This is the multi-page printable view of this section. Click here to print.

Return to the regular view of this page.

Configuring PVs

1: Configuring Dynamic PVs
2: Configuring a Static PV
3: Managing PVs

Huawei CSI allows storage resources (LUNs or file systems) to be created on Huawei storage and provided for containers based on user settings. For details about the supported features, see Compatibility and Features of the storage device.

PV configurations can be classified into configuring dynamic PVs or static PVs, and managing PVs.

Configuring dynamic PV does not require a PV to be created in advance. Huawei CSI automatically creates resources required by a PV on storage devices based on a StorageClass. In addition, you can create a PV when creating a PVC.
Configuring a static PV requires the administrator to create required resources on a storage device in advance and use existing resources by creating a PV. In addition, you can specify the associated PV when creating a PVC.
Managing PVs does not require a PV to be created in advance. You can specify the StorageClass and resource information on the storage device in a PVC. Create the PVC and PV at the same time, and manage all existing storage resources in the cluster.

1 - Configuring Dynamic PVs

Dynamic volume provisioning allows storage volumes to be created on demand. Dynamic volume provisioning depends on the StorageClass objects. The cluster administrator can define multiple StorageClass objects as required and specify a StorageClass that meets service requirements when declaring a PV or PVC. When applying for resources from Huawei storage devices, Huawei CSI creates storage resources that meet service requirements based on the preset StorageClass.

Configuration Description

Perform the following steps to configure and use dynamic PVs:

Preparation
Configuring a PVC
Using a PVC

Preparation

Before configuring dynamic PVs, configure StorageClass by referring to Configuring a StorageClass .

Configuring a PVC

Prepare the PVC configuration file mypv.yaml. The following is an example. For details about other parameters, see Table 1 .

kind: PersistentVolumeClaim
apiVersion: v1
metadata:
  name: mypvc
spec:
  accessModes:
    - ReadWriteOnce
  volumeMode: Filesystem
  storageClassName: mysc
  resources:
    requests:
      storage: 100Gi

Run the following command to create a PVC using the configuration file.
```
kubectl create -f mypvc.yaml
```
After a period of time, run the following command to view the information about the created PVC.
```
kubectl get pvc mypvc
```
The following is an example of the command output. If the PVC status is Bound, the PVC has been created and can be used by a Pod.
```
NAME        STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS   AGE
mypvc       Bound    pvc-840054d3-1d5b-4153-b73f-826f980abf9e   100Gi      RWO            mysc           12s
```
- After the PVC is created, if the PVC is in the Pending state after a long time (for example, one minute), refer to When a PVC Is Created, the PVC Is in the Pending State .
- You are advised to create or delete a maximum of 100 PVCs in a batch.

Table 1 PVC parameters for configuring a dynamic PV

Parameter	Description	Mandatory	Default Value	Remarks
metadata.name	User-defined name of a PVC object.	Yes	-	Take Kubernetes v1.22.1 as an example. The value can contain digits, lowercase letters, hyphens (-), and periods (.), and must start and end with a letter or digit.
spec.volumeMode	Volume mode. This parameter is optional. When LUN volumes are used, the following types are supported: Filesystem: local file system. Block: raw device.	No	Filesystem	This parameter takes effect when a PV is mounted. The default value is Filesystem. Filesystem indicates that a container accesses a PV using a local file system. The local file system type is specified by the fsType field in the specified StorageClass. Storage of the Dtree type also uses this parameter. Block indicates that a PV is accessed in raw volume mode.
spec.storageClassName	Name of the StorageClass object.	Yes	-	Name of the StorageClass object required by services.
spec.resources.requests.storage	Size of the volume to be created. The format is ***Gi and the unit is GiB.	Yes	10Gi	The PVC capacity depends on storage specifications and host specifications. For example, OceanStor Dorado 6.1.2 or OceanStor Pacific series 8.1.0 is connected to CentOS 7. If ext4 file systems are used, see Table 2. If XFS file systems are used, see Table 3. If NFS or raw devices are used, the capacity must meet the specifications of the used Huawei storage device model and version. If the PVC capacity does not meet the specifications, a PVC or Pod may fail to be created due to the limitations of storage specifications or host file system specifications.
spec.accessModes	Access mode of the volume. RWO (ReadWriteOnce): A volume can be mounted to a node in read/write mode. This mode also allows multiple Pods running on the same node to access the volume. ROX (ReadOnlyMany): A volume can be mounted to multiple nodes in read-only mode. RWX (ReadWriteMany): A volume can be mounted to multiple nodes in read/write mode. RWOP (ReadWriteOncePod): A volume can only be mounted to a single Pod in read/write mode. Kubernetes 1.22 and later versions support this feature.	Yes	ReadWriteOnce	RWO/ROX/RWOP: supported by all types of volumes. RWOP is supported only by Kubernetes 1.22 and later versions. For versions earlier than Kubernetes 1.29, you need to enable this feature by following the instructions in Enabling the ReadWriteOncePod Feature Gate. The support for RWX is as follows: NAS storage: supported by all volumes SAN storage: supported only by volumes whose volumeMode is set to Block

