Black lives matter.
We stand in solidarity with the Black community.
Racism is unacceptable.
It conflicts with the core values of the Kubernetes project and our community does not tolerate it.
We stand in solidarity with the Black community.
Racism is unacceptable.
It conflicts with the core values of the Kubernetes project and our community does not tolerate it.
This document describes the current state of persistent volumes in Kubernetes. Familiarity with volumes is suggested.
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 once read/write or many times read-only).
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 just 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.
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:
There are two ways PVs may be provisioned: statically or dynamically.
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.
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.
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 master 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.
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.
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.
Note: PVC is in active use by a Pod when a Pod object exists that is using the PVC.
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>
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.
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.
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, GCE PD, Azure Disk, or Cinder 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.
Warning: TheRecycle
reclaim policy is deprecated. Instead, the recommended approach is to use dynamic provisioning.
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 here. 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: "k8s.gcr.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.
Kubernetes v1.11 [beta]
Support for expanding PersistentVolumeClaims (PVCs) is now enabled by default. You can expand the following types of volumes:
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: gluster-vol-default
provisioner: kubernetes.io/glusterfs
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.
Kubernetes v1.16 [beta]
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.
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 allow resize if the driver is set with the RequiresFSResize
capability to true
.
The FlexVolume can be resized on Pod restart.
Kubernetes v1.15 [beta]
Note: Expanding in-use PVCs is available as beta since Kubernetes 1.15, and as alpha since 1.11. TheExpandInUsePersistentVolumes
feature must be enabled, which is the case automatically for many clusters for beta features. Refer to the feature gate documentation for more information.
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.
Note: FlexVolume resize is possible only when the underlying driver supports resize.
Note: Expanding EBS volumes is a time-consuming operation. Also, there is a per-volume quota of one modification every 6 hours.
PersistentVolume types are implemented as plugins. Kubernetes currently supports the following plugins:
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
Note: Helper programs relating to the volume type may be required for consumption of a PersistentVolume within a cluster. In this example, the PersistentVolume is of type NFS and the helper program /sbin/mount.nfs is required to support the mounting of NFS filesystems.
Generally, a PV will have a specific storage capacity. This is set using the PV's capacity
attribute. See the Kubernetes Resource Model 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.
Kubernetes v1.18 [stable]
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, Kuberneretes 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.
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:
In the CLI, the access modes are abbreviated to:
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.
Volume Plugin | ReadWriteOnce | ReadOnlyMany | ReadWriteMany |
---|---|---|---|
AWSElasticBlockStore | ✓ | - | - |
AzureFile | ✓ | ✓ | ✓ |
AzureDisk | ✓ | - | - |
CephFS | ✓ | ✓ | ✓ |
Cinder | ✓ | - | - |
CSI | depends on the driver | depends on the driver | depends on the driver |
FC | ✓ | ✓ | - |
FlexVolume | ✓ | ✓ | depends on the driver |
Flocker | ✓ | - | - |
GCEPersistentDisk | ✓ | ✓ | - |
Glusterfs | ✓ | ✓ | ✓ |
HostPath | ✓ | - | - |
iSCSI | ✓ | ✓ | - |
Quobyte | ✓ | ✓ | ✓ |
NFS | ✓ | ✓ | ✓ |
RBD | ✓ | ✓ | - |
VsphereVolume | ✓ | - | - (works when Pods are collocated) |
PortworxVolume | ✓ | - | ✓ |
ScaleIO | ✓ | ✓ | - |
StorageOS | ✓ | - | - |
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.
Current reclaim policies are:
rm -rf /thevolume/*
)Currently, only NFS and HostPath support recycling. AWS EBS, GCE PD, Azure Disk, and Cinder volumes support deletion.
A Kubernetes administrator can specify additional mount options for when a Persistent Volume is mounted on a node.
Note: Not all Persistent Volume types support mount options.
The following volume types support mount options:
Mount options are not validated, so mount will simply fail if one is invalid.
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.
Note: For most volume types, you do not need to set this field. It is automatically populated for AWS EBS, GCE PD and Azure Disk volume block types. You need to explicitly set this for local volumes.
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.
A volume will be in one of the following phases:
The CLI will show the name of the PVC bound to the PV.
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]}
Claims use the same conventions as volumes when requesting storage with specific access modes.
Claims use the same convention as volumes to indicate the consumption of the volume as either a filesystem or block device.
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.
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 valuematchExpressions
- 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.
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.
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 is specified, the admission plugin forbids the creation of
all PVCs.storageClassName
can be bound only to PVs that
have no class. In this case, the PVCs that have no storageClassName
are treated the
same way as PVCs that have their storageClassName
set to ""
.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.
Note: Currently, a PVC with a non-emptyselector
can't have a PV dynamically provisioned for it.
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.
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
PersistentVolumes binds are exclusive, and since PersistentVolumeClaims are namespaced objects, mounting claims with "Many" modes (ROX
, RWX
) is only possible within one namespace.
Kubernetes v1.18 [stable]
The following volume plugins support raw block volumes, including dynamic provisioning where applicable:
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
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
name: block-pvc
spec:
accessModes:
- ReadWriteOnce
volumeMode: Block
resources:
requests:
storage: 10Gi
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
Note: When adding a raw block device for a Pod, you specify the device path in the container instead of a mount path.
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:
PV volumeMode | PVC volumeMode | Result |
---|---|---|
unspecified | unspecified | BIND |
unspecified | Block | NO BIND |
unspecified | Filesystem | BIND |
Block | unspecified | NO BIND |
Block | Block | BIND |
Block | Filesystem | NO BIND |
Filesystem | Filesystem | BIND |
Filesystem | Block | NO BIND |
Filesystem | unspecified | BIND |
Note: Only statically provisioned volumes are supported for alpha release. Administrators should take care to consider these values when working with raw block devices.
Kubernetes v1.17 [beta]
Volume snapshot feature was added to support CSI Volume Plugins only. For details, see volume snapshots.
To enable support for restoring a volume from a volume snapshot data source, enable the
VolumeSnapshotDataSource
feature gate on the apiserver and controller-manager.
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 only available for CSI volume plugins.
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
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.
persistentVolumeClaim.storageClassName
field.
This will cause the PVC to match the right storage
class if the cluster has StorageClasses enabled by the admin.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).