Kubernetes Container Storage Interface (CSI)

Introduction to Kubernetes Container Storage Interface (CSI)

The Container Storage Interface (CSI) is an industry standard that enables storage vendors to develop plugins for exposing block and filesystem storage systems to containers. The goal of CSI is to provide an extensible layer to manage Kubernetes volumes.

CSI replaces earlier Kubernetes attempts to manage storage such as in-tree volume plugins and FlexVolume plugins. These plugins were hosted in the Kubernetes codebase and were shipped with the core Kubernetes binaries. This meant storage providers were forced to follow the Kubernetes release process to release bug fixes and new functionality.

CSI Architecture

The CSI standard defines a set of services to reduce coupling with storage drives and provide more flexibility. These services are deployed along side a storage driver as sidecar containers. The sidecar containers are responsible for interacting with the Kubernetes API.

Konvoy will automatically configure native storage drivers for the supported cloud provisioners (AWS, Azure, or GCP) which deploy the following sidecar containers:

Sidecar container Description
external-attacher Watches Kubernetes VolumeAttachment objects and triggers ControllerPublish and ControllerUnpublish operations against a CSI endpoint.
external-provisioner Watches Kubernetes PersistentVolumeClaim objects and triggers CreateVolume and DeleteVolume operations against a CSI endpoint.
node-driver-registrar Registers the CSI driver with the Kubelet using the Kubelet device plugin registration mechanism.
external-snapshotter Watches Kubernetes VolumeSnapshot CRD objects and triggers CreateSnapshot and DeleteSnapshot operations against a CSI endpoint.
livenessprobe Optional CSI plugin to enable the Kubernetes Liveness Probe mechanism.

Using a CSI volume with Konvoy on AWS

In Konvoy, you can use CSI volumes through the Kubernetes storage API objects like PersistentVolumeClaims (PVCs), PersistentVolumes (PVs), and StorageClasses.

Here’s an example of a StorageClass for the AWS EBS CSI driver. Save it in file named ebs-sc.yaml:

kind: StorageClass
apiVersion: storage.k8s.io/v1
  name: ebs-storage-class # The name of the StorageClass.
provisioner: ebs.csi.aws.com # The name of the AWS CSI driver.
volumeBindingMode: WaitForFirstConsumer
  csi.storage.k8s.io/fstype: xfs # The filesystem type to format the volume.
  type: io1 # The type of AWS EBS volume.
  iopsPerGB: "50"
  encrypted: "true"
- matchLabelExpressions:
  - key: topology.ebs.csi.aws.com/zone
    - us-east-1a

Next, apply the file.

kubectl apply -f ebs-sc.yaml

Ensure that it is applied.

kubectl get storageclass

The output should be similar to:

ebs-storageclass   ebs.csi.aws.com   Delete          WaitForFirstConsumer   true                   10s

This is an example of a PVC configured to use the StorageClass. Save it in a file named ebs-pvc.yaml.

apiVersion: v1
kind: PersistentVolumeClaim
  name: ebs-pvc # The name of the PersistentVolumeClaim.
    - ReadWriteOnce
  storageClassName: ebs-storage-class # The storage class that it should use.
      storage: 4Gi # The size of the volume.

Next, apply the file.

kubectl apply -f ebs-pvc.yaml

Ensure that it is applied.

kubectl get pvc

The output should be similar to:

NAMESPACE    NAME      STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS        AGE
default      ebs-pvc   Bound    pvc-c0dff51c-a8c1-48b7-97f0-01b56fabca08   30Gi       RWO            ebs-storage-class   50s

The following sample Kubernetes Pod specification has the EBS volume created by the AWS EBS CSI driver and mounted in the container:

apiVersion: v1
kind: Pod
  name: example-application
  - name: app
    image: centos
    command: ["/bin/sh"]
    args: ["-c", "while true; do echo $(date -u) >> /data/out.txt; sleep 5; done"]
    - name: persistent-storage
      mountPath: /data # The path inside the container where the volume is mounted.
  - name: persistent-storage
      claimName: ebs-claim # The PVC that it should use.