount:csi:cinder-csi
scc "privileged" added to: ["system:serviceaccount:csi:cinder-csi"]

Apply this YAML file to create the deployment with the external CSI attacher and provisioner and DaemonSet with the CSI driver.

# This YAML file contains all API objects that are necessary to run Cinder CSI
# driver.
#
# In production, this needs to be in separate files, e.g. service account and
# role and role binding needs to be created once.
#
# It server as an example how to use external attacher and external provisioner
# images shipped with {product-title} with a community CSI driver.

kind: ClusterRole
apiVersion: rbac.authorization.k8s.io/v1
metadata:
  name: cinder-csi-role
rules:
  - apiGroups: [""]
    resources: ["persistentvolumes"]
    verbs: ["create", "delete", "get", "list", "watch", "update", "patch"]
  - apiGroups: [""]
    resources: ["events"]
    verbs: ["create", "get", "list", "watch", "update", "patch"]
  - apiGroups: [""]
    resources: ["persistentvolumeclaims"]
    verbs: ["get", "list", "watch", "update", "patch"]
  - apiGroups: [""]
    resources: ["nodes"]
    verbs: ["get", "list", "watch", "update", "patch"]
  - apiGroups: ["storage.k8s.io"]
    resources: ["storageclasses"]
    verbs: ["get", "list", "watch"]
  - apiGroups: ["storage.k8s.io"]
    resources: ["volumeattachments"]
    verbs: ["get", "list", "watch", "update", "patch"]
  - apiGroups: [""]
    resources: ["configmaps"]
    verbs: ["get", "list", "watch", "create", "update", "patch"]

---

kind: ClusterRoleBinding
apiVersion: rbac.authorization.k8s.io/v1
metadata:
  name: cinder-csi-role
subjects:
  - kind: ServiceAccount
    name: cinder-csi
    namespace: csi
roleRef:
  kind: ClusterRole
  name: cinder-csi-role
  apiGroup: rbac.authorization.k8s.io

---
apiVersion: v1
data:
  cloud.conf: W0dsb2JhbF0KYXV0aC11cmwgPSBodHRwczovL2V4YW1wbGUuY29tOjEzMDAwL3YyLjAvCnVzZXJuYW1lID0gYWxhZGRpbgpwYXNzd29yZCA9IG9wZW5zZXNhbWUKdGVuYW50LWlkID0gZTBmYTg1YjZhMDY0NDM5NTlkMmQzYjQ5NzE3NGJlZDYKcmVnaW9uID0gcmVnaW9uT25lCg== (1)
kind: Secret
metadata:
  creationTimestamp: null
  name: cloudconfig
---
kind: deployment
apiVersion: apps/v1
metadata:
  name: cinder-csi-controller
spec:
  replicas: 2
  selector:
    matchLabels:
      app: cinder-csi-controllers
  template:
    metadata:
      labels:
        app: cinder-csi-controllers
    spec:
      serviceAccount: cinder-csi
      containers:
        - name: csi-attacher
          image: registry.access.redhat.com/openshift3/csi-attacher:v3.10
          args:
            - "--v=5"
            - "--csi-address=$(ADDRESS)"
            - "--leader-election"
            - "--leader-election-namespace=$(MY_NAMESPACE)"
            - "--leader-election-identity=$(MY_NAME)"
          env:
            - name: MY_NAME
              valueFrom:
                fieldRef:
                  fieldPath: metadata.name
            - name: MY_NAMESPACE
              valueFrom:
                fieldRef:
                  fieldPath: metadata.namespace
            - name: ADDRESS
              value: /csi/csi.sock
          volumeMounts:
            - name: socket-dir
              mountPath: /csi
        - name: csi-provisioner
          image: registry.access.redhat.com/openshift3/csi-provisioner:v3.10
          args:
            - "--v=5"
            - "--provisioner=csi-cinderplugin"
            - "--csi-address=$(ADDRESS)"
          env:
            - name: ADDRESS
              value: /csi/csi.sock
          volumeMounts:
            - name: socket-dir
              mountPath: /csi
        - name: cinder-driver
          image: quay.io/jsafrane/cinder-csi-plugin
          command: [ "/bin/cinder-csi-plugin" ]
          args:
            - "--nodeid=$(NODEID)"
            - "--endpoint=unix://$(ADDRESS)"
            - "--cloud-config=/etc/cloudconfig/cloud.conf"
          env:
            - name: NODEID
              valueFrom:
                fieldRef:
                  fieldPath: spec.nodeName
            - name: ADDRESS
              value: /csi/csi.sock
          volumeMounts:
            - name: socket-dir
              mountPath: /csi
            - name: cloudconfig
              mountPath: /etc/cloudconfig
      volumes:
        - name: socket-dir
          emptyDir:
        - name: cloudconfig
          secret:
            secretName: cloudconfig

