Storage

• 1: Volumes
• 2: Persistent Volumes
• 3: Projected Volumes
• 4: Ephemeral Volumes
• 5: Storage Classes
• 6: Dynamic Volume Provisioning
• 7: Volume Snapshots
• 8: Volume Snapshot Classes
• 9: CSI Volume Cloning
• 10: Storage Capacity
• 11: Node-specific Volume Limits
• 12: Volume Health Monitoring
• 13: Windows Storage

1 - Volumes

On-disk files in a container are ephemeral, which presents some problems for non-trivial applications when running in containers. One problem occurs when a container crashes or is stopped. Container state is not saved so all of the files that were created or modified during the lifetime of the container are lost. During a crash, kubelet restarts the container with a clean state. Another problem occurs when multiple containers are running in a Pod and need to share files. It can be challenging to setup and access a shared filesystem across all of the containers. The Kubernetes volume abstraction solves both of these problems. Familiarity with Pods is suggested.

Background

Kubernetes supports many types of volumes. A Pod can use any number of volume types simultaneously. Ephemeral volume types have a lifetime of a pod, but persistent volumes exist beyond the lifetime of a pod. When a pod ceases to exist, Kubernetes destroys ephemeral volumes; however, Kubernetes does not destroy persistent volumes. For any kind of volume in a given pod, data is preserved across container restarts.

At its core, a volume is a directory, possibly with some data in it, which is accessible to the containers in a pod. How that directory comes to be, the medium that backs it, and the contents of it are determined by the particular volume type used.

To use a volume, specify the volumes to provide for the Pod in .spec.volumes and declare where to mount those volumes into containers in .spec.containers[*].volumeMounts. A process in a container sees a filesystem view composed from the initial contents of the container image, plus volumes (if defined) mounted inside the container. The process sees a root filesystem that initially matches the contents of the container image. Any writes to within that filesystem hierarchy, if allowed, affect what that process views when it performs a subsequent filesystem access. Volumes mount at the specified paths within the image. For each container defined within a Pod, you must independently specify where to mount each volume that the container uses.

Volumes cannot mount within other volumes (but see Using subPath for a related mechanism). Also, a volume cannot contain a hard link to anything in a different volume.

Types of volumes

Kubernetes supports several types of volumes.

awsElasticBlockStore (removed)

Kubernetes 1.28 does not include a awsElasticBlockStore volume type.

The AWSElasticBlockStore in-tree storage driver was deprecated in the Kubernetes v1.19 release and then removed entirely in the v1.27 release.

The Kubernetes project suggests that you use the AWS EBS third party storage driver instead.

azureDisk (removed)

Kubernetes 1.28 does not include a azureDisk volume type.

The AzureDisk in-tree storage driver was deprecated in the Kubernetes v1.19 release and then removed entirely in the v1.27 release.

The Kubernetes project suggests that you use the Azure Disk third party storage driver instead.

azureFile (deprecated)

The azureFile volume type mounts a Microsoft Azure File volume (SMB 2.1 and 3.0) into a pod.

For more details, see the azureFile volume plugin.

azureFile CSI migration

The CSIMigration feature for azureFile, when enabled, redirects all plugin operations from the existing in-tree plugin to the file.csi.azure.com Container Storage Interface (CSI) Driver. In order to use this feature, the Azure File CSI Driver must be installed on the cluster and the CSIMigrationAzureFile feature gates must be enabled.

Azure File CSI driver does not support using same volume with different fsgroups. If CSIMigrationAzureFile is enabled, using same volume with different fsgroups won't be supported at all.

azureFile CSI migration complete

To disable the azureFile storage plugin from being loaded by the controller manager and the kubelet, set the InTreePluginAzureFileUnregister flag to true.

cephfs

A cephfs volume allows an existing CephFS volume to be mounted into your Pod. Unlike emptyDir, which is erased when a pod is removed, the contents of a cephfs volume are preserved and the volume is merely unmounted. This means that a cephfs volume can be pre-populated with data, and that data can be shared between pods. The cephfs volume can be mounted by multiple writers simultaneously.

See the CephFS example for more details.

cinder (removed)

Kubernetes 1.28 does not include a cinder volume type.

The OpenStack Cinder in-tree storage driver was deprecated in the Kubernetes v1.11 release and then removed entirely in the v1.26 release.

The Kubernetes project suggests that you use the OpenStack Cinder third party storage driver instead.

configMap

A ConfigMap provides a way to inject configuration data into pods. The data stored in a ConfigMap can be referenced in a volume of type configMap and then consumed by containerized applications running in a pod.

When referencing a ConfigMap, you provide the name of the ConfigMap in the volume. You can customize the path to use for a specific entry in the ConfigMap. The following configuration shows how to mount the log-config ConfigMap onto a Pod called configmap-pod:

apiVersion: v1
kind: Pod
metadata:
  name: configmap-pod
spec:
  containers:
    - name: test
      image: busybox:1.28
      command: ['sh', '-c', 'echo "The app is running!" && tail -f /dev/null']
      volumeMounts:
        - name: config-vol
          mountPath: /etc/config
  volumes:
    - name: config-vol
      configMap:
        name: log-config
        items:
          - key: log_level
            path: log_level

The log-config ConfigMap is mounted as a volume, and all contents stored in its log_level entry are mounted into the Pod at path /etc/config/log_level. Note that this path is derived from the volume's mountPath and the path keyed with log_level.

downwardAPI

A downwardAPI volume makes downward API data available to applications. Within the volume, you can find the exposed data as read-only files in plain text format.

See Expose Pod Information to Containers Through Files to learn more.

emptyDir

For a Pod that defines an emptyDir volume, the volume is created when the Pod is assigned to a node. As the name says, the emptyDir volume is initially empty. All containers in the Pod can read and write the same files in the emptyDir volume, though that volume can be mounted at the same or different paths in each container. When a Pod is removed from a node for any reason, the data in the emptyDir is deleted permanently.

Some uses for an emptyDir are:

• scratch space, such as for a disk-based merge sort
• checkpointing a long computation for recovery from crashes
• holding files that a content-manager container fetches while a webserver container serves the data

The emptyDir.medium field controls where emptyDir volumes are stored. By default emptyDir volumes are stored on whatever medium that backs the node such as disk, SSD, or network storage, depending on your environment. If you set the emptyDir.medium field to "Memory", Kubernetes mounts a tmpfs (RAM-backed filesystem) for you instead. While tmpfs is very fast be aware that, unlike disks, files you write count against the memory limit of the container that wrote them.

A size limit can be specified for the default medium, which limits the capacity of the emptyDir volume. The storage is allocated from node ephemeral storage. If that is filled up from another source (for example, log files or image overlays), the emptyDir may run out of capacity before this limit.

emptyDir configuration example

apiVersion: v1
kind: Pod
metadata:
  name: test-pd
spec:
  containers:
  - image: registry.k8s.io/test-webserver
    name: test-container
    volumeMounts:
    - mountPath: /cache
      name: cache-volume
  volumes:
  - name: cache-volume
    emptyDir:
      sizeLimit: 500Mi

fc (fibre channel)

An fc volume type allows an existing fibre channel block storage volume to mount in a Pod. You can specify single or multiple target world wide names (WWNs) using the parameter targetWWNs in your Volume configuration. If multiple WWNs are specified, targetWWNs expect that those WWNs are from multi-path connections.

See the fibre channel example for more details.

gcePersistentDisk (deprecated)

A gcePersistentDisk volume mounts a Google Compute Engine (GCE) persistent disk (PD) into your Pod. Unlike emptyDir, which is erased when a pod is removed, the contents of a PD are preserved and the volume is merely unmounted. This means that a PD can be pre-populated with data, and that data can be shared between pods.

There are some restrictions when using a gcePersistentDisk:

• the nodes on which Pods are running must be GCE VMs
• those VMs need to be in the same GCE project and zone as the persistent disk

One feature of GCE persistent disk is concurrent read-only access to a persistent disk. A gcePersistentDisk volume permits multiple consumers to simultaneously mount a persistent disk as read-only. This means that you can pre-populate a PD with your dataset and then serve it in parallel from as many Pods as you need. Unfortunately, PDs can only be mounted by a single consumer in read-write mode. Simultaneous writers are not allowed.

Using a GCE persistent disk with a Pod controlled by a ReplicaSet will fail unless the PD is read-only or the replica count is 0 or 1.

Creating a GCE persistent disk

Before you can use a GCE persistent disk with a Pod, you need to create it.

gcloud compute disks create --size=500GB --zone=us-central1-a my-data-disk

GCE persistent disk configuration example

apiVersion: v1
kind: Pod
metadata:
  name: test-pd
spec:
  containers:
  - image: registry.k8s.io/test-webserver
    name: test-container
    volumeMounts:
    - mountPath: /test-pd
      name: test-volume
  volumes:
  - name: test-volume
    # This GCE PD must already exist.
    gcePersistentDisk:
      pdName: my-data-disk
      fsType: ext4

Regional persistent disks

The Regional persistent disks feature allows the creation of persistent disks that are available in two zones within the same region. In order to use this feature, the volume must be provisioned as a PersistentVolume; referencing the volume directly from a pod is not supported.

Manually provisioning a Regional PD PersistentVolume

Dynamic provisioning is possible using a StorageClass for GCE PD. Before creating a PersistentVolume, you must create the persistent disk:

gcloud compute disks create --size=500GB my-data-disk
  --region us-central1
  --replica-zones us-central1-a,us-central1-b

Regional persistent disk configuration example

apiVersion: v1
kind: PersistentVolume
metadata:
  name: test-volume
spec:
  capacity:
    storage: 400Gi
  accessModes:
  - ReadWriteOnce
  gcePersistentDisk:
    pdName: my-data-disk
    fsType: ext4
  nodeAffinity:
    required:
      nodeSelectorTerms:
      - matchExpressions:
        # failure-domain.beta.kubernetes.io/zone should be used prior to 1.21
        - key: topology.kubernetes.io/zone
          operator: In
          values:
          - us-central1-a
          - us-central1-b

GCE CSI migration

The CSIMigration feature for GCE PD, when enabled, redirects all plugin operations from the existing in-tree plugin to the pd.csi.storage.gke.io Container Storage Interface (CSI) Driver. In order to use this feature, the GCE PD CSI Driver must be installed on the cluster.

GCE CSI migration complete

To disable the gcePersistentDisk storage plugin from being loaded by the controller manager and the kubelet, set the InTreePluginGCEUnregister flag to true.

gitRepo (deprecated)

A gitRepo volume is an example of a volume plugin. This plugin mounts an empty directory and clones a git repository into this directory for your Pod to use.

Here is an example of a gitRepo volume:

apiVersion: v1
kind: Pod
metadata:
  name: server
spec:
  containers:
  - image: nginx
    name: nginx
    volumeMounts:
    - mountPath: /mypath
      name: git-volume
  volumes:
  - name: git-volume
    gitRepo:
      repository: "git@somewhere:me/my-git-repository.git"
      revision: "22f1d8406d464b0c0874075539c1f2e96c253775"

glusterfs (removed)

Kubernetes 1.28 does not include a glusterfs volume type.

The GlusterFS in-tree storage driver was deprecated in the Kubernetes v1.25 release and then removed entirely in the v1.26 release.

hostPath

A hostPath volume mounts a file or directory from the host node's filesystem into your Pod. This is not something that most Pods will need, but it offers a powerful escape hatch for some applications.

For example, some uses for a hostPath are:

• running a container that needs access to Docker internals; use a hostPath of /var/lib/docker
• running cAdvisor in a container; use a hostPath of /sys
• allowing a Pod to specify whether a given hostPath should exist prior to the Pod running, whether it should be created, and what it should exist as

In addition to the required path property, you can optionally specify a type for a hostPath volume.

The supported values for field type are:

Watch out when using this type of volume, because:

• HostPaths can expose privileged system credentials (such as for the Kubelet) or privileged APIs (such as container runtime socket), which can be used for container escape or to attack other parts of the cluster.
• Pods with identical configuration (such as created from a PodTemplate) may behave differently on different nodes due to different files on the nodes
• The files or directories created on the underlying hosts are only writable by root. You either need to run your process as root in a privileged Container or modify the file permissions on the host to be able to write to a hostPath volume

hostPath configuration example

apiVersion: v1
kind: Pod
metadata:
  name: test-pd
spec:
  containers:
  - image: registry.k8s.io/test-webserver
    name: test-container
    volumeMounts:
    - mountPath: /test-pd
      name: test-volume
  volumes:
  - name: test-volume
    hostPath:
      # directory location on host
      path: /data
      # this field is optional
      type: Directory

hostPath FileOrCreate configuration example

apiVersion: v1
kind: Pod
metadata:
  name: test-webserver
spec:
  containers:
  - name: test-webserver
    image: registry.k8s.io/test-webserver:latest
    volumeMounts:
    - mountPath: /var/local/aaa
      name: mydir
    - mountPath: /var/local/aaa/1.txt
      name: myfile
  volumes:
  - name: mydir
    hostPath:
      # Ensure the file directory is created.
      path: /var/local/aaa
      type: DirectoryOrCreate
  - name: myfile
    hostPath:
      path: /var/local/aaa/1.txt
      type: FileOrCreate

iscsi

An iscsi volume allows an existing iSCSI (SCSI over IP) volume to be mounted into your Pod. Unlike emptyDir, which is erased when a Pod is removed, the contents of an iscsi volume are preserved and the volume is merely unmounted. This means that an iscsi volume can be pre-populated with data, and that data can be shared between pods.

A feature of iSCSI is that it can be mounted as read-only by multiple consumers simultaneously. This means that you can pre-populate a volume with your dataset and then serve it in parallel from as many Pods as you need. Unfortunately, iSCSI volumes can only be mounted by a single consumer in read-write mode. Simultaneous writers are not allowed.

See the iSCSI example for more details.

local

A local volume represents a mounted local storage device such as a disk, partition or directory.

Local volumes can only be used as a statically created PersistentVolume. Dynamic provisioning is not supported.

Compared to hostPath volumes, local volumes are used in a durable and portable manner without manually scheduling pods to nodes. The system is aware of the volume's node constraints by looking at the node affinity on the PersistentVolume.

However, local volumes are subject to the availability of the underlying node and are not suitable for all applications. If a node becomes unhealthy, then the local volume becomes inaccessible by the pod. The pod using this volume is unable to run. Applications using local volumes must be able to tolerate this reduced availability, as well as potential data loss, depending on the durability characteristics of the underlying disk.

The following example shows a PersistentVolume using a local volume and nodeAffinity:

apiVersion: v1
kind: PersistentVolume
metadata:
  name: example-pv
spec:
  capacity:
    storage: 100Gi
  volumeMode: Filesystem
  accessModes:
  - ReadWriteOnce
  persistentVolumeReclaimPolicy: Delete
  storageClassName: local-storage
  local:
    path: /mnt/disks/ssd1
  nodeAffinity:
    required:
      nodeSelectorTerms:
      - matchExpressions:
        - key: kubernetes.io/hostname
          operator: In
          values:
          - example-node

You must set a PersistentVolume nodeAffinity when using local volumes. The Kubernetes scheduler uses the PersistentVolume nodeAffinity to schedule these Pods to the correct node.