Table 2 ext4 capacity specifications

Storage Type	Storage Specifications	ext4 Specifications	CSI Specifications
OceanStor Dorado	512 Ki to 256 Ti	50 Ti	512 Ki to 50 Ti
OceanStor Pacific series	64 Mi to 512 Ti	50 Ti	64 Mi to 50 Ti
OceanDisk	512 Ki to 256 Ti	50 Ti	512 Ki to 50 Ti

Table 3 XFS capacity specifications

Storage Type	Storage Specifications	XFS Specifications	CSI Specifications
OceanStor Dorado	512 Ki to 256 Ti	500 Ti	512 Ki to 256 Ti
OceanStor Pacific series	64 Mi to 512 Ti	500 Ti	64 Mi to 500 Ti
OceanDisk	512 Ki to 256 Ti	500 Ti	512 Ki to 256 Ti

Using a PVC

After a PVC is created, you can use the PVC to create a Pod. The following is a simple example of using a PVC. In this example, the created Pod uses the newly created mypvc.

apiVersion: apps/v1
kind: Deployment
metadata:
  name: nginx-deployment
spec:
  selector:
    matchLabels:
      app: nginx
  replicas: 2
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers: 
      - image: nginx:alpine
        name: container-0 
        volumeMounts: 
        - mountPath: /tmp
          name: pvc-mypvc 
      restartPolicy: Always 
      volumes: 
      - name: pvc-mypvc 
        persistentVolumeClaim: 
          claimName:  mypvc  # name of PVC

If Pods are batch created using PVCs, the Pods are in the ContainerCreating status for a long time, and the huawei-csi-node service is in the OOMKilled status, the memory of the huawei-csi-node service is insufficient. Increase the memory limit of huawei-csi-node by referring to Table 1 .

2 - Configuring a Static PV

Static volume provisioning allows administrators to use a resource created on the storage side as a PV for containers in the cluster.

Configuration Description

Perform the following steps to configure a static PV:

Preparation
Configuring a PV
Configuring a PVC
Using a PVC

Preparation

A storage resource, such as a LUN or file system, required by the PV to be created exists on the storage device. If the storage resource is a file system, you also need to create the share and client information of the file system.

Configuring a PV

Prepare the PV configuration file mypv.yaml. The following is an example. For details about other parameters, see Table 1 .

kind: PersistentVolume
apiVersion: v1
metadata:
  name: mypv
spec:
  volumeMode: Filesystem
  storageClassName: "" # The value must be to "".
  accessModes:
    - ReadWriteOnce
  csi:
    driver: csi.huawei.com # Enter the CSI driver name.
    volumeHandle: iscsi-dorado-181.lun0001 # Enter the volume name.
    fsType: xfs # Set the file system type.
  capacity:
    storage: 100Gi

In the configuration file for static volume provisioning, storageClassName must be set to "". Otherwise, Kubernetes will use the default StorageClass.

Run the following command to create a PV based on the prepared .yaml file.
```
kubectl create -f mypv.yaml
```

After a period of time, run the following command to view the information about the created PV.

kubectl get pv

The following is an example of the command output. If the PV status is Available, the PV is successfully created.