---

kind: DaemonSet
apiVersion: apps/v1
metadata:
  name: cinder-csi-ds
spec:
  selector:
    matchLabels:
      app: cinder-csi-driver
  template:
    metadata:
      labels:
        app: cinder-csi-driver
    spec:
      nodeSelector:
          role: node
      serviceAccount: cinder-csi
      containers:
        - name: csi-driver-registrar
          image: registry.access.redhat.com/openshift3/csi-driver-registrar:v3.10
          securityContext:
            privileged: true
          args:
            - "--v=5"
            - "--csi-address=$(ADDRESS)"
          env:
            - name: ADDRESS
              value: /csi/csi.sock
            - name: KUBE_NODE_NAME
              valueFrom:
                fieldRef:
                  fieldPath: spec.nodeName
          volumeMounts:
            - name: socket-dir
              mountPath: /csi
        - name: cinder-driver
          securityContext:
            privileged: true
            capabilities:
              add: ["SYS_ADMIN"]
            allowPrivilegeEscalation: true
          image: quay.io/jsafrane/cinder-csi-plugin
          command: [ "/bin/cinder-csi-plugin" ]
          args:
            - "--nodeid=$(NODEID)"
            - "--endpoint=unix://$(ADDRESS)"
            - "--cloud-config=/etc/cloudconfig/cloud.conf"
          env:
            - name: NODEID
              valueFrom:
                fieldRef:
                  fieldPath: spec.nodeName
            - name: ADDRESS
              value: /csi/csi.sock
          volumeMounts:
            - name: socket-dir
              mountPath: /csi
            - name: cloudconfig
              mountPath: /etc/cloudconfig
            - name: mountpoint-dir
              mountPath: /var/lib/origin/openshift.local.volumes/pods/
              mountPropagation: "Bidirectional"
            - name: cloud-metadata
              mountPath: /var/lib/cloud/data/
            - name: dev
              mountPath: /dev
      volumes:
        - name: cloud-metadata
          hostPath:
            path: /var/lib/cloud/data/
        - name: socket-dir
          hostPath:
            path: /var/lib/kubelet/plugins/csi-cinderplugin
            type: DirectoryOrCreate
        - name: mountpoint-dir
          hostPath:
            path: /var/lib/origin/openshift.local.volumes/pods/
            type: Directory
        - name: cloudconfig
          secret:
            secretName: cloudconfig
        - name: dev
          hostPath:
            path: /dev

1	Replace with `cloud.conf` for your OpenStack deployment, as described in OpenStack configuration. For example, the Secret can be generated using the `oc create secret generic cloudconfig --from-file cloud.conf --dry-run -o yaml`.

Dynamic Provisioning

Dynamic provisioning of persistent storage depends on the capabilities of the CSI driver and underlying storage backend. The provider of the CSI driver should document how to create a StorageClass in OpenShift Container Platform and the parameters available for configuration.

As seen in the OpenStack Cinder example, you can deploy this StorageClass to enable dynamic provisioning. The following example creates a new default storage class that ensures that all PVCs that do not require any special storage class are provisioned by the installed CSI driver:

# oc create -f - << EOF
apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: cinder
  annotations:
    storageclass.kubernetes.io/is-default-class: "true"
provisioner: csi-cinderplugin
parameters:
EOF

Usage

Once the CSI driver is deployed and the StorageClass for dynamic provisioning is created, OpenShift Container Platform is ready to use CSI. The following example installs a default MySQL template without any changes to the template:

# oc new-app mysql-persistent
--> Deploying template "openshift/mysql-persistent" to project default
...

# oc get pvc
NAME              STATUS    VOLUME                                   CAPACITY   ACCESS MODES   STORAGECLASS   AGE
mysql             Bound     kubernetes-dynamic-pv-3271ffcb4e1811e8   1Gi        RWO            cinder         3s

Persistent Storage Using Container Storage Interface (CSI)