PersistentVolume volumeMode can be set to "Block" (instead of the default value "Filesystem") to expose the local volume as a raw block device.

When using local volumes, it is recommended to create a StorageClass with volumeBindingMode set to WaitForFirstConsumer. For more details, see the local StorageClass example. Delaying volume binding ensures that the PersistentVolumeClaim binding decision will also be evaluated with any other node constraints the Pod may have, such as node resource requirements, node selectors, Pod affinity, and Pod anti-affinity.

An external static provisioner can be run separately for improved management of the local volume lifecycle. Note that this provisioner does not support dynamic provisioning yet. For an example on how to run an external local provisioner, see the local volume provisioner user guide.

nfs

An nfs volume allows an existing NFS (Network File System) share to be mounted into a Pod. Unlike emptyDir, which is erased when a Pod is removed, the contents of an nfs volume are preserved and the volume is merely unmounted. This means that an NFS volume can be pre-populated with data, and that data can be shared between pods. NFS can be mounted by multiple writers simultaneously.

apiVersion: v1
kind: Pod
metadata:
  name: test-pd
spec:
  containers:
  - image: registry.k8s.io/test-webserver
    name: test-container
    volumeMounts:
    - mountPath: /my-nfs-data
      name: test-volume
  volumes:
  - name: test-volume
    nfs:
      server: my-nfs-server.example.com
      path: /my-nfs-volume
      readOnly: true

See the NFS example for an example of mounting NFS volumes with PersistentVolumes.

persistentVolumeClaim

A persistentVolumeClaim volume is used to mount a PersistentVolume into a Pod. PersistentVolumeClaims are a way for users to "claim" durable storage (such as a GCE PersistentDisk or an iSCSI volume) without knowing the details of the particular cloud environment.

See the information about PersistentVolumes for more details.

portworxVolume (deprecated)

A portworxVolume is an elastic block storage layer that runs hyperconverged with Kubernetes. Portworx fingerprints storage in a server, tiers based on capabilities, and aggregates capacity across multiple servers. Portworx runs in-guest in virtual machines or on bare metal Linux nodes.

A portworxVolume can be dynamically created through Kubernetes or it can also be pre-provisioned and referenced inside a Pod. Here is an example Pod referencing a pre-provisioned Portworx volume:

apiVersion: v1
kind: Pod
metadata:
  name: test-portworx-volume-pod
spec:
  containers:
  - image: registry.k8s.io/test-webserver
    name: test-container
    volumeMounts:
    - mountPath: /mnt
      name: pxvol
  volumes:
  - name: pxvol
    # This Portworx volume must already exist.
    portworxVolume:
      volumeID: "pxvol"
      fsType: "<fs-type>"

For more details, see the Portworx volume examples.

Portworx CSI migration

The CSIMigration feature for Portworx has been added but disabled by default in Kubernetes 1.23 since it's in alpha state. It has been beta now since v1.25 but it is still turned off by default. It redirects all plugin operations from the existing in-tree plugin to the pxd.portworx.com Container Storage Interface (CSI) Driver. Portworx CSI Driver must be installed on the cluster. To enable the feature, set CSIMigrationPortworx=true in kube-controller-manager and kubelet.

projected

A projected volume maps several existing volume sources into the same directory. For more details, see projected volumes.

rbd

An rbd volume allows a Rados Block Device (RBD) volume to mount into your Pod. Unlike emptyDir, which is erased when a pod is removed, the contents of an rbd volume are preserved and the volume is unmounted. This means that a RBD volume can be pre-populated with data, and that data can be shared between pods.

A feature of RBD is that it can be mounted as read-only by multiple consumers simultaneously. This means that you can pre-populate a volume with your dataset and then serve it in parallel from as many pods as you need. Unfortunately, RBD volumes can only be mounted by a single consumer in read-write mode. Simultaneous writers are not allowed.

See the RBD example for more details.

RBD CSI migration

The CSIMigration feature for RBD, when enabled, redirects all plugin operations from the existing in-tree plugin to the rbd.csi.ceph.com CSI driver. In order to use this feature, the Ceph CSI driver must be installed on the cluster and the CSIMigrationRBD feature gate must be enabled. (Note that the csiMigrationRBD flag has been removed and replaced with CSIMigrationRBD in release v1.24)

secret

A secret volume is used to pass sensitive information, such as passwords, to Pods. You can store secrets in the Kubernetes API and mount them as files for use by pods without coupling to Kubernetes directly. secret volumes are backed by tmpfs (a RAM-backed filesystem) so they are never written to non-volatile storage.

For more details, see Configuring Secrets.

vsphereVolume (deprecated)

A vsphereVolume is used to mount a vSphere VMDK volume into your Pod. The contents of a volume are preserved when it is unmounted. It supports both VMFS and VSAN datastore.

For more information, see the vSphere volume examples.

vSphere CSI migration

In Kubernetes 1.28, all operations for the in-tree vsphereVolume type are redirected to the csi.vsphere.vmware.com CSI driver.

vSphere CSI driver must be installed on the cluster. You can find additional advice on how to migrate in-tree vsphereVolume in VMware's documentation page Migrating In-Tree vSphere Volumes to vSphere Container Storage lug-in. If vSphere CSI Driver is not installed volume operations can not be performed on the PV created with the in-tree vsphereVolume type.

You must run vSphere 7.0u2 or later in order to migrate to the vSphere CSI driver.

If you are running a version of Kubernetes other than v1.28, consult the documentation for that version of Kubernetes.

vSphere CSI migration complete

To turn off the vsphereVolume plugin from being loaded by the controller manager and the kubelet, you need to set InTreePluginvSphereUnregister feature flag to true. You must install a csi.vsphere.vmware.com CSI driver on all worker nodes.

Using subPath

Sometimes, it is useful to share one volume for multiple uses in a single pod. The volumeMounts.subPath property specifies a sub-path inside the referenced volume instead of its root.

The following example shows how to configure a Pod with a LAMP stack (Linux Apache MySQL PHP) using a single, shared volume. This sample subPath configuration is not recommended for production use.

The PHP application's code and assets map to the volume's html folder and the MySQL database is stored in the volume's mysql folder. For example:

apiVersion: v1
kind: Pod
metadata:
  name: my-lamp-site
spec:
    containers:
    - name: mysql
      image: mysql
      env:
      - name: MYSQL_ROOT_PASSWORD
        value: "rootpasswd"
      volumeMounts:
      - mountPath: /var/lib/mysql
        name: site-data
        subPath: mysql
    - name: php
      image: php:7.0-apache
      volumeMounts:
      - mountPath: /var/www/html
        name: site-data
        subPath: html
    volumes:
    - name: site-data
      persistentVolumeClaim:
        claimName: my-lamp-site-data

Using subPath with expanded environment variables

Use the subPathExpr field to construct subPath directory names from downward API environment variables. The subPath and subPathExpr properties are mutually exclusive.

In this example, a Pod uses subPathExpr to create a directory pod1 within the hostPath volume /var/log/pods. The hostPath volume takes the Pod name from the downwardAPI. The host directory /var/log/pods/pod1 is mounted at /logs in the container.

apiVersion: v1
kind: Pod
metadata:
  name: pod1
spec:
  containers:
  - name: container1
    env:
    - name: POD_NAME
      valueFrom:
        fieldRef:
          apiVersion: v1
          fieldPath: metadata.name
    image: busybox:1.28
    command: [ "sh", "-c", "while [ true ]; do echo 'Hello'; sleep 10; done | tee -a /logs/hello.txt" ]
    volumeMounts:
    - name: workdir1
      mountPath: /logs
      # The variable expansion uses round brackets (not curly brackets).
      subPathExpr: $(POD_NAME)
  restartPolicy: Never
  volumes:
  - name: workdir1
    hostPath:
      path: /var/log/pods

Resources

The storage media (such as Disk or SSD) of an emptyDir volume is determined by the medium of the filesystem holding the kubelet root dir (typically /var/lib/kubelet). There is no limit on how much space an emptyDir or hostPath volume can consume, and no isolation between containers or between pods.

To learn about requesting space using a resource specification, see how to manage resources.

Out-of-tree volume plugins

The out-of-tree volume plugins include Container Storage Interface (CSI), and also FlexVolume (which is deprecated). These plugins enable storage vendors to create custom storage plugins without adding their plugin source code to the Kubernetes repository.

Previously, all volume plugins were "in-tree". The "in-tree" plugins were built, linked, compiled, and shipped with the core Kubernetes binaries. This meant that adding a new storage system to Kubernetes (a volume plugin) required checking code into the core Kubernetes code repository.

Both CSI and FlexVolume allow volume plugins to be developed independent of the Kubernetes code base, and deployed (installed) on Kubernetes clusters as extensions.

For storage vendors looking to create an out-of-tree volume plugin, please refer to the volume plugin FAQ.

csi

Container Storage Interface (CSI) defines a standard interface for container orchestration systems (like Kubernetes) to expose arbitrary storage systems to their container workloads.

Please read the CSI design proposal for more information.

Once a CSI compatible volume driver is deployed on a Kubernetes cluster, users may use the csi volume type to attach or mount the volumes exposed by the CSI driver.

A csi volume can be used in a Pod in three different ways:

• through a reference to a PersistentVolumeClaim
• with a generic ephemeral volume
• with a CSI ephemeral volume if the driver supports that

The following fields are available to storage administrators to configure a CSI persistent volume:

• driver: A string value that specifies the name of the volume driver to use. This value must correspond to the value returned in the GetPluginInfoResponse by the CSI driver as defined in the CSI spec. It is used by Kubernetes to identify which CSI driver to call out to, and by CSI driver components to identify which PV objects belong to the CSI driver.
• volumeHandle: A string value that uniquely identifies the volume. This value must correspond to the value returned in the volume.id field of the CreateVolumeResponse by the CSI driver as defined in the CSI spec. The value is passed as volume_id on all calls to the CSI volume driver when referencing the volume.
• readOnly: An optional boolean value indicating whether the volume is to be "ControllerPublished" (attached) as read only. Default is false. This value is passed to the CSI driver via the readonly field in the ControllerPublishVolumeRequest.
• fsType: If the PV's VolumeMode is Filesystem then this field may be used to specify the filesystem that should be used to mount the volume. If the volume has not been formatted and formatting is supported, this value will be used to format the volume. This value is passed to the CSI driver via the VolumeCapability field of ControllerPublishVolumeRequest, NodeStageVolumeRequest, and NodePublishVolumeRequest.
• volumeAttributes: A map of string to string that specifies static properties of a volume. This map must correspond to the map returned in the volume.attributes field of the CreateVolumeResponse by the CSI driver as defined in the CSI spec. The map is passed to the CSI driver via the volume_context field in the ControllerPublishVolumeRequest, NodeStageVolumeRequest, and NodePublishVolumeRequest.
• controllerPublishSecretRef: A reference to the secret object containing sensitive information to pass to the CSI driver to complete the CSI ControllerPublishVolume and ControllerUnpublishVolume calls. This field is optional, and may be empty if no secret is required. If the Secret contains more than one secret, all secrets are passed.
• nodeExpandSecretRef: A reference to the secret containing sensitive information to pass to the CSI driver to complete the CSI NodeExpandVolume call. This field is optional, and may be empty if no secret is required. If the object contains more than one secret, all secrets are passed. When you have configured secret data for node-initiated volume expansion, the kubelet passes that data via the NodeExpandVolume() call to the CSI driver. In order to use the nodeExpandSecretRef field, your cluster should be running Kubernetes version 1.25 or later.
• If you are running Kubernetes Version 1.25 or 1.26, you must enable the feature gate named CSINodeExpandSecret for each kube-apiserver and for the kubelet on every node. In Kubernetes version 1.27 this feature has been enabled by default and no explicit enablement of the feature gate is required. You must also be using a CSI driver that supports or requires secret data during node-initiated storage resize operations.
• nodePublishSecretRef: A reference to the secret object containing sensitive information to pass to the CSI driver to complete the CSI NodePublishVolume call. This field is optional, and may be empty if no secret is required. If the secret object contains more than one secret, all secrets are passed.
• nodeStageSecretRef: A reference to the secret object containing sensitive information to pass to the CSI driver to complete the CSI NodeStageVolume call. This field is optional, and may be empty if no secret is required. If the Secret contains more than one secret, all secrets are passed.

CSI raw block volume support

Vendors with external CSI drivers can implement raw block volume support in Kubernetes workloads.

You can set up your PersistentVolume/PersistentVolumeClaim with raw block volume support as usual, without any CSI specific changes.

CSI ephemeral volumes

You can directly configure CSI volumes within the Pod specification. Volumes specified in this way are ephemeral and do not persist across pod restarts. See Ephemeral Volumes for more information.

For more information on how to develop a CSI driver, refer to the kubernetes-csi documentation

Windows CSI proxy

CSI node plugins need to perform various privileged operations like scanning of disk devices and mounting of file systems. These operations differ for each host operating system. For Linux worker nodes, containerized CSI node plugins are typically deployed as privileged containers. For Windows worker nodes, privileged operations for containerized CSI node plugins is supported using csi-proxy, a community-managed, stand-alone binary that needs to be pre-installed on each Windows node.

For more details, refer to the deployment guide of the CSI plugin you wish to deploy.

Migrating to CSI drivers from in-tree plugins

The CSIMigration feature directs operations against existing in-tree plugins to corresponding CSI plugins (which are expected to be installed and configured). As a result, operators do not have to make any configuration changes to existing Storage Classes, PersistentVolumes or PersistentVolumeClaims (referring to in-tree plugins) when transitioning to a CSI driver that supersedes an in-tree plugin.

The operations and features that are supported include: provisioning/delete, attach/detach, mount/unmount and resizing of volumes.

In-tree plugins that support CSIMigration and have a corresponding CSI driver implemented are listed in Types of Volumes.

The following in-tree plugins support persistent storage on Windows nodes:

• azureFile
• gcePersistentDisk
• vsphereVolume

flexVolume (deprecated)

FlexVolume is an out-of-tree plugin interface that uses an exec-based model to interface with storage drivers. The FlexVolume driver binaries must be installed in a pre-defined volume plugin path on each node and in some cases the control plane nodes as well.

Pods interact with FlexVolume drivers through the flexVolume in-tree volume plugin. For more details, see the FlexVolume README document.

The following FlexVolume plugins, deployed as PowerShell scripts on the host, support Windows nodes:

• SMB
• iSCSI

Mount propagation

Mount propagation allows for sharing volumes mounted by a container to other containers in the same pod, or even to other pods on the same node.

Mount propagation of a volume is controlled by the mountPropagation field in Container.volumeMounts. Its values are:

• None - This volume mount will not receive any subsequent mounts that are mounted to this volume or any of its subdirectories by the host. In similar fashion, no mounts created by the container will be visible on the host. This is the default mode.

  This mode is equal to rprivate mount propagation as described in mount(8)

  However, the CRI runtime may choose rslave mount propagation (i.e., HostToContainer) instead, when rprivate propagation is not applicable. cri-dockerd (Docker) is known to choose rslave mount propagation when the mount source contains the Docker daemon's root directory (/var/lib/docker).