NAME       CAPACITY   ACCESS MODES   RECLAIM POLICY   STATUS      CLAIM               STORAGECLASS   REASON   AGE
mypv       100Gi      RWO            Retain           Available                                               4s

Table 1 Parameters for configuring a static PV

Parameter	Description	Mandatory	Default Value	Remarks
metadata.name	User-defined name of a PV object.	Yes	-	Take Kubernetes v1.22.1 as an example. The value can contain digits, lowercase letters, hyphens (-), and periods (.), and must start and end with a letter or digit.
spec.volumeMode	Volume mode. This parameter is optional. When LUN volumes are used, the following types are supported: Filesystem: local file system. Block: raw device.	No	Filesystem	This parameter takes effect when a PV is mounted. The default value is Filesystem. Filesystem indicates that a container accesses a PV using a local file system. The local file system type is specified by the fsType field in the specified StorageClass. Block indicates that a PV is accessed in raw volume mode.
spec.storageClassName	Name of the StorageClass object. This parameter is mandatory.	Yes	-	Set the parameter to an empty string, that is, enter "".
spec.accessModes	Access mode of the volume. RWO (ReadWriteOnce): A volume can be mounted to a node in read/write mode. This mode also allows multiple Pods running on the same node to access the volume. ROX (ReadOnlyMany): A volume can be mounted to multiple nodes in read-only mode. RWX (ReadWriteMany): A volume can be mounted to multiple nodes in read/write mode. RWOP (ReadWriteOncePod): A volume can only be mounted to a single Pod in read/write mode. Kubernetes 1.22 and later versions support this feature.	Yes	ReadWriteOnce	RWO/ROX/RWOP: supported by all types of volumes. RWOP is supported only by Kubernetes 1.22 and later versions. Check whether this feature is enabled for your Kubernetes cluster by referring to Enabling the ReadWriteOncePod Feature Gate. The support for RWX is as follows: NAS storage: supported by all volumes SAN storage: supported only by volumes whose volumeMode is set to Block
spec.csi.driver	CSI driver name.	Yes	csi.huawei.com	Set this parameter to the driver name set during Huawei CSI installation.
spec.csi.volumeHandle	Unique identifier of a storage resource. This parameter is mandatory. Format: <backendName>.<volume-name>	Yes	-	The value of this parameter consists of the following parts: <backendName>: indicates the name of the backend where the volume resides. You can run the following command to obtain the configured backend information. oceanctl get backend <volume-name>: indicates the name of a resource (LUN/file system) on the storage. You can obtain the value from DeviceManager.
spec.csi.fsType	Type of a host file system. This parameter is optional. The supported types are: ext2 ext3 ext4 xfs	No	-	If this parameter is not set, the default value ext4 is used. This parameter is available only when volumeMode is set to Filesystem.
spec.csi.volumeAttributes.dTreeParentName	Name of the parent file system when the volume resource type is dtree.	Conditionally mandatory	-	This parameter is mandatory when the managed object is a dtree resource and the parentname parameter is not configured in the storage backend. If dTreeParentName is configured only in the PV but parentname is not configured in the corresponding storage backend, set CSIDriverObject.attachRequired to true during CSI installation according to Table 5.
spec.capacity.storage	Volume size.	Yes	100Gi	Ensure that the size is the same as that of the corresponding resource on the storage. Kubernetes will not invoke CSI to check whether the value of this parameter is correct. Therefore, the PV can be successfully created even if its capacity is inconsistent with that of the corresponding resource on the storage.
spec.mountOptions.nfsvers	NFS mount option on the host. The following mount option is supported: nfsvers: protocol version for NFS mounting. The value can be 3, 4, 4.0, 4.1, or 4.2.	No	-	This parameter is optional after the -o parameter when the mount command is executed on the host. The value is in list format. If the NFS version is specified for mounting, NFS 3, 4.0, 4.1, and 4.2 protocols are supported (the protocol must be supported and enabled on storage devices). If nfsvers is set to 4, the latest protocol version NFS 4 may be used for mounting due to different OS configurations, for example, 4.2. If the 4.0 protocol is required, you are advised to set nfsvers to 4.0.
spec.mountOptions.acl	The DPC namespace supports the ACL function. The DPC client supports POSIX ACL, NFSv4 ACL, and NT ACL authentication.	No	-	The descriptions of acl, aclonlyposix, cnflush, and cflush are for reference only. For details about the parameters, see OceanStor Pacific Series Product Documentation and choose Configuration > Basic Service Configuration Guide for File > Configuring Basic Services (DPC Scenario) > Accessing a DPC Share on a Client > Step 2.
spec.mountOptions.aclonlyposix	The DPC namespace supports POSIX ACL, and the DPC client supports POSIX ACL authentication. The following protocols support POSIX ACL: DPC, NFSv3, and HDFS. If NFSv4 ACL or NT ACL is used, the DPC client cannot identify the ACL of this type. As a result, the ACL of this type does not take effect.	No	-	If aclonlyposix and acl are used together, only acl takes effect. That is, the namespace supports the ACL function.
spec.mountOptions.cnflush	Asynchronous disk flushing mode. That is, data is not flushed to disks immediately when files in the namespace are closed.	No	-	Asynchronous flushing mode: When a file is closed, data in the cache is not flushed to storage media in synchronous mode. Instead, data is written from the cache to the storage media in asynchronous flushing mode. After the write service is complete, data is flushed from the cache to disks periodically based on the flushing period. In a multi-client scenario, if concurrent operations are performed on the same file, the file size update is affected by the disk flushing period. That is, the file size is updated only after the disk flushing is complete. Generally, the update is completed within several seconds. Synchronous I/Os are not affected by the disk flushing period.
spec.mountOptions.cflush	Synchronous disk flushing mode. That is, data is flushed to disks immediately when files in the namespace are closed.	No	-	By default, the synchronous disk flushing mode is used.

Configuring a PVC

After a PV is created in static volume provisioning mode, you can create a PVC based on the PV for containers.

Prepare the PVC configuration file mypv.yaml. The following is an example. For details about other parameters, see Table 2 .

kind: PersistentVolumeClaim
apiVersion: v1
metadata:
  name: mypvc
spec:
  storageClassName: ""
  accessModes:
    - ReadWriteOnce
  volumeMode: Filesystem
  resources:
    requests:
      storage: 100Gi
  volumeName: mypv # Enter the name of the corresponding PV.

Run the following command to create a PVC based on the configured .yaml file.
```
kubectl create -f mypvc.yaml
```
After a period of time, run the following command to view the information about the created PVC.
```
kubectl get pvc
```
The following is an example of the command output. If the PVC status is Bound, the PVC is successfully created.
```
NAME        STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS   AGE
mypvc       Bound    pvc-840054d3-1d5b-4153-b73f-826f980abf9e   100Gi      RWO                           12s
```
- After the PVC is created, if the PVC is in the Pending state after a long time, see When a PVC Is Created, the PVC Is in the Pending State .
- You are advised to create or delete a maximum of 100 PVCs in a batch.

Table 2 PVC parameters for configuring a static PV

Parameter	Description	Mandatory	Default Value	Remarks
metadata.name	User-defined name of a PVC object.	Yes	-	Take Kubernetes v1.22.1 as an example. The value can contain digits, lowercase letters, hyphens (-), and periods (.), and must start and end with a letter or digit.
spec.accessModes	Access mode of the volume. RWO (ReadWriteOnce): A volume can be mounted to a node in read/write mode. This mode also allows multiple Pods running on the same node to access the volume. ROX (ReadOnlyMany): A volume can be mounted to multiple nodes in read-only mode. RWX (ReadWriteMany): A volume can be mounted to multiple nodes in read/write mode. RWOP (ReadWriteOncePod): A volume can only be mounted to a single Pod in read/write mode. Kubernetes 1.22 and later versions support this feature.	Yes	ReadWriteOnce	RWO/ROX/RWOP: supported by all types of volumes. RWOP is supported only by Kubernetes 1.22 and later versions. For versions earlier than Kubernetes 1.29, you need to enable this feature by following the instructions in Enabling the ReadWriteOncePod Feature Gate. The support for RWX is as follows: NAS storage: supported by all volumes SAN storage: supported only by volumes whose volumeMode is set to Block
spec.volumeMode	Volume mode.	No	Filesystem	This parameter is optional. The value can be Filesystem or Block. The default value is Filesystem. This parameter takes effect when a Pod is created. Filesystem indicates that a file system is created on a PVC to access the storage. Block indicates that a raw volume is used to access the storage.
spec.resources.requests.storage	Size of the volume to be created.	Yes	-	Size of the volume to be created. The format is ***Gi and the unit is GiB. The PVC capacity depends on storage specifications and host specifications. For example, OceanStor Dorado 6.1.2 or OceanStor Pacific series 8.1.0 is connected to CentOS 7. If ext4 file systems are used, see Table 2. If XFS file systems are used, see Table 3. If NFS or raw devices are used, the capacity must meet the specifications of the used Huawei storage device model and version. If the PVC capacity does not meet the specifications, a PVC or Pod may fail to be created due to the limitations of storage specifications or host file system specifications. When a PVC is created using a static PV and the PVC capacity is smaller than the capacity of the bound PV, the PVC capacity is set to the capacity of the bound PV. If the PVC capacity is greater than the capacity of the bound PV, the PVC cannot be created.
spec.volumeName	Name of the PV object.	Yes	-	This parameter is mandatory when a PVC is created statically.
spec.storageClassName	Name of the StorageClass object.	Yes	-	When a PVC is created, an empty character string is transferred. If this parameter is not set, the default StorageClass object name will be used.