• HostToContainer - This volume mount will receive all subsequent mounts that are mounted to this volume or any of its subdirectories.

  In other words, if the host mounts anything inside the volume mount, the container will see it mounted there.

  Similarly, if any Pod with Bidirectional mount propagation to the same volume mounts anything there, the container with HostToContainer mount propagation will see it.

  This mode is equal to rslave mount propagation as described in the mount(8)

• Bidirectional - This volume mount behaves the same the HostToContainer mount. In addition, all volume mounts created by the container will be propagated back to the host and to all containers of all pods that use the same volume.

  A typical use case for this mode is a Pod with a FlexVolume or CSI driver or a Pod that needs to mount something on the host using a hostPath volume.

  This mode is equal to rshared mount propagation as described in the mount(8)

  Warning: Bidirectional mount propagation can be dangerous. It can damage the host operating system and therefore it is allowed only in privileged containers. Familiarity with Linux kernel behavior is strongly recommended. In addition, any volume mounts created by containers in pods must be destroyed (unmounted) by the containers on termination.

What's next

Follow an example of deploying WordPress and MySQL with Persistent Volumes.

2 - Persistent Volumes

This document describes persistent volumes in Kubernetes. Familiarity with volumes is suggested.

Introduction

Managing storage is a distinct problem from managing compute instances. The PersistentVolume subsystem provides an API for users and administrators that abstracts details of how storage is provided from how it is consumed. To do this, we introduce two new API resources: PersistentVolume and PersistentVolumeClaim.

A PersistentVolume (PV) is a piece of storage in the cluster that has been provisioned by an administrator or dynamically provisioned using Storage Classes. It is a resource in the cluster just like a node is a cluster resource. PVs are volume plugins like Volumes, but have a lifecycle independent of any individual Pod that uses the PV. This API object captures the details of the implementation of the storage, be that NFS, iSCSI, or a cloud-provider-specific storage system.

A PersistentVolumeClaim (PVC) is a request for storage by a user. It is similar to a Pod. Pods consume node resources and PVCs consume PV resources. Pods can request specific levels of resources (CPU and Memory). Claims can request specific size and access modes (e.g., they can be mounted ReadWriteOnce, ReadOnlyMany or ReadWriteMany, see AccessModes).

While PersistentVolumeClaims allow a user to consume abstract storage resources, it is common that users need PersistentVolumes with varying properties, such as performance, for different problems. Cluster administrators need to be able to offer a variety of PersistentVolumes that differ in more ways than size and access modes, without exposing users to the details of how those volumes are implemented. For these needs, there is the StorageClass resource.

See the detailed walkthrough with working examples.

Lifecycle of a volume and claim

PVs are resources in the cluster. PVCs are requests for those resources and also act as claim checks to the resource. The interaction between PVs and PVCs follows this lifecycle:

Provisioning

There are two ways PVs may be provisioned: statically or dynamically.

Static

A cluster administrator creates a number of PVs. They carry the details of the real storage, which is available for use by cluster users. They exist in the Kubernetes API and are available for consumption.

Dynamic

When none of the static PVs the administrator created match a user's PersistentVolumeClaim, the cluster may try to dynamically provision a volume specially for the PVC. This provisioning is based on StorageClasses: the PVC must request a storage class and the administrator must have created and configured that class for dynamic provisioning to occur. Claims that request the class "" effectively disable dynamic provisioning for themselves.

To enable dynamic storage provisioning based on storage class, the cluster administrator needs to enable the DefaultStorageClass admission controller on the API server. This can be done, for example, by ensuring that DefaultStorageClass is among the comma-delimited, ordered list of values for the --enable-admission-plugins flag of the API server component. For more information on API server command-line flags, check kube-apiserver documentation.

Binding

A user creates, or in the case of dynamic provisioning, has already created, a PersistentVolumeClaim with a specific amount of storage requested and with certain access modes. A control loop in the control plane watches for new PVCs, finds a matching PV (if possible), and binds them together. If a PV was dynamically provisioned for a new PVC, the loop will always bind that PV to the PVC. Otherwise, the user will always get at least what they asked for, but the volume may be in excess of what was requested. Once bound, PersistentVolumeClaim binds are exclusive, regardless of how they were bound. A PVC to PV binding is a one-to-one mapping, using a ClaimRef which is a bi-directional binding between the PersistentVolume and the PersistentVolumeClaim.

Claims will remain unbound indefinitely if a matching volume does not exist. Claims will be bound as matching volumes become available. For example, a cluster provisioned with many 50Gi PVs would not match a PVC requesting 100Gi. The PVC can be bound when a 100Gi PV is added to the cluster.

Using

Pods use claims as volumes. The cluster inspects the claim to find the bound volume and mounts that volume for a Pod. For volumes that support multiple access modes, the user specifies which mode is desired when using their claim as a volume in a Pod.

Once a user has a claim and that claim is bound, the bound PV belongs to the user for as long as they need it. Users schedule Pods and access their claimed PVs by including a persistentVolumeClaim section in a Pod's volumes block. See Claims As Volumes for more details on this.

Storage Object in Use Protection

The purpose of the Storage Object in Use Protection feature is to ensure that PersistentVolumeClaims (PVCs) in active use by a Pod and PersistentVolume (PVs) that are bound to PVCs are not removed from the system, as this may result in data loss.

If a user deletes a PVC in active use by a Pod, the PVC is not removed immediately. PVC removal is postponed until the PVC is no longer actively used by any Pods. Also, if an admin deletes a PV that is bound to a PVC, the PV is not removed immediately. PV removal is postponed until the PV is no longer bound to a PVC.

You can see that a PVC is protected when the PVC's status is Terminating and the Finalizers list includes kubernetes.io/pvc-protection:

kubectl describe pvc hostpath
Name:          hostpath
Namespace:     default
StorageClass:  example-hostpath
Status:        Terminating
Volume:
Labels:        <none>
Annotations:   volume.beta.kubernetes.io/storage-class=example-hostpath
               volume.beta.kubernetes.io/storage-provisioner=example.com/hostpath
Finalizers:    [kubernetes.io/pvc-protection]
...

You can see that a PV is protected when the PV's status is Terminating and the Finalizers list includes kubernetes.io/pv-protection too:

kubectl describe pv task-pv-volume
Name:            task-pv-volume
Labels:          type=local
Annotations:     <none>
Finalizers:      [kubernetes.io/pv-protection]
StorageClass:    standard
Status:          Terminating
Claim:
Reclaim Policy:  Delete
Access Modes:    RWO
Capacity:        1Gi
Message:
Source:
    Type:          HostPath (bare host directory volume)
    Path:          /tmp/data
    HostPathType:
Events:            <none>

Reclaiming

When a user is done with their volume, they can delete the PVC objects from the API that allows reclamation of the resource. The reclaim policy for a PersistentVolume tells the cluster what to do with the volume after it has been released of its claim. Currently, volumes can either be Retained, Recycled, or Deleted.

Retain

The Retain reclaim policy allows for manual reclamation of the resource. When the PersistentVolumeClaim is deleted, the PersistentVolume still exists and the volume is considered "released". But it is not yet available for another claim because the previous claimant's data remains on the volume. An administrator can manually reclaim the volume with the following steps.

1. Delete the PersistentVolume. The associated storage asset in external infrastructure (such as an AWS EBS or GCE PD volume) still exists after the PV is deleted.
2. Manually clean up the data on the associated storage asset accordingly.
3. Manually delete the associated storage asset.

If you want to reuse the same storage asset, create a new PersistentVolume with the same storage asset definition.

Delete

For volume plugins that support the Delete reclaim policy, deletion removes both the PersistentVolume object from Kubernetes, as well as the associated storage asset in the external infrastructure, such as an AWS EBS or GCE PD volume. Volumes that were dynamically provisioned inherit the reclaim policy of their StorageClass, which defaults to Delete. The administrator should configure the StorageClass according to users' expectations; otherwise, the PV must be edited or patched after it is created. See Change the Reclaim Policy of a PersistentVolume.

Recycle

If supported by the underlying volume plugin, the Recycle reclaim policy performs a basic scrub (rm -rf /thevolume/*) on the volume and makes it available again for a new claim.

However, an administrator can configure a custom recycler Pod template using the Kubernetes controller manager command line arguments as described in the reference. The custom recycler Pod template must contain a volumes specification, as shown in the example below:

apiVersion: v1
kind: Pod
metadata:
  name: pv-recycler
  namespace: default
spec:
  restartPolicy: Never
  volumes:
  - name: vol
    hostPath:
      path: /any/path/it/will/be/replaced
  containers:
  - name: pv-recycler
    image: "registry.k8s.io/busybox"
    command: ["/bin/sh", "-c", "test -e /scrub && rm -rf /scrub/..?* /scrub/.[!.]* /scrub/*  && test -z \"$(ls -A /scrub)\" || exit 1"]
    volumeMounts:
    - name: vol
      mountPath: /scrub

However, the particular path specified in the custom recycler Pod template in the volumes part is replaced with the particular path of the volume that is being recycled.

PersistentVolume deletion protection finalizer

Finalizers can be added on a PersistentVolume to ensure that PersistentVolumes having Delete reclaim policy are deleted only after the backing storage are deleted.

The newly introduced finalizers kubernetes.io/pv-controller and external-provisioner.volume.kubernetes.io/finalizer are only added to dynamically provisioned volumes.

The finalizer kubernetes.io/pv-controller is added to in-tree plugin volumes. The following is an example

kubectl describe pv pvc-74a498d6-3929-47e8-8c02-078c1ece4d78
Name:            pvc-74a498d6-3929-47e8-8c02-078c1ece4d78
Labels:          <none>
Annotations:     kubernetes.io/createdby: vsphere-volume-dynamic-provisioner
                 pv.kubernetes.io/bound-by-controller: yes
                 pv.kubernetes.io/provisioned-by: kubernetes.io/vsphere-volume
Finalizers:      [kubernetes.io/pv-protection kubernetes.io/pv-controller]
StorageClass:    vcp-sc
Status:          Bound
Claim:           default/vcp-pvc-1
Reclaim Policy:  Delete
Access Modes:    RWO
VolumeMode:      Filesystem
Capacity:        1Gi
Node Affinity:   <none>
Message:         
Source:
    Type:               vSphereVolume (a Persistent Disk resource in vSphere)
    VolumePath:         [vsanDatastore] d49c4a62-166f-ce12-c464-020077ba5d46/kubernetes-dynamic-pvc-74a498d6-3929-47e8-8c02-078c1ece4d78.vmdk
    FSType:             ext4
    StoragePolicyName:  vSAN Default Storage Policy
Events:                 <none>

The finalizer external-provisioner.volume.kubernetes.io/finalizer is added for CSI volumes. The following is an example:

Name:            pvc-2f0bab97-85a8-4552-8044-eb8be45cf48d
Labels:          <none>
Annotations:     pv.kubernetes.io/provisioned-by: csi.vsphere.vmware.com
Finalizers:      [kubernetes.io/pv-protection external-provisioner.volume.kubernetes.io/finalizer]
StorageClass:    fast
Status:          Bound
Claim:           demo-app/nginx-logs
Reclaim Policy:  Delete
Access Modes:    RWO
VolumeMode:      Filesystem
Capacity:        200Mi
Node Affinity:   <none>
Message:         
Source:
    Type:              CSI (a Container Storage Interface (CSI) volume source)
    Driver:            csi.vsphere.vmware.com
    FSType:            ext4
    VolumeHandle:      44830fa8-79b4-406b-8b58-621ba25353fd
    ReadOnly:          false
    VolumeAttributes:      storage.kubernetes.io/csiProvisionerIdentity=1648442357185-8081-csi.vsphere.vmware.com
                           type=vSphere CNS Block Volume
Events:                <none>

When the CSIMigration{provider} feature flag is enabled for a specific in-tree volume plugin, the kubernetes.io/pv-controller finalizer is replaced by the external-provisioner.volume.kubernetes.io/finalizer finalizer.

Reserving a PersistentVolume

The control plane can bind PersistentVolumeClaims to matching PersistentVolumes in the cluster. However, if you want a PVC to bind to a specific PV, you need to pre-bind them.

By specifying a PersistentVolume in a PersistentVolumeClaim, you declare a binding between that specific PV and PVC. If the PersistentVolume exists and has not reserved PersistentVolumeClaims through its claimRef field, then the PersistentVolume and PersistentVolumeClaim will be bound.

The binding happens regardless of some volume matching criteria, including node affinity. The control plane still checks that storage class, access modes, and requested storage size are valid.

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: foo-pvc
  namespace: foo
spec:
  storageClassName: "" # Empty string must be explicitly set otherwise default StorageClass will be set
  volumeName: foo-pv
  ...

This method does not guarantee any binding privileges to the PersistentVolume. If other PersistentVolumeClaims could use the PV that you specify, you first need to reserve that storage volume. Specify the relevant PersistentVolumeClaim in the claimRef field of the PV so that other PVCs can not bind to it.

apiVersion: v1
kind: PersistentVolume
metadata:
  name: foo-pv
spec:
  storageClassName: ""
  claimRef:
    name: foo-pvc
    namespace: foo
  ...

This is useful if you want to consume PersistentVolumes that have their claimPolicy set to Retain, including cases where you are reusing an existing PV.

Expanding Persistent Volumes Claims

Support for expanding PersistentVolumeClaims (PVCs) is enabled by default. You can expand the following types of volumes:

• azureFile (deprecated)
• csi
• flexVolume (deprecated)
• gcePersistentDisk (deprecated)
• rbd (deprecated)
• portworxVolume (deprecated)

You can only expand a PVC if its storage class's allowVolumeExpansion field is set to true.

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: example-vol-default
provisioner: vendor-name.example/magicstorage
parameters:
  resturl: "http://192.168.10.100:8080"
  restuser: ""
  secretNamespace: ""
  secretName: ""
allowVolumeExpansion: true

To request a larger volume for a PVC, edit the PVC object and specify a larger size. This triggers expansion of the volume that backs the underlying PersistentVolume. A new PersistentVolume is never created to satisfy the claim. Instead, an existing volume is resized.

CSI Volume expansion

Support for expanding CSI volumes is enabled by default but it also requires a specific CSI driver to support volume expansion. Refer to documentation of the specific CSI driver for more information.

Resizing a volume containing a file system

You can only resize volumes containing a file system if the file system is XFS, Ext3, or Ext4.

When a volume contains a file system, the file system is only resized when a new Pod is using the PersistentVolumeClaim in ReadWrite mode. File system expansion is either done when a Pod is starting up or when a Pod is running and the underlying file system supports online expansion.

FlexVolumes (deprecated since Kubernetes v1.23) allow resize if the driver is configured with the RequiresFSResize capability to true. The FlexVolume can be resized on Pod restart.

Resizing an in-use PersistentVolumeClaim

In this case, you don't need to delete and recreate a Pod or deployment that is using an existing PVC. Any in-use PVC automatically becomes available to its Pod as soon as its file system has been expanded. This feature has no effect on PVCs that are not in use by a Pod or deployment. You must create a Pod that uses the PVC before the expansion can complete.

Similar to other volume types - FlexVolume volumes can also be expanded when in-use by a Pod.

Recovering from Failure when Expanding Volumes

If a user specifies a new size that is too big to be satisfied by underlying storage system, expansion of PVC will be continuously retried until user or cluster administrator takes some action. This can be undesirable and hence Kubernetes provides following methods of recovering from such failures.

• Manually with Cluster Administrator access
• By requesting expansion to smaller size

Types of Persistent Volumes

PersistentVolume types are implemented as plugins. Kubernetes currently supports the following plugins:

• csi - Container Storage Interface (CSI)
• fc - Fibre Channel (FC) storage
• hostPath - HostPath volume (for single node testing only; WILL NOT WORK in a multi-node cluster; consider using local volume instead)
• iscsi - iSCSI (SCSI over IP) storage
• local - local storage devices mounted on nodes.
• nfs - Network File System (NFS) storage

The following types of PersistentVolume are deprecated. This means that support is still available but will be removed in a future Kubernetes release.

• azureFile - Azure File (deprecated in v1.21)
• flexVolume - FlexVolume (deprecated in v1.23)
• gcePersistentDisk - GCE Persistent Disk (deprecated in v1.17)
• portworxVolume - Portworx volume (deprecated in v1.25)
• vsphereVolume - vSphere VMDK volume (deprecated in v1.19)
• cephfs - CephFS volume (deprecated in v1.28)
• rbd - Rados Block Device (RBD) volume (deprecated in v1.28)

Older versions of Kubernetes also supported the following in-tree PersistentVolume types:

• awsElasticBlockStore - AWS Elastic Block Store (EBS) (not available in v1.27)
• azureDisk - Azure Disk (not available in v1.27)
• cinder - Cinder (OpenStack block storage) (not available in v1.26)
• photonPersistentDisk - Photon controller persistent disk. (not available starting v1.15)
• scaleIO - ScaleIO volume. (not available starting v1.21)
• flocker - Flocker storage. (not available starting v1.25)
• quobyte - Quobyte volume. (not available starting v1.25)
• storageos - StorageOS volume. (not available starting v1.25)

Persistent Volumes

Each PV contains a spec and status, which is the specification and status of the volume. The name of a PersistentVolume object must be a valid DNS subdomain name.

apiVersion: v1
kind: PersistentVolume
metadata:
  name: pv0003
spec:
  capacity:
    storage: 5Gi
  volumeMode: Filesystem
  accessModes:
    - ReadWriteOnce
  persistentVolumeReclaimPolicy: Recycle
  storageClassName: slow
  mountOptions:
    - hard
    - nfsvers=4.1
  nfs:
    path: /tmp
    server: 172.17.0.2

Capacity

Generally, a PV will have a specific storage capacity. This is set using the PV's capacity attribute. Read the glossary term Quantity to understand the units expected by capacity.

Currently, storage size is the only resource that can be set or requested. Future attributes may include IOPS, throughput, etc.

Volume Mode

Kubernetes supports two volumeModes of PersistentVolumes: Filesystem and Block.

volumeMode is an optional API parameter. Filesystem is the default mode used when volumeMode parameter is omitted.

A volume with volumeMode: Filesystem is mounted into Pods into a directory. If the volume is backed by a block device and the device is empty, Kubernetes creates a filesystem on the device before mounting it for the first time.

You can set the value of volumeMode to Block to use a volume as a raw block device. Such volume is presented into a Pod as a block device, without any filesystem on it. This mode is useful to provide a Pod the fastest possible way to access a volume, without any filesystem layer between the Pod and the volume. On the other hand, the application running in the Pod must know how to handle a raw block device. See Raw Block Volume Support for an example on how to use a volume with volumeMode: Block in a Pod.

Access Modes

A PersistentVolume can be mounted on a host in any way supported by the resource provider. As shown in the table below, providers will have different capabilities and each PV's access modes are set to the specific modes supported by that particular volume. For example, NFS can support multiple read/write clients, but a specific NFS PV might be exported on the server as read-only. Each PV gets its own set of access modes describing that specific PV's capabilities.

The access modes are:

ReadWriteOnce

the volume can be mounted as read-write by a single node. ReadWriteOnce access mode still can allow multiple pods to access the volume when the pods are running on the same node.

ReadOnlyMany

the volume can be mounted as read-only by many nodes.

ReadWriteMany

the volume can be mounted as read-write by many nodes.

ReadWriteOncePod

FEATURE STATE: Kubernetes v1.27 [beta]

the volume can be mounted as read-write by a single Pod. Use ReadWriteOncePod access mode if you want to ensure that only one pod across the whole cluster can read that PVC or write to it. This is only supported for CSI volumes and Kubernetes version 1.22+.

The blog article Introducing Single Pod Access Mode for PersistentVolumes covers this in more detail.

In the CLI, the access modes are abbreviated to:

• RWO - ReadWriteOnce
• ROX - ReadOnlyMany
• RWX - ReadWriteMany
• RWOP - ReadWriteOncePod

Important! A volume can only be mounted using one access mode at a time, even if it supports many. For example, a GCEPersistentDisk can be mounted as ReadWriteOnce by a single node or ReadOnlyMany by many nodes, but not at the same time.

Class

A PV can have a class, which is specified by setting the storageClassName attribute to the name of a StorageClass. A PV of a particular class can only be bound to PVCs requesting that class. A PV with no storageClassName has no class and can only be bound to PVCs that request no particular class.

In the past, the annotation volume.beta.kubernetes.io/storage-class was used instead of the storageClassName attribute. This annotation is still working; however, it will become fully deprecated in a future Kubernetes release.

Reclaim Policy

Current reclaim policies are:

• Retain -- manual reclamation
• Recycle -- basic scrub (rm -rf /thevolume/*)
• Delete -- associated storage asset such as AWS EBS or GCE PD volume is deleted

Currently, only NFS and HostPath support recycling. AWS EBS and GCE PD volumes support deletion.

Mount Options

A Kubernetes administrator can specify additional mount options for when a Persistent Volume is mounted on a node.

The following volume types support mount options:

• azureFile
• cephfs (deprecated in v1.28)
• cinder (deprecated in v1.18)
• gcePersistentDisk (deprecated in v1.28)
• iscsi
• nfs
• rbd (deprecated in v1.28)
• vsphereVolume

Mount options are not validated. If a mount option is invalid, the mount fails.

In the past, the annotation volume.beta.kubernetes.io/mount-options was used instead of the mountOptions attribute. This annotation is still working; however, it will become fully deprecated in a future Kubernetes release.

Node Affinity

A PV can specify node affinity to define constraints that limit what nodes this volume can be accessed from. Pods that use a PV will only be scheduled to nodes that are selected by the node affinity. To specify node affinity, set nodeAffinity in the .spec of a PV. The PersistentVolume API reference has more details on this field.

Phase

A PersistentVolume will be in one of the following phases:

Available

a free resource that is not yet bound to a claim

Bound

the volume is bound to a claim

Released

the claim has been deleted, but the associated storage resource is not yet reclaimed by the cluster

Failed

the volume has failed its (automated) reclamation

You can see the name of the PVC bound to the PV using kubectl describe persistentvolume <name>.

Phase transition timestamp

The .status field for a PersistentVolume can include an alpha lastPhaseTransitionTime field. This field records the timestamp of when the volume last transitioned its phase. For newly created volumes the phase is set to Pending and lastPhaseTransitionTime is set to the current time.

PersistentVolumeClaims

Each PVC contains a spec and status, which is the specification and status of the claim. The name of a PersistentVolumeClaim object must be a valid DNS subdomain name.

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: myclaim
spec:
  accessModes:
    - ReadWriteOnce
  volumeMode: Filesystem
  resources:
    requests:
      storage: 8Gi
  storageClassName: slow
  selector:
    matchLabels:
      release: "stable"
    matchExpressions:
      - {key: environment, operator: In, values: [dev]}

Access Modes

Claims use the same conventions as volumes when requesting storage with specific access modes.

Volume Modes

Claims use the same convention as volumes to indicate the consumption of the volume as either a filesystem or block device.

Resources

Claims, like Pods, can request specific quantities of a resource. In this case, the request is for storage. The same resource model applies to both volumes and claims.

Selector

Claims can specify a label selector to further filter the set of volumes. Only the volumes whose labels match the selector can be bound to the claim. The selector can consist of two fields:

• matchLabels - the volume must have a label with this value
• matchExpressions - a list of requirements made by specifying key, list of values, and operator that relates the key and values. Valid operators include In, NotIn, Exists, and DoesNotExist.

All of the requirements, from both matchLabels and matchExpressions, are ANDed together – they must all be satisfied in order to match.

Class

A claim can request a particular class by specifying the name of a StorageClass using the attribute storageClassName. Only PVs of the requested class, ones with the same storageClassName as the PVC, can be bound to the PVC.

PVCs don't necessarily have to request a class. A PVC with its storageClassName set equal to "" is always interpreted to be requesting a PV with no class, so it can only be bound to PVs with no class (no annotation or one set equal to ""). A PVC with no storageClassName is not quite the same and is treated differently by the cluster, depending on whether the DefaultStorageClass admission plugin is turned on.

• If the admission plugin is turned on, the administrator may specify a default StorageClass. All PVCs that have no storageClassName can be bound only to PVs of that default. Specifying a default StorageClass is done by setting the annotation storageclass.kubernetes.io/is-default-class equal to true in a StorageClass object. If the administrator does not specify a default, the cluster responds to PVC creation as if the admission plugin were turned off. If more than one default StorageClass is specified, the newest default is used when the PVC is dynamically provisioned.
• If the admission plugin is turned off, there is no notion of a default StorageClass. All PVCs that have storageClassName set to "" can be bound only to PVs that have storageClassName also set to "". However, PVCs with missing storageClassName can be updated later once default StorageClass becomes available. If the PVC gets updated it will no longer bind to PVs that have storageClassName also set to "".

See retroactive default StorageClass assignment for more details.

Depending on installation method, a default StorageClass may be deployed to a Kubernetes cluster by addon manager during installation.

When a PVC specifies a selector in addition to requesting a StorageClass, the requirements are ANDed together: only a PV of the requested class and with the requested labels may be bound to the PVC.

In the past, the annotation volume.beta.kubernetes.io/storage-class was used instead of storageClassName attribute. This annotation is still working; however, it won't be supported in a future Kubernetes release.

Retroactive default StorageClass assignment

You can create a PersistentVolumeClaim without specifying a storageClassName for the new PVC, and you can do so even when no default StorageClass exists in your cluster. In this case, the new PVC creates as you defined it, and the storageClassName of that PVC remains unset until default becomes available.

When a default StorageClass becomes available, the control plane identifies any existing PVCs without storageClassName. For the PVCs that either have an empty value for storageClassName or do not have this key, the control plane then updates those PVCs to set storageClassName to match the new default StorageClass. If you have an existing PVC where the storageClassName is "", and you configure a default StorageClass, then this PVC will not get updated.

In order to keep binding to PVs with storageClassName set to "" (while a default StorageClass is present), you need to set the storageClassName of the associated PVC to "".

This behavior helps administrators change default StorageClass by removing the old one first and then creating or setting another one. This brief window while there is no default causes PVCs without storageClassName created at that time to not have any default, but due to the retroactive default StorageClass assignment this way of changing defaults is safe.

Claims As Volumes

Pods access storage by using the claim as a volume. Claims must exist in the same namespace as the Pod using the claim. The cluster finds the claim in the Pod's namespace and uses it to get the PersistentVolume backing the claim. The volume is then mounted to the host and into the Pod.

apiVersion: v1
kind: Pod
metadata:
  name: mypod
spec:
  containers:
    - name: myfrontend
      image: nginx
      volumeMounts:
      - mountPath: "/var/www/html"
        name: mypd
  volumes:
    - name: mypd
      persistentVolumeClaim:
        claimName: myclaim

A Note on Namespaces

PersistentVolumes binds are exclusive, and since PersistentVolumeClaims are namespaced objects, mounting claims with "Many" modes (ROX, RWX) is only possible within one namespace.

PersistentVolumes typed hostPath

A hostPath PersistentVolume uses a file or directory on the Node to emulate network-attached storage. See an example of hostPath typed volume.

Raw Block Volume Support

The following volume plugins support raw block volumes, including dynamic provisioning where applicable:

• CSI
• FC (Fibre Channel)
• GCEPersistentDisk (deprecated)
• iSCSI
• Local volume
• OpenStack Cinder
• RBD (deprecated)
• RBD (Ceph Block Device; deprecated)
• VsphereVolume

PersistentVolume using a Raw Block Volume

apiVersion: v1
kind: PersistentVolume
metadata:
  name: block-pv
spec:
  capacity:
    storage: 10Gi
  accessModes:
    - ReadWriteOnce
  volumeMode: Block
  persistentVolumeReclaimPolicy: Retain
  fc:
    targetWWNs: ["50060e801049cfd1"]
    lun: 0
    readOnly: false

PersistentVolumeClaim requesting a Raw Block Volume

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: block-pvc
spec:
  accessModes:
    - ReadWriteOnce
  volumeMode: Block
  resources:
    requests:
      storage: 10Gi

Pod specification adding Raw Block Device path in container

apiVersion: v1
kind: Pod
metadata:
  name: pod-with-block-volume
spec:
  containers:
    - name: fc-container
      image: fedora:26
      command: ["/bin/sh", "-c"]
      args: [ "tail -f /dev/null" ]
      volumeDevices:
        - name: data
          devicePath: /dev/xvda
  volumes:
    - name: data
      persistentVolumeClaim:
        claimName: block-pvc

Binding Block Volumes

If a user requests a raw block volume by indicating this using the volumeMode field in the PersistentVolumeClaim spec, the binding rules differ slightly from previous releases that didn't consider this mode as part of the spec. Listed is a table of possible combinations the user and admin might specify for requesting a raw block device. The table indicates if the volume will be bound or not given the combinations: Volume binding matrix for statically provisioned volumes:

Volume Snapshot and Restore Volume from Snapshot Support

Volume snapshots only support the out-of-tree CSI volume plugins. For details, see Volume Snapshots. In-tree volume plugins are deprecated. You can read about the deprecated volume plugins in the Volume Plugin FAQ.

Create a PersistentVolumeClaim from a Volume Snapshot

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: restore-pvc
spec:
  storageClassName: csi-hostpath-sc
  dataSource:
    name: new-snapshot-test
    kind: VolumeSnapshot
    apiGroup: snapshot.storage.k8s.io
  accessModes:
    - ReadWriteOnce
  resources:
    requests:
      storage: 10Gi

Volume Cloning

Volume Cloning only available for CSI volume plugins.

Create PersistentVolumeClaim from an existing PVC

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: cloned-pvc
spec:
  storageClassName: my-csi-plugin
  dataSource:
    name: existing-src-pvc-name
    kind: PersistentVolumeClaim
  accessModes:
    - ReadWriteOnce
  resources:
    requests:
      storage: 10Gi

Volume populators and data sources

Kubernetes supports custom volume populators. To use custom volume populators, you must enable the AnyVolumeDataSource feature gate for the kube-apiserver and kube-controller-manager.

Volume populators take advantage of a PVC spec field called dataSourceRef. Unlike the dataSource field, which can only contain either a reference to another PersistentVolumeClaim or to a VolumeSnapshot, the dataSourceRef field can contain a reference to any object in the same namespace, except for core objects other than PVCs. For clusters that have the feature gate enabled, use of the dataSourceRef is preferred over dataSource.

Cross namespace data sources

Kubernetes supports cross namespace volume data sources. To use cross namespace volume data sources, you must enable the AnyVolumeDataSource and CrossNamespaceVolumeDataSource feature gates for the kube-apiserver and kube-controller-manager. Also, you must enable the CrossNamespaceVolumeDataSource feature gate for the csi-provisioner.

Enabling the CrossNamespaceVolumeDataSource feature gate allows you to specify a namespace in the dataSourceRef field.

Data source references

The dataSourceRef field behaves almost the same as the dataSource field. If one is specified while the other is not, the API server will give both fields the same value. Neither field can be changed after creation, and attempting to specify different values for the two fields will result in a validation error. Therefore the two fields will always have the same contents.

There are two differences between the dataSourceRef field and the dataSource field that users should be aware of:

• The dataSource field ignores invalid values (as if the field was blank) while the dataSourceRef field never ignores values and will cause an error if an invalid value is used. Invalid values are any core object (objects with no apiGroup) except for PVCs.
• The dataSourceRef field may contain different types of objects, while the dataSource field only allows PVCs and VolumeSnapshots.

When the CrossNamespaceVolumeDataSource feature is enabled, there are additional differences:

• The dataSource field only allows local objects, while the dataSourceRef field allows objects in any namespaces.
• When namespace is specified, dataSource and dataSourceRef are not synced.

Users should always use dataSourceRef on clusters that have the feature gate enabled, and fall back to dataSource on clusters that do not. It is not necessary to look at both fields under any circumstance. The duplicated values with slightly different semantics exist only for backwards compatibility. In particular, a mixture of older and newer controllers are able to interoperate because the fields are the same.

Using volume populators

Volume populators are controllers that can create non-empty volumes, where the contents of the volume are determined by a Custom Resource. Users create a populated volume by referring to a Custom Resource using the dataSourceRef field:

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: populated-pvc
spec:
  dataSourceRef:
    name: example-name
    kind: ExampleDataSource
    apiGroup: example.storage.k8s.io
  accessModes:
    - ReadWriteOnce
  resources:
    requests:
      storage: 10Gi

Because volume populators are external components, attempts to create a PVC that uses one can fail if not all the correct components are installed. External controllers should generate events on the PVC to provide feedback on the status of the creation, including warnings if the PVC cannot be created due to some missing component.

You can install the alpha volume data source validator controller into your cluster. That controller generates warning Events on a PVC in the case that no populator is registered to handle that kind of data source. When a suitable populator is installed for a PVC, it's the responsibility of that populator controller to report Events that relate to volume creation and issues during the process.

Using a cross-namespace volume data source

Create a ReferenceGrant to allow the namespace owner to accept the reference. You define a populated volume by specifying a cross namespace volume data source using the dataSourceRef field. You must already have a valid ReferenceGrant in the source namespace:

apiVersion: gateway.networking.k8s.io/v1beta1
kind: ReferenceGrant
metadata:
  name: allow-ns1-pvc
  namespace: default
spec:
  from:
  - group: ""
    kind: PersistentVolumeClaim
    namespace: ns1
  to:
  - group: snapshot.storage.k8s.io
    kind: VolumeSnapshot
    name: new-snapshot-demo

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: foo-pvc
  namespace: ns1
spec:
  storageClassName: example
  accessModes:
  - ReadWriteOnce
  resources:
    requests:
      storage: 1Gi
  dataSourceRef:
    apiGroup: snapshot.storage.k8s.io
    kind: VolumeSnapshot
    name: new-snapshot-demo
    namespace: default
  volumeMode: Filesystem

Writing Portable Configuration

If you're writing configuration templates or examples that run on a wide range of clusters and need persistent storage, it is recommended that you use the following pattern:

• Include PersistentVolumeClaim objects in your bundle of config (alongside Deployments, ConfigMaps, etc).
• Do not include PersistentVolume objects in the config, since the user instantiating the config may not have permission to create PersistentVolumes.
• Give the user the option of providing a storage class name when instantiating the template.
  • If the user provides a storage class name, put that value into the persistentVolumeClaim.storageClassName field. This will cause the PVC to match the right storage class if the cluster has StorageClasses enabled by the admin.
  • If the user does not provide a storage class name, leave the persistentVolumeClaim.storageClassName field as nil. This will cause a PV to be automatically provisioned for the user with the default StorageClass in the cluster. Many cluster environments have a default StorageClass installed, or administrators can create their own default StorageClass.
• In your tooling, watch for PVCs that are not getting bound after some time and surface this to the user, as this may indicate that the cluster has no dynamic storage support (in which case the user should create a matching PV) or the cluster has no storage system (in which case the user cannot deploy config requiring PVCs).

What's next

• Learn more about Creating a PersistentVolume.
• Learn more about Creating a PersistentVolumeClaim.
• Read the Persistent Storage design document.

API references

Read about the APIs described in this page:

• PersistentVolume
• PersistentVolumeClaim

3 - Projected Volumes

This document describes projected volumes in Kubernetes. Familiarity with volumes is suggested.

Introduction

A projected volume maps several existing volume sources into the same directory.

Currently, the following types of volume sources can be projected:

• secret
• downwardAPI
• configMap
• serviceAccountToken

All sources are required to be in the same namespace as the Pod. For more details, see the all-in-one volume design document.

Example configuration with a secret, a downwardAPI, and a configMap

pods/storage/projected-secret-downwardapi-configmap.yaml

apiVersion: v1
kind: Pod
metadata:
  name: volume-test
spec:
  containers:
  - name: container-test
    image: busybox:1.28
    command: ["sleep", "3600"]
    volumeMounts:
    - name: all-in-one
      mountPath: "/projected-volume"
      readOnly: true
  volumes:
  - name: all-in-one
    projected:
      sources:
      - secret:
          name: mysecret
          items:
            - key: username
              path: my-group/my-username
      - downwardAPI:
          items:
            - path: "labels"
              fieldRef:
                fieldPath: metadata.labels
            - path: "cpu_limit"
              resourceFieldRef:
                containerName: container-test
                resource: limits.cpu
      - configMap:
          name: myconfigmap
          items:
            - key: config
              path: my-group/my-config

Example configuration: secrets with a non-default permission mode set

pods/storage/projected-secrets-nondefault-permission-mode.yaml

apiVersion: v1
kind: Pod
metadata:
  name: volume-test
spec:
  containers:
  - name: container-test
    image: busybox:1.28
    command: ["sleep", "3600"]
    volumeMounts:
    - name: all-in-one
      mountPath: "/projected-volume"
      readOnly: true
  volumes:
  - name: all-in-one
    projected:
      sources:
      - secret:
          name: mysecret
          items:
            - key: username
              path: my-group/my-username
      - secret:
          name: mysecret2
          items:
            - key: password
              path: my-group/my-password
              mode: 511

Each projected volume source is listed in the spec under sources. The parameters are nearly the same with two exceptions:

• For secrets, the secretName field has been changed to name to be consistent with ConfigMap naming.
• The defaultMode can only be specified at the projected level and not for each volume source. However, as illustrated above, you can explicitly set the mode for each individual projection.

serviceAccountToken projected volumes

You can inject the token for the current service account into a Pod at a specified path. For example:

pods/storage/projected-service-account-token.yaml

apiVersion: v1
kind: Pod
metadata:
  name: sa-token-test
spec:
  containers:
  - name: container-test
    image: busybox:1.28
    command: ["sleep", "3600"]
    volumeMounts:
    - name: token-vol
      mountPath: "/service-account"
      readOnly: true
  serviceAccountName: default
  volumes:
  - name: token-vol
    projected:
      sources:
      - serviceAccountToken:
          audience: api
          expirationSeconds: 3600
          path: token

The example Pod has a projected volume containing the injected service account token. Containers in this Pod can use that token to access the Kubernetes API server, authenticating with the identity of the pod's ServiceAccount. The audience field contains the intended audience of the token. A recipient of the token must identify itself with an identifier specified in the audience of the token, and otherwise should reject the token. This field is optional and it defaults to the identifier of the API server.

The expirationSeconds is the expected duration of validity of the service account token. It defaults to 1 hour and must be at least 10 minutes (600 seconds). An administrator can also limit its maximum value by specifying the --service-account-max-token-expiration option for the API server. The path field specifies a relative path to the mount point of the projected volume.

SecurityContext interactions

The proposal for file permission handling in projected service account volume enhancement introduced the projected files having the correct owner permissions set.

Linux

In Linux pods that have a projected volume and RunAsUser set in the Pod SecurityContext, the projected files have the correct ownership set including container user ownership.

When all containers in a pod have the same runAsUser set in their PodSecurityContext or container SecurityContext, then the kubelet ensures that the contents of the serviceAccountToken volume are owned by that user, and the token file has its permission mode set to 0600.

Windows

In Windows pods that have a projected volume and RunAsUsername set in the Pod SecurityContext, the ownership is not enforced due to the way user accounts are managed in Windows. Windows stores and manages local user and group accounts in a database file called Security Account Manager (SAM). Each container maintains its own instance of the SAM database, to which the host has no visibility into while the container is running. Windows containers are designed to run the user mode portion of the OS in isolation from the host, hence the maintenance of a virtual SAM database. As a result, the kubelet running on the host does not have the ability to dynamically configure host file ownership for virtualized container accounts. It is recommended that if files on the host machine are to be shared with the container then they should be placed into their own volume mount outside of C:\.

By default, the projected files will have the following ownership as shown for an example projected volume file:

PS C:\> Get-Acl C:\var\run\secrets\kubernetes.io\serviceaccount\..2021_08_31_22_22_18.318230061\ca.crt | Format-List

Path   : Microsoft.PowerShell.Core\FileSystem::C:\var\run\secrets\kubernetes.io\serviceaccount\..2021_08_31_22_22_18.318230061\ca.crt
Owner  : BUILTIN\Administrators
Group  : NT AUTHORITY\SYSTEM
Access : NT AUTHORITY\SYSTEM Allow  FullControl
         BUILTIN\Administrators Allow  FullControl
         BUILTIN\Users Allow  ReadAndExecute, Synchronize
Audit  :
Sddl   : O:BAG:SYD:AI(A;ID;FA;;;SY)(A;ID;FA;;;BA)(A;ID;0x1200a9;;;BU)

This implies all administrator users like ContainerAdministrator will have read, write and execute access while, non-administrator users will have read and execute access.

4 - Ephemeral Volumes

This document describes ephemeral volumes in Kubernetes. Familiarity with volumes is suggested, in particular PersistentVolumeClaim and PersistentVolume.

Some application need additional storage but don't care whether that data is stored persistently across restarts. For example, caching services are often limited by memory size and can move infrequently used data into storage that is slower than memory with little impact on overall performance.

Other applications expect some read-only input data to be present in files, like configuration data or secret keys.

Ephemeral volumes are designed for these use cases. Because volumes follow the Pod's lifetime and get created and deleted along with the Pod, Pods can be stopped and restarted without being limited to where some persistent volume is available.

Ephemeral volumes are specified inline in the Pod spec, which simplifies application deployment and management.

Types of ephemeral volumes

Kubernetes supports several different kinds of ephemeral volumes for different purposes:

• emptyDir: empty at Pod startup, with storage coming locally from the kubelet base directory (usually the root disk) or RAM
• configMap, downwardAPI, secret: inject different kinds of Kubernetes data into a Pod
• CSI ephemeral volumes: similar to the previous volume kinds, but provided by special CSI drivers which specifically support this feature
• generic ephemeral volumes, which can be provided by all storage drivers that also support persistent volumes

emptyDir, configMap, downwardAPI, secret are provided as local ephemeral storage. They are managed by kubelet on each node.

CSI ephemeral volumes must be provided by third-party CSI storage drivers.

Generic ephemeral volumes can be provided by third-party CSI storage drivers, but also by any other storage driver that supports dynamic provisioning. Some CSI drivers are written specifically for CSI ephemeral volumes and do not support dynamic provisioning: those then cannot be used for generic ephemeral volumes.

The advantage of using third-party drivers is that they can offer functionality that Kubernetes itself does not support, for example storage with different performance characteristics than the disk that is managed by kubelet, or injecting different data.

CSI ephemeral volumes

Conceptually, CSI ephemeral volumes are similar to configMap, downwardAPI and secret volume types: the storage is managed locally on each node and is created together with other local resources after a Pod has been scheduled onto a node. Kubernetes has no concept of rescheduling Pods anymore at this stage. Volume creation has to be unlikely to fail, otherwise Pod startup gets stuck. In particular, storage capacity aware Pod scheduling is not supported for these volumes. They are currently also not covered by the storage resource usage limits of a Pod, because that is something that kubelet can only enforce for storage that it manages itself.

Here's an example manifest for a Pod that uses CSI ephemeral storage:

kind: Pod
apiVersion: v1
metadata:
  name: my-csi-app
spec:
  containers:
    - name: my-frontend
      image: busybox:1.28
      volumeMounts:
      - mountPath: "/data"
        name: my-csi-inline-vol
      command: [ "sleep", "1000000" ]
  volumes:
    - name: my-csi-inline-vol
      csi:
        driver: inline.storage.kubernetes.io
        volumeAttributes:
          foo: bar

The volumeAttributes determine what volume is prepared by the driver. These attributes are specific to each driver and not standardized. See the documentation of each CSI driver for further instructions.

CSI driver restrictions

CSI ephemeral volumes allow users to provide volumeAttributes directly to the CSI driver as part of the Pod spec. A CSI driver allowing volumeAttributes that are typically restricted to administrators is NOT suitable for use in an inline ephemeral volume. For example, parameters that are normally defined in the StorageClass should not be exposed to users through the use of inline ephemeral volumes.

Cluster administrators who need to restrict the CSI drivers that are allowed to be used as inline volumes within a Pod spec may do so by:

• Removing Ephemeral from volumeLifecycleModes in the CSIDriver spec, which prevents the driver from being used as an inline ephemeral volume.
• Using an admission webhook to restrict how this driver is used.

Generic ephemeral volumes

Generic ephemeral volumes are similar to emptyDir volumes in the sense that they provide a per-pod directory for scratch data that is usually empty after provisioning. But they may also have additional features:

• Storage can be local or network-attached.
• Volumes can have a fixed size that Pods are not able to exceed.
• Volumes may have some initial data, depending on the driver and parameters.
• Typical operations on volumes are supported assuming that the driver supports them, including snapshotting, cloning, resizing, and storage capacity tracking.

Example:

kind: Pod
apiVersion: v1
metadata:
  name: my-app
spec:
  containers:
    - name: my-frontend
      image: busybox:1.28
      volumeMounts:
      - mountPath: "/scratch"
        name: scratch-volume
      command: [ "sleep", "1000000" ]
  volumes:
    - name: scratch-volume
      ephemeral:
        volumeClaimTemplate:
          metadata:
            labels:
              type: my-frontend-volume
          spec:
            accessModes: [ "ReadWriteOnce" ]
            storageClassName: "scratch-storage-class"
            resources:
              requests:
                storage: 1Gi

Lifecycle and PersistentVolumeClaim

The key design idea is that the parameters for a volume claim are allowed inside a volume source of the Pod. Labels, annotations and the whole set of fields for a PersistentVolumeClaim are supported. When such a Pod gets created, the ephemeral volume controller then creates an actual PersistentVolumeClaim object in the same namespace as the Pod and ensures that the PersistentVolumeClaim gets deleted when the Pod gets deleted.

That triggers volume binding and/or provisioning, either immediately if the StorageClass uses immediate volume binding or when the Pod is tentatively scheduled onto a node (WaitForFirstConsumer volume binding mode). The latter is recommended for generic ephemeral volumes because then the scheduler is free to choose a suitable node for the Pod. With immediate binding, the scheduler is forced to select a node that has access to the volume once it is available.

In terms of resource ownership, a Pod that has generic ephemeral storage is the owner of the PersistentVolumeClaim(s) that provide that ephemeral storage. When the Pod is deleted, the Kubernetes garbage collector deletes the PVC, which then usually triggers deletion of the volume because the default reclaim policy of storage classes is to delete volumes. You can create quasi-ephemeral local storage using a StorageClass with a reclaim policy of retain: the storage outlives the Pod, and in this case you need to ensure that volume clean up happens separately.

While these PVCs exist, they can be used like any other PVC. In particular, they can be referenced as data source in volume cloning or snapshotting. The PVC object also holds the current status of the volume.

PersistentVolumeClaim naming

Naming of the automatically created PVCs is deterministic: the name is a combination of the Pod name and volume name, with a hyphen (-) in the middle. In the example above, the PVC name will be my-app-scratch-volume. This deterministic naming makes it easier to interact with the PVC because one does not have to search for it once the Pod name and volume name are known.

The deterministic naming also introduces a potential conflict between different Pods (a Pod "pod-a" with volume "scratch" and another Pod with name "pod" and volume "a-scratch" both end up with the same PVC name "pod-a-scratch") and between Pods and manually created PVCs.

Such conflicts are detected: a PVC is only used for an ephemeral volume if it was created for the Pod. This check is based on the ownership relationship. An existing PVC is not overwritten or modified. But this does not resolve the conflict because without the right PVC, the Pod cannot start.

Security

Using generic ephemeral volumes allows users to create PVCs indirectly if they can create Pods, even if they do not have permission to create PVCs directly. Cluster administrators must be aware of this. If this does not fit their security model, they should use an admission webhook that rejects objects like Pods that have a generic ephemeral volume.

The normal namespace quota for PVCs still applies, so even if users are allowed to use this new mechanism, they cannot use it to circumvent other policies.

What's next

Ephemeral volumes managed by kubelet

See local ephemeral storage.

CSI ephemeral volumes

• For more information on the design, see the Ephemeral Inline CSI volumes KEP.
• For more information on further development of this feature, see the enhancement tracking issue #596.

Generic ephemeral volumes

• For more information on the design, see the Generic ephemeral inline volumes KEP.

5 - Storage Classes

This document describes the concept of a StorageClass in Kubernetes. Familiarity with volumes and persistent volumes is suggested.

Introduction

A StorageClass provides a way for administrators to describe the "classes" of storage they offer. Different classes might map to quality-of-service levels, or to backup policies, or to arbitrary policies determined by the cluster administrators. Kubernetes itself is unopinionated about what classes represent. This concept is sometimes called "profiles" in other storage systems.

The StorageClass Resource

Each StorageClass contains the fields provisioner, parameters, and reclaimPolicy, which are used when a PersistentVolume belonging to the class needs to be dynamically provisioned.

The name of a StorageClass object is significant, and is how users can request a particular class. Administrators set the name and other parameters of a class when first creating StorageClass objects.

Administrators can specify a default StorageClass only for PVCs that don't request any particular class to bind to: see the PersistentVolumeClaim section for details.

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: standard
provisioner: kubernetes.io/aws-ebs
parameters:
  type: gp2
reclaimPolicy: Retain
allowVolumeExpansion: true
mountOptions:
  - debug
volumeBindingMode: Immediate

Default StorageClass

When a PVC does not specify a storageClassName, the default StorageClass is used. The cluster can only have one default StorageClass. If more than one default StorageClass is accidentally set, the newest default is used when the PVC is dynamically provisioned.

For instructions on setting the default StorageClass, see Change the default StorageClass. Note that certain cloud providers may already define a default StorageClass.

Provisioner

Each StorageClass has a provisioner that determines what volume plugin is used for provisioning PVs. This field must be specified.

You are not restricted to specifying the "internal" provisioners listed here (whose names are prefixed with "kubernetes.io" and shipped alongside Kubernetes). You can also run and specify external provisioners, which are independent programs that follow a specification defined by Kubernetes. Authors of external provisioners have full discretion over where their code lives, how the provisioner is shipped, how it needs to be run, what volume plugin it uses (including Flex), etc. The repository kubernetes-sigs/sig-storage-lib-external-provisioner houses a library for writing external provisioners that implements the bulk of the specification. Some external provisioners are listed under the repository kubernetes-sigs/sig-storage-lib-external-provisioner.

For example, NFS doesn't provide an internal provisioner, but an external provisioner can be used. There are also cases when 3rd party storage vendors provide their own external provisioner.

Reclaim Policy

PersistentVolumes that are dynamically created by a StorageClass will have the reclaim policy specified in the reclaimPolicy field of the class, which can be either Delete or Retain. If no reclaimPolicy is specified when a StorageClass object is created, it will default to Delete.

PersistentVolumes that are created manually and managed via a StorageClass will have whatever reclaim policy they were assigned at creation.

Allow Volume Expansion

PersistentVolumes can be configured to be expandable. This feature when set to true, allows the users to resize the volume by editing the corresponding PVC object.

The following types of volumes support volume expansion, when the underlying StorageClass has the field allowVolumeExpansion set to true.

Mount Options

PersistentVolumes that are dynamically created by a StorageClass will have the mount options specified in the mountOptions field of the class.

If the volume plugin does not support mount options but mount options are specified, provisioning will fail. Mount options are not validated on either the class or PV. If a mount option is invalid, the PV mount fails.

Volume Binding Mode

The volumeBindingMode field controls when volume binding and dynamic provisioning should occur. When unset, "Immediate" mode is used by default.

The Immediate mode indicates that volume binding and dynamic provisioning occurs once the PersistentVolumeClaim is created. For storage backends that are topology-constrained and not globally accessible from all Nodes in the cluster, PersistentVolumes will be bound or provisioned without knowledge of the Pod's scheduling requirements. This may result in unschedulable Pods.

A cluster administrator can address this issue by specifying the WaitForFirstConsumer mode which will delay the binding and provisioning of a PersistentVolume until a Pod using the PersistentVolumeClaim is created. PersistentVolumes will be selected or provisioned conforming to the topology that is specified by the Pod's scheduling constraints. These include, but are not limited to, resource requirements, node selectors, pod affinity and anti-affinity, and taints and tolerations.

The following plugins support WaitForFirstConsumer with dynamic provisioning:

• GCEPersistentDisk

The following plugins support WaitForFirstConsumer with pre-created PersistentVolume binding:

• All of the above
• Local

CSI volumes are also supported with dynamic provisioning and pre-created PVs, but you'll need to look at the documentation for a specific CSI driver to see its supported topology keys and examples.

apiVersion: v1
kind: Pod
metadata:
  name: task-pv-pod
spec:
  nodeSelector:
    kubernetes.io/hostname: kube-01
  volumes:
    - name: task-pv-storage
      persistentVolumeClaim:
        claimName: task-pv-claim
  containers:
    - name: task-pv-container
      image: nginx
      ports:
        - containerPort: 80
          name: "http-server"
      volumeMounts:
        - mountPath: "/usr/share/nginx/html"
          name: task-pv-storage

Allowed Topologies

When a cluster operator specifies the WaitForFirstConsumer volume binding mode, it is no longer necessary to restrict provisioning to specific topologies in most situations. However, if still required, allowedTopologies can be specified.

This example demonstrates how to restrict the topology of provisioned volumes to specific zones and should be used as a replacement for the zone and zones parameters for the supported plugins.

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: standard
provisioner: kubernetes.io/gce-pd
parameters:
  type: pd-standard
volumeBindingMode: WaitForFirstConsumer
allowedTopologies:
- matchLabelExpressions:
  - key: topology.kubernetes.io/zone
    values:
    - us-central-1a
    - us-central-1b

Parameters

Storage Classes have parameters that describe volumes belonging to the storage class. Different parameters may be accepted depending on the provisioner. For example, the value io1, for the parameter type, and the parameter iopsPerGB are specific to EBS. When a parameter is omitted, some default is used.

There can be at most 512 parameters defined for a StorageClass. The total length of the parameters object including its keys and values cannot exceed 256 KiB.

AWS EBS

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: slow
provisioner: kubernetes.io/aws-ebs
parameters:
  type: io1
  iopsPerGB: "10"
  fsType: ext4

• type: io1, gp2, sc1, st1. See AWS docs for details. Default: gp2.
• zone (Deprecated): AWS zone. If neither zone nor zones is specified, volumes are generally round-robin-ed across all active zones where Kubernetes cluster has a node. zone and zones parameters must not be used at the same time.
• zones (Deprecated): A comma separated list of AWS zone(s). If neither zone nor zones is specified, volumes are generally round-robin-ed across all active zones where Kubernetes cluster has a node. zone and zones parameters must not be used at the same time.
• iopsPerGB: only for io1 volumes. I/O operations per second per GiB. AWS volume plugin multiplies this with size of requested volume to compute IOPS of the volume and caps it at 20 000 IOPS (maximum supported by AWS, see AWS docs). A string is expected here, i.e. "10", not 10.
• fsType: fsType that is supported by kubernetes. Default: "ext4".
• encrypted: denotes whether the EBS volume should be encrypted or not. Valid values are "true" or "false". A string is expected here, i.e. "true", not true.
• kmsKeyId: optional. The full Amazon Resource Name of the key to use when encrypting the volume. If none is supplied but encrypted is true, a key is generated by AWS. See AWS docs for valid ARN value.

GCE PD

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: slow
provisioner: kubernetes.io/gce-pd
parameters:
  type: pd-standard
  fstype: ext4
  replication-type: none

• type: pd-standard or pd-ssd. Default: pd-standard

• zone (Deprecated): GCE zone. If neither zone nor zones is specified, volumes are generally round-robin-ed across all active zones where Kubernetes cluster has a node. zone and zones parameters must not be used at the same time.

• zones (Deprecated): A comma separated list of GCE zone(s). If neither zone nor zones is specified, volumes are generally round-robin-ed across all active zones where Kubernetes cluster has a node. zone and zones parameters must not be used at the same time.

• fstype: ext4 or xfs. Default: ext4. The defined filesystem type must be supported by the host operating system.

• replication-type: none or regional-pd. Default: none.

If replication-type is set to none, a regular (zonal) PD will be provisioned.

If replication-type is set to regional-pd, a Regional Persistent Disk will be provisioned. It's highly recommended to have volumeBindingMode: WaitForFirstConsumer set, in which case when you create a Pod that consumes a PersistentVolumeClaim which uses this StorageClass, a Regional Persistent Disk is provisioned with two zones. One zone is the same as the zone that the Pod is scheduled in. The other zone is randomly picked from the zones available to the cluster. Disk zones can be further constrained using allowedTopologies.

NFS

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: example-nfs
provisioner: example.com/external-nfs
parameters:
  server: nfs-server.example.com
  path: /share
  readOnly: "false"

• server: Server is the hostname or IP address of the NFS server.
• path: Path that is exported by the NFS server.
• readOnly: A flag indicating whether the storage will be mounted as read only (default false).

Kubernetes doesn't include an internal NFS provisioner. You need to use an external provisioner to create a StorageClass for NFS. Here are some examples:

• NFS Ganesha server and external provisioner
• NFS subdir external provisioner

vSphere

There are two types of provisioners for vSphere storage classes:

• CSI provisioner: csi.vsphere.vmware.com
• vCP provisioner: kubernetes.io/vsphere-volume

In-tree provisioners are deprecated. For more information on the CSI provisioner, see Kubernetes vSphere CSI Driver and vSphereVolume CSI migration.

CSI Provisioner

The vSphere CSI StorageClass provisioner works with Tanzu Kubernetes clusters. For an example, refer to the vSphere CSI repository.

vCP Provisioner

The following examples use the VMware Cloud Provider (vCP) StorageClass provisioner.

1. Create a StorageClass with a user specified disk format.

   apiVersion: storage.k8s.io/v1
   kind: StorageClass
   metadata:
     name: fast
   provisioner: kubernetes.io/vsphere-volume
   parameters:
     diskformat: zeroedthick
   

   diskformat: thin, zeroedthick and eagerzeroedthick. Default: "thin".

2. Create a StorageClass with a disk format on a user specified datastore.

   apiVersion: storage.k8s.io/v1
   kind: StorageClass
   metadata:
     name: fast
   provisioner: kubernetes.io/vsphere-volume
   parameters:
     diskformat: zeroedthick
     datastore: VSANDatastore
   

   datastore: The user can also specify the datastore in the StorageClass. The volume will be created on the datastore specified in the StorageClass, which in this case is VSANDatastore. This field is optional. If the datastore is not specified, then the volume will be created on the datastore specified in the vSphere config file used to initialize the vSphere Cloud Provider.

3. Storage Policy Management inside kubernetes

   • Using existing vCenter SPBM policy

     One of the most important features of vSphere for Storage Management is policy based Management. Storage Policy Based Management (SPBM) is a storage policy framework that provides a single unified control plane across a broad range of data services and storage solutions. SPBM enables vSphere administrators to overcome upfront storage provisioning challenges, such as capacity planning, differentiated service levels and managing capacity headroom.

     The SPBM policies can be specified in the StorageClass using the storagePolicyName parameter.

   • Virtual SAN policy support inside Kubernetes

     Vsphere Infrastructure (VI) Admins will have the ability to specify custom Virtual SAN Storage Capabilities during dynamic volume provisioning. You can now define storage requirements, such as performance and availability, in the form of storage capabilities during dynamic volume provisioning. The storage capability requirements are converted into a Virtual SAN policy which are then pushed down to the Virtual SAN layer when a persistent volume (virtual disk) is being created. The virtual disk is distributed across the Virtual SAN datastore to meet the requirements.

     You can see Storage Policy Based Management for dynamic provisioning of volumes for more details on how to use storage policies for persistent volumes management.

There are few vSphere examples which you try out for persistent volume management inside Kubernetes for vSphere.

Ceph RBD

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: fast
provisioner: kubernetes.io/rbd
parameters:
  monitors: 10.16.153.105:6789
  adminId: kube
  adminSecretName: ceph-secret
  adminSecretNamespace: kube-system
  pool: kube
  userId: kube
  userSecretName: ceph-secret-user
  userSecretNamespace: default
  fsType: ext4
  imageFormat: "2"
  imageFeatures: "layering"

• monitors: Ceph monitors, comma delimited. This parameter is required.

• adminId: Ceph client ID that is capable of creating images in the pool. Default is "admin".

• adminSecretName: Secret Name for adminId. This parameter is required. The provided secret must have type "kubernetes.io/rbd".

• adminSecretNamespace: The namespace for adminSecretName. Default is "default".

• pool: Ceph RBD pool. Default is "rbd".

• userId: Ceph client ID that is used to map the RBD image. Default is the same as adminId.

• userSecretName: The name of Ceph Secret for userId to map RBD image. It must exist in the same namespace as PVCs. This parameter is required. The provided secret must have type "kubernetes.io/rbd", for example created in this way:

  kubectl create secret generic ceph-secret --type="kubernetes.io/rbd" \
    --from-literal=key='QVFEQ1pMdFhPUnQrSmhBQUFYaERWNHJsZ3BsMmNjcDR6RFZST0E9PQ==' \
    --namespace=kube-system
  

• userSecretNamespace: The namespace for userSecretName.

• fsType: fsType that is supported by kubernetes. Default: "ext4".

• imageFormat: Ceph RBD image format, "1" or "2". Default is "2".

• imageFeatures: This parameter is optional and should only be used if you set imageFormat to "2". Currently supported features are layering only. Default is "", and no features are turned on.

Azure Disk

Azure Unmanaged Disk storage class

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: slow
provisioner: kubernetes.io/azure-disk
parameters:
  skuName: Standard_LRS
  location: eastus
  storageAccount: azure_storage_account_name

• skuName: Azure storage account Sku tier. Default is empty.
• location: Azure storage account location. Default is empty.
• storageAccount: Azure storage account name. If a storage account is provided, it must reside in the same resource group as the cluster, and location is ignored. If a storage account is not provided, a new storage account will be created in the same resource group as the cluster.

Azure Disk storage class (starting from v1.7.2)

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: slow
provisioner: kubernetes.io/azure-disk
parameters:
  storageaccounttype: Standard_LRS
  kind: managed

• storageaccounttype: Azure storage account Sku tier. Default is empty.
• kind: Possible values are shared, dedicated, and managed (default). When kind is shared, all unmanaged disks are created in a few shared storage accounts in the same resource group as the cluster. When kind is dedicated, a new dedicated storage account will be created for the new unmanaged disk in the same resource group as the cluster. When kind is managed, all managed disks are created in the same resource group as the cluster.
• resourceGroup: Specify the resource group in which the Azure disk will be created. It must be an existing resource group name. If it is unspecified, the disk will be placed in the same resource group as the current Kubernetes cluster.

• Premium VM can attach both Standard_LRS and Premium_LRS disks, while Standard VM can only attach Standard_LRS disks.
• Managed VM can only attach managed disks and unmanaged VM can only attach unmanaged disks.

Azure File

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: azurefile
provisioner: kubernetes.io/azure-file
parameters:
  skuName: Standard_LRS
  location: eastus
  storageAccount: azure_storage_account_name

• skuName: Azure storage account Sku tier. Default is empty.
• location: Azure storage account location. Default is empty.
• storageAccount: Azure storage account name. Default is empty. If a storage account is not provided, all storage accounts associated with the resource group are searched to find one that matches skuName and location. If a storage account is provided, it must reside in the same resource group as the cluster, and skuName and location are ignored.
• secretNamespace: the namespace of the secret that contains the Azure Storage Account Name and Key. Default is the same as the Pod.
• secretName: the name of the secret that contains the Azure Storage Account Name and Key. Default is azure-storage-account-<accountName>-secret
• readOnly: a flag indicating whether the storage will be mounted as read only. Defaults to false which means a read/write mount. This setting will impact the ReadOnly setting in VolumeMounts as well.

During storage provisioning, a secret named by secretName is created for the mounting credentials. If the cluster has enabled both RBAC and Controller Roles, add the create permission of resource secret for clusterrole system:controller:persistent-volume-binder.

In a multi-tenancy context, it is strongly recommended to set the value for secretNamespace explicitly, otherwise the storage account credentials may be read by other users.

Portworx Volume

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: portworx-io-priority-high
provisioner: kubernetes.io/portworx-volume
parameters:
  repl: "1"
  snap_interval: "70"
  priority_io: "high"

• fs: filesystem to be laid out: none/xfs/ext4 (default: ext4).
• block_size: block size in Kbytes (default: 32).
• repl: number of synchronous replicas to be provided in the form of replication factor 1..3 (default: 1) A string is expected here i.e. "1" and not 1.
• priority_io: determines whether the volume will be created from higher performance or a lower priority storage high/medium/low (default: low).
• snap_interval: clock/time interval in minutes for when to trigger snapshots. Snapshots are incremental based on difference with the prior snapshot, 0 disables snaps (default: 0). A string is expected here i.e. "70" and not 70.
• aggregation_level: specifies the number of chunks the volume would be distributed into, 0 indicates a non-aggregated volume (default: 0). A string is expected here i.e. "0" and not 0
• ephemeral: specifies whether the volume should be cleaned-up after unmount or should be persistent. emptyDir use case can set this value to true and persistent volumes use case such as for databases like Cassandra should set to false, true/false (default false). A string is expected here i.e. "true" and not true.

Local

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: local-storage
provisioner: kubernetes.io/no-provisioner
volumeBindingMode: WaitForFirstConsumer

Local volumes do not currently support dynamic provisioning, however a StorageClass should still be created to delay volume binding until Pod scheduling. This is specified by the WaitForFirstConsumer volume binding mode.

Delaying volume binding allows the scheduler to consider all of a Pod's scheduling constraints when choosing an appropriate PersistentVolume for a PersistentVolumeClaim.

6 - Dynamic Volume Provisioning

Dynamic volume provisioning allows storage volumes to be created on-demand. Without dynamic provisioning, cluster administrators have to manually make calls to their cloud or storage provider to create new storage volumes, and then create PersistentVolume objects to represent them in Kubernetes. The dynamic provisioning feature eliminates the need for cluster administrators to pre-provision storage. Instead, it automatically provisions storage when it is requested by users.

Background

The implementation of dynamic volume provisioning is based on the API object StorageClass from the API group storage.k8s.io. A cluster administrator can define as many StorageClass objects as needed, each specifying a volume plugin (aka provisioner) that provisions a volume and the set of parameters to pass to that provisioner when provisioning. A cluster administrator can define and expose multiple flavors of storage (from the same or different storage systems) within a cluster, each with a custom set of parameters. This design also ensures that end users don't have to worry about the complexity and nuances of how storage is provisioned, but still have the ability to select from multiple storage options.

More information on storage classes can be found here.

Enabling Dynamic Provisioning

To enable dynamic provisioning, a cluster administrator needs to pre-create one or more StorageClass objects for users. StorageClass objects define which provisioner should be used and what parameters should be passed to that provisioner when dynamic provisioning is invoked. The name of a StorageClass object must be a valid DNS subdomain name.

The following manifest creates a storage class "slow" which provisions standard disk-like persistent disks.

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: slow
provisioner: kubernetes.io/gce-pd
parameters:
  type: pd-standard

The following manifest creates a storage class "fast" which provisions SSD-like persistent disks.

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: fast
provisioner: kubernetes.io/gce-pd
parameters:
  type: pd-ssd

Using Dynamic Provisioning

Users request dynamically provisioned storage by including a storage class in their PersistentVolumeClaim. Before Kubernetes v1.6, this was done via the volume.beta.kubernetes.io/storage-class annotation. However, this annotation is deprecated since v1.9. Users now can and should instead use the storageClassName field of the PersistentVolumeClaim object. The value of this field must match the name of a StorageClass configured by the administrator (see below).

To select the "fast" storage class, for example, a user would create the following PersistentVolumeClaim:

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: claim1
spec:
  accessModes:
    - ReadWriteOnce
  storageClassName: fast
  resources:
    requests:
      storage: 30Gi

This claim results in an SSD-like Persistent Disk being automatically provisioned. When the claim is deleted, the volume is destroyed.

Defaulting Behavior

Dynamic provisioning can be enabled on a cluster such that all claims are dynamically provisioned if no storage class is specified. A cluster administrator can enable this behavior by:

• Marking one StorageClass object as default;
• Making sure that the DefaultStorageClass admission controller is enabled on the API server.

An administrator can mark a specific StorageClass as default by adding the storageclass.kubernetes.io/is-default-class annotation to it. When a default StorageClass exists in a cluster and a user creates a PersistentVolumeClaim with storageClassName unspecified, the DefaultStorageClass admission controller automatically adds the storageClassName field pointing to the default storage class.

Note that there can be at most one default storage class on a cluster, or a PersistentVolumeClaim without storageClassName explicitly specified cannot be created.

Topology Awareness

In Multi-Zone clusters, Pods can be spread across Zones in a Region. Single-Zone storage backends should be provisioned in the Zones where Pods are scheduled. This can be accomplished by setting the Volume Binding Mode.

7 - Volume Snapshots

In Kubernetes, a VolumeSnapshot represents a snapshot of a volume on a storage system. This document assumes that you are already familiar with Kubernetes persistent volumes.

Introduction

Similar to how API resources PersistentVolume and PersistentVolumeClaim are used to provision volumes for users and administrators, VolumeSnapshotContent and VolumeSnapshot API resources are provided to create volume snapshots for users and administrators.

A VolumeSnapshotContent is a snapshot taken from a volume in the cluster that has been provisioned by an administrator. It is a resource in the cluster just like a PersistentVolume is a cluster resource.

A VolumeSnapshot is a request for snapshot of a volume by a user. It is similar to a PersistentVolumeClaim.

VolumeSnapshotClass allows you to specify different attributes belonging to a VolumeSnapshot. These attributes may differ among snapshots taken from the same volume on the storage system and therefore cannot be expressed by using the same StorageClass of a PersistentVolumeClaim.

Volume snapshots provide Kubernetes users with a standardized way to copy a volume's contents at a particular point in time without creating an entirely new volume. This functionality enables, for example, database administrators to backup databases before performing edit or delete modifications.

Users need to be aware of the following when using this feature:

• API Objects VolumeSnapshot, VolumeSnapshotContent, and VolumeSnapshotClass are CRDs, not part of the core API.
• VolumeSnapshot support is only available for CSI drivers.
• As part of the deployment process of VolumeSnapshot, the Kubernetes team provides a snapshot controller to be deployed into the control plane, and a sidecar helper container called csi-snapshotter to be deployed together with the CSI driver. The snapshot controller watches VolumeSnapshot and VolumeSnapshotContent objects and is responsible for the creation and deletion of VolumeSnapshotContent object. The sidecar csi-snapshotter watches VolumeSnapshotContent objects and triggers CreateSnapshot and DeleteSnapshot operations against a CSI endpoint.
• There is also a validating webhook server which provides tightened validation on snapshot objects. This should be installed by the Kubernetes distros along with the snapshot controller and CRDs, not CSI drivers. It should be installed in all Kubernetes clusters that has the snapshot feature enabled.
• CSI drivers may or may not have implemented the volume snapshot functionality. The CSI drivers that have provided support for volume snapshot will likely use the csi-snapshotter. See CSI Driver documentation for details.
• The CRDs and snapshot controller installations are the responsibility of the Kubernetes distribution.

Lifecycle of a volume snapshot and volume snapshot content

VolumeSnapshotContents are resources in the cluster. VolumeSnapshots are requests for those resources. The interaction between VolumeSnapshotContents and VolumeSnapshots follow this lifecycle:

Provisioning Volume Snapshot

There are two ways snapshots may be provisioned: pre-provisioned or dynamically provisioned.

Pre-provisioned

A cluster administrator creates a number of VolumeSnapshotContents. They carry the details of the real volume snapshot on the storage system which is available for use by cluster users. They exist in the Kubernetes API and are available for consumption.

Dynamic

Instead of using a pre-existing snapshot, you can request that a snapshot to be dynamically taken from a PersistentVolumeClaim. The VolumeSnapshotClass specifies storage provider-specific parameters to use when taking a snapshot.

Binding

The snapshot controller handles the binding of a VolumeSnapshot object with an appropriate VolumeSnapshotContent object, in both pre-provisioned and dynamically provisioned scenarios. The binding is a one-to-one mapping.

In the case of pre-provisioned binding, the VolumeSnapshot will remain unbound until the requested VolumeSnapshotContent object is created.

Persistent Volume Claim as Snapshot Source Protection

The purpose of this protection is to ensure that in-use PersistentVolumeClaim API objects are not removed from the system while a snapshot is being taken from it (as this may result in data loss).

While a snapshot is being taken of a PersistentVolumeClaim, that PersistentVolumeClaim is in-use. If you delete a PersistentVolumeClaim API object in active use as a snapshot source, the PersistentVolumeClaim object is not removed immediately. Instead, removal of the PersistentVolumeClaim object is postponed until the snapshot is readyToUse or aborted.

Delete

Deletion is triggered by deleting the VolumeSnapshot object, and the DeletionPolicy will be followed. If the DeletionPolicy is Delete, then the underlying storage snapshot will be deleted along with the VolumeSnapshotContent object. If the DeletionPolicy is Retain, then both the underlying snapshot and VolumeSnapshotContent remain.

VolumeSnapshots

Each VolumeSnapshot contains a spec and a status.

apiVersion: snapshot.storage.k8s.io/v1
kind: VolumeSnapshot
metadata:
  name: new-snapshot-test
spec:
  volumeSnapshotClassName: csi-hostpath-snapclass
  source:
    persistentVolumeClaimName: pvc-test

persistentVolumeClaimName is the name of the PersistentVolumeClaim data source for the snapshot. This field is required for dynamically provisioning a snapshot.

A volume snapshot can request a particular class by specifying the name of a VolumeSnapshotClass using the attribute volumeSnapshotClassName. If nothing is set, then the default class is used if available.

For pre-provisioned snapshots, you need to specify a volumeSnapshotContentName as the source for the snapshot as shown in the following example. The volumeSnapshotContentName source field is required for pre-provisioned snapshots.

apiVersion: snapshot.storage.k8s.io/v1
kind: VolumeSnapshot
metadata:
  name: test-snapshot
spec:
  source:
    volumeSnapshotContentName: test-content

Volume Snapshot Contents

Each VolumeSnapshotContent contains a spec and status. In dynamic provisioning, the snapshot common controller creates VolumeSnapshotContent objects. Here is an example:

apiVersion: snapshot.storage.k8s.io/v1
kind: VolumeSnapshotContent
metadata:
  name: snapcontent-72d9a349-aacd-42d2-a240-d775650d2455
spec:
  deletionPolicy: Delete
  driver: hostpath.csi.k8s.io
  source:
    volumeHandle: ee0cfb94-f8d4-11e9-b2d8-0242ac110002
  sourceVolumeMode: Filesystem
  volumeSnapshotClassName: csi-hostpath-snapclass
  volumeSnapshotRef:
    name: new-snapshot-test
    namespace: default
    uid: 72d9a349-aacd-42d2-a240-d775650d2455

volumeHandle is the unique identifier of the volume created on the storage backend and returned by the CSI driver during the volume creation. This field is required for dynamically provisioning a snapshot. It specifies the volume source of the snapshot.

For pre-provisioned snapshots, you (as cluster administrator) are responsible for creating the VolumeSnapshotContent object as follows.

apiVersion: snapshot.storage.k8s.io/v1
kind: VolumeSnapshotContent
metadata:
  name: new-snapshot-content-test
spec:
  deletionPolicy: Delete
  driver: hostpath.csi.k8s.io
  source:
    snapshotHandle: 7bdd0de3-aaeb-11e8-9aae-0242ac110002
  sourceVolumeMode: Filesystem
  volumeSnapshotRef:
    name: new-snapshot-test
    namespace: default

snapshotHandle is the unique identifier of the volume snapshot created on the storage backend. This field is required for the pre-provisioned snapshots. It specifies the CSI snapshot id on the storage system that this VolumeSnapshotContent represents.

sourceVolumeMode is the mode of the volume whose snapshot is taken. The value of the sourceVolumeMode field can be either Filesystem or Block. If the source volume mode is not specified, Kubernetes treats the snapshot as if the source volume's mode is unknown.

volumeSnapshotRef is the reference of the corresponding VolumeSnapshot. Note that when the VolumeSnapshotContent is being created as a pre-provisioned snapshot, the VolumeSnapshot referenced in volumeSnapshotRef might not exist yet.

Converting the volume mode of a Snapshot

If the VolumeSnapshots API installed on your cluster supports the sourceVolumeMode field, then the API has the capability to prevent unauthorized users from converting the mode of a volume.

To check if your cluster has capability for this feature, run the following command:

$ kubectl get crd volumesnapshotcontent -o yaml

If you want to allow users to create a PersistentVolumeClaim from an existing VolumeSnapshot, but with a different volume mode than the source, the annotation snapshot.storage.kubernetes.io/allow-volume-mode-change: "true"needs to be added to the VolumeSnapshotContent that corresponds to the VolumeSnapshot.

For pre-provisioned snapshots, spec.sourceVolumeMode needs to be populated by the cluster administrator.

An example VolumeSnapshotContent resource with this feature enabled would look like:

apiVersion: snapshot.storage.k8s.io/v1
kind: VolumeSnapshotContent
metadata:
  name: new-snapshot-content-test
  annotations:
    - snapshot.storage.kubernetes.io/allow-volume-mode-change: "true"
spec:
  deletionPolicy: Delete
  driver: hostpath.csi.k8s.io
  source:
    snapshotHandle: 7bdd0de3-aaeb-11e8-9aae-0242ac110002
  sourceVolumeMode: Filesystem
  volumeSnapshotRef:
    name: new-snapshot-test
    namespace: default

Provisioning Volumes from Snapshots

You can provision a new volume, pre-populated with data from a snapshot, by using the dataSource field in the PersistentVolumeClaim object.

For more details, see Volume Snapshot and Restore Volume from Snapshot.

8 - Volume Snapshot Classes

This document describes the concept of VolumeSnapshotClass in Kubernetes. Familiarity with volume snapshots and storage classes is suggested.

Introduction

Just like StorageClass provides a way for administrators to describe the "classes" of storage they offer when provisioning a volume, VolumeSnapshotClass provides a way to describe the "classes" of storage when provisioning a volume snapshot.

The VolumeSnapshotClass Resource

Each VolumeSnapshotClass contains the fields driver, deletionPolicy, and parameters, which are used when a VolumeSnapshot belonging to the class needs to be dynamically provisioned.

The name of a VolumeSnapshotClass object is significant, and is how users can request a particular class. Administrators set the name and other parameters of a class when first creating VolumeSnapshotClass objects, and the objects cannot be updated once they are created.

apiVersion: snapshot.storage.k8s.io/v1
kind: VolumeSnapshotClass
metadata:
  name: csi-hostpath-snapclass
driver: hostpath.csi.k8s.io
deletionPolicy: Delete
parameters:

Administrators can specify a default VolumeSnapshotClass for VolumeSnapshots that don't request any particular class to bind to by adding the snapshot.storage.kubernetes.io/is-default-class: "true" annotation:

apiVersion: snapshot.storage.k8s.io/v1
kind: VolumeSnapshotClass
metadata:
  name: csi-hostpath-snapclass
  annotations:
    snapshot.storage.kubernetes.io/is-default-class: "true"
driver: hostpath.csi.k8s.io
deletionPolicy: Delete
parameters:

Driver

Volume snapshot classes have a driver that determines what CSI volume plugin is used for provisioning VolumeSnapshots. This field must be specified.

DeletionPolicy

Volume snapshot classes have a deletionPolicy. It enables you to configure what happens to a VolumeSnapshotContent when the VolumeSnapshot object it is bound to is to be deleted. The deletionPolicy of a volume snapshot class can either be Retain or Delete. This field must be specified.

If the deletionPolicy is Delete, then the underlying storage snapshot will be deleted along with the VolumeSnapshotContent object. If the deletionPolicy is Retain, then both the underlying snapshot and VolumeSnapshotContent remain.

Parameters

Volume snapshot classes have parameters that describe volume snapshots belonging to the volume snapshot class. Different parameters may be accepted depending on the driver.

9 - CSI Volume Cloning

This document describes the concept of cloning existing CSI Volumes in Kubernetes. Familiarity with Volumes is suggested.

Introduction

The CSI Volume Cloning feature adds support for specifying existing PVCs in the dataSource field to indicate a user would like to clone a Volume.

A Clone is defined as a duplicate of an existing Kubernetes Volume that can be consumed as any standard Volume would be. The only difference is that upon provisioning, rather than creating a "new" empty Volume, the back end device creates an exact duplicate of the specified Volume.

The implementation of cloning, from the perspective of the Kubernetes API, adds the ability to specify an existing PVC as a dataSource during new PVC creation. The source PVC must be bound and available (not in use).

Users need to be aware of the following when using this feature:

• Cloning support (VolumePVCDataSource) is only available for CSI drivers.
• Cloning support is only available for dynamic provisioners.
• CSI drivers may or may not have implemented the volume cloning functionality.
• You can only clone a PVC when it exists in the same namespace as the destination PVC (source and destination must be in the same namespace).
• Cloning is supported with a different Storage Class.
  • Destination volume can be the same or a different storage class as the source.
  • Default storage class can be used and storageClassName omitted in the spec.
• Cloning can only be performed between two volumes that use the same VolumeMode setting (if you request a block mode volume, the source MUST also be block mode)

Provisioning

Clones are provisioned like any other PVC with the exception of adding a dataSource that references an existing PVC in the same namespace.

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
    name: clone-of-pvc-1
    namespace: myns
spec:
  accessModes:
  - ReadWriteOnce
  storageClassName: cloning
  resources:
    requests:
      storage: 5Gi
  dataSource:
    kind: PersistentVolumeClaim
    name: pvc-1

The result is a new PVC with the name clone-of-pvc-1 that has the exact same content as the specified source pvc-1.

Usage

Upon availability of the new PVC, the cloned PVC is consumed the same as other PVC. It's also expected at this point that the newly created PVC is an independent object. It can be consumed, cloned, snapshotted, or deleted independently and without consideration for it's original dataSource PVC. This also implies that the source is not linked in any way to the newly created clone, it may also be modified or deleted without affecting the newly created clone.

10 - Storage Capacity

Storage capacity is limited and may vary depending on the node on which a pod runs: network-attached storage might not be accessible by all nodes, or storage is local to a node to begin with.

This page describes how Kubernetes keeps track of storage capacity and how the scheduler uses that information to schedule Pods onto nodes that have access to enough storage capacity for the remaining missing volumes. Without storage capacity tracking, the scheduler may choose a node that doesn't have enough capacity to provision a volume and multiple scheduling retries will be needed.

Before you begin

Kubernetes v1.28 includes cluster-level API support for storage capacity tracking. To use this you must also be using a CSI driver that supports capacity tracking. Consult the documentation for the CSI drivers that you use to find out whether this support is available and, if so, how to use it. If you are not running Kubernetes v1.28, check the documentation for that version of Kubernetes.

API

There are two API extensions for this feature:

• CSIStorageCapacity objects: these get produced by a CSI driver in the namespace where the driver is installed. Each object contains capacity information for one storage class and defines which nodes have access to that storage.
• The CSIDriverSpec.StorageCapacity field: when set to true, the Kubernetes scheduler will consider storage capacity for volumes that use the CSI driver.

Scheduling

Storage capacity information is used by the Kubernetes scheduler if:

• a Pod uses a volume that has not been created yet,
• that volume uses a StorageClass which references a CSI driver and uses WaitForFirstConsumer volume binding mode, and
• the CSIDriver object for the driver has StorageCapacity set to true.

In that case, the scheduler only considers nodes for the Pod which have enough storage available to them. This check is very simplistic and only compares the size of the volume against the capacity listed in CSIStorageCapacity objects with a topology that includes the node.

For volumes with Immediate volume binding mode, the storage driver decides where to create the volume, independently of Pods that will use the volume. The scheduler then schedules Pods onto nodes where the volume is available after the volume has been created.

For CSI ephemeral volumes, scheduling always happens without considering storage capacity. This is based on the assumption that this volume type is only used by special CSI drivers which are local to a node and do not need significant resources there.

Rescheduling

When a node has been selected for a Pod with WaitForFirstConsumer volumes, that decision is still tentative. The next step is that the CSI storage driver gets asked to create the volume with a hint that the volume is supposed to be available on the selected node.

Because Kubernetes might have chosen a node based on out-dated capacity information, it is possible that the volume cannot really be created. The node selection is then reset and the Kubernetes scheduler tries again to find a node for the Pod.

Limitations

Storage capacity tracking increases the chance that scheduling works on the first try, but cannot guarantee this because the scheduler has to decide based on potentially out-dated information. Usually, the same retry mechanism as for scheduling without any storage capacity information handles scheduling failures.

One situation where scheduling can fail permanently is when a Pod uses multiple volumes: one volume might have been created already in a topology segment which then does not have enough capacity left for another volume. Manual intervention is necessary to recover from this, for example by increasing capacity or deleting the volume that was already created.

What's next

• For more information on the design, see the Storage Capacity Constraints for Pod Scheduling KEP.

11 - Node-specific Volume Limits

This page describes the maximum number of volumes that can be attached to a Node for various cloud providers.

Cloud providers like Google, Amazon, and Microsoft typically have a limit on how many volumes can be attached to a Node. It is important for Kubernetes to respect those limits. Otherwise, Pods scheduled on a Node could get stuck waiting for volumes to attach.

Kubernetes default limits

The Kubernetes scheduler has default limits on the number of volumes that can be attached to a Node:

Custom limits

You can change these limits by setting the value of the KUBE_MAX_PD_VOLS environment variable, and then starting the scheduler. CSI drivers might have a different procedure, see their documentation on how to customize their limits.

Use caution if you set a limit that is higher than the default limit. Consult the cloud provider's documentation to make sure that Nodes can actually support the limit you set.

The limit applies to the entire cluster, so it affects all Nodes.

Dynamic volume limits

Dynamic volume limits are supported for following volume types.

• Amazon EBS
• Google Persistent Disk
• Azure Disk
• CSI

For volumes managed by in-tree volume plugins, Kubernetes automatically determines the Node type and enforces the appropriate maximum number of volumes for the node. For example:

• On Google Compute Engine, up to 127 volumes can be attached to a node, depending on the node type.

• For Amazon EBS disks on M5,C5,R5,T3 and Z1D instance types, Kubernetes allows only 25 volumes to be attached to a Node. For other instance types on Amazon Elastic Compute Cloud (EC2), Kubernetes allows 39 volumes to be attached to a Node.

• On Azure, up to 64 disks can be attached to a node, depending on the node type. For more details, refer to Sizes for virtual machines in Azure.

• If a CSI storage driver advertises a maximum number of volumes for a Node (using NodeGetInfo), the kube-scheduler honors that limit. Refer to the CSI specifications for details.

• For volumes managed by in-tree plugins that have been migrated to a CSI driver, the maximum number of volumes will be the one reported by the CSI driver.

12 - Volume Health Monitoring

CSI volume health monitoring allows CSI Drivers to detect abnormal volume conditions from the underlying storage systems and report them as events on PVCs or Pods.

Volume health monitoring

Kubernetes volume health monitoring is part of how Kubernetes implements the Container Storage Interface (CSI). Volume health monitoring feature is implemented in two components: an External Health Monitor controller, and the kubelet.

If a CSI Driver supports Volume Health Monitoring feature from the controller side, an event will be reported on the related PersistentVolumeClaim (PVC) when an abnormal volume condition is detected on a CSI volume.

The External Health Monitor controller also watches for node failure events. You can enable node failure monitoring by setting the enable-node-watcher flag to true. When the external health monitor detects a node failure event, the controller reports an Event will be reported on the PVC to indicate that pods using this PVC are on a failed node.

If a CSI Driver supports Volume Health Monitoring feature from the node side, an Event will be reported on every Pod using the PVC when an abnormal volume condition is detected on a CSI volume. In addition, Volume Health information is exposed as Kubelet VolumeStats metrics. A new metric kubelet_volume_stats_health_status_abnormal is added. This metric includes two labels: namespace and persistentvolumeclaim. The count is either 1 or 0. 1 indicates the volume is unhealthy, 0 indicates volume is healthy. For more information, please check KEP.

What's next

See the CSI driver documentation to find out which CSI drivers have implemented this feature.

13 - Windows Storage

This page provides an storage overview specific to the Windows operating system.

Persistent storage

Windows has a layered filesystem driver to mount container layers and create a copy filesystem based on NTFS. All file paths in the container are resolved only within the context of that container.

• With Docker, volume mounts can only target a directory in the container, and not an individual file. This limitation does not apply to containerd.
• Volume mounts cannot project files or directories back to the host filesystem.
• Read-only filesystems are not supported because write access is always required for the Windows registry and SAM database. However, read-only volumes are supported.
• Volume user-masks and permissions are not available. Because the SAM is not shared between the host & container, there's no mapping between them. All permissions are resolved within the context of the container.

As a result, the following storage functionality is not supported on Windows nodes:

• Volume subpath mounts: only the entire volume can be mounted in a Windows container
• Subpath volume mounting for Secrets
• Host mount projection
• Read-only root filesystem (mapped volumes still support readOnly)
• Block device mapping
• Memory as the storage medium (for example, emptyDir.medium set to Memory)
• File system features like uid/gid; per-user Linux filesystem permissions
• Setting secret permissions with DefaultMode (due to UID/GID dependency)
• NFS based storage/volume support
• Expanding the mounted volume (resizefs)

Kubernetes volumes enable complex applications, with data persistence and Pod volume sharing requirements, to be deployed on Kubernetes. Management of persistent volumes associated with a specific storage back-end or protocol includes actions such as provisioning/de-provisioning/resizing of volumes, attaching/detaching a volume to/from a Kubernetes node and mounting/dismounting a volume to/from individual containers in a pod that needs to persist data.

Volume management components are shipped as Kubernetes volume plugin. The following broad classes of Kubernetes volume plugins are supported on Windows:

• FlexVolume plugins
  • Please note that FlexVolumes have been deprecated as of 1.23
• CSI Plugins

In-tree volume plugins

The following in-tree plugins support persistent storage on Windows nodes:

• azureFile
• gcePersistentDisk
• vsphereVolume