Using a PVC

The use method is the same as that for dynamic volume provisioning in Using a PVC .

3 - Managing PVs

Manage Volume Provisioning allows administrators to use resources created on storage as PVs and supports features of dynamic volumes, such as capacity expansion, snapshot, and clone. This is a custom capability of Huawei CSI. This feature applies to the following scenarios:

In the reconstruction containerized applications, existing storage volumes need to be used.
The Kubernetes cluster is rebuilt.
Storage data is migrated in disaster recovery (DR) scenarios.

Manage Volume Provisioning allows existing storage resources to be managed by Kubernetes. You are not allowed to manage a storage resource for multiple times and concurrently delete or create a storage resource. When a storage resource is managed by multiple clusters, operations on the managed volume in a single cluster take effect only in the cluster and will not be synchronized to other clusters. Instead, you need to perform these operations on the managed volume in other clusters. For example, when you expand the capacity of a PVC in a cluster, the capacity of the corresponding PVC in other clusters will not be automatically expanded. In this case, you need to manually expand the capacity in other clusters by running the expansion commands in Expanding the Capacity of a PV .

Configuration Description

Perform the following steps to manage and use PVs:

Preparation
Configuring a PVC
Using a PVC

Preparation

You have registered the storage where the volume to be managed resides with CSI.
You have logged in to the storage device to obtain the name and capacity of the volume to be managed.
The StorageClass has been configured. For details, see Configuring a StorageClass (pay attention to the Whether the Volume Management Takes Effect field in the table).

Configuring a PVC

Prepare the PVC configuration file mypv.yaml. The following is an example. For details about other parameters, see Table 1 .

kind: PersistentVolumeClaim
apiVersion: v1
metadata:
  name: mypvc
  annotations:
    csi.huawei.com/manageVolumeName: "*"  # Enter the storage resource name.
    csi.huawei.com/manageBackendName: "*" # Enter the storage backend name.
  labels:
    provisioner: csi.huawei.com
spec:
  accessModes:
    - ReadWriteOnce
  volumeMode: Filesystem
  storageClassName: mysc
  resources:
    requests:
      storage: 100Gi

Run the following command to create a PVC using the configuration file.
```
kubectl create -f mypvc.yaml
```
After a period of time, run the following command to view the information about the created PVC.
```
kubectl get pvc mypvc
```
The following is an example of the command output. If the PVC status is Bound, the PVC has been created and can be used by a Pod.
```
NAME        STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS   AGE
mypvc       Bound    pvc-840054d3-1d5b-4153-b73f-826f980abf9e   100Gi      RWO            mysc           12s
```
- After the PVC is created, if the PVC is in the Pending state after a long time (for example, one minute), refer to When a PVC Is Created, the PVC Is in the Pending State .
- You are advised to create or delete a maximum of 100 PVCs in a batch.

Table 1 PVC parameters for managing PVs

Parameter	Description	Mandatory	Default Value	Remarks
metadata.annotations	PVC object annotations. Set the following parameters: Driver name/manageVolumeName: volume name on the storage. Driver name/manageBackendName: name of the backend to which the volume belongs.	Yes	csi.huawei.com/manageVolumeName: * csi.huawei.com/manageBackendName: *	For details about how to obtain Driver name, see Table 4. Driver name/manageVolumeName: name of an existing volume on the storage. Only English characters are supported. Driver name/manageBackendName: name of the storage backend in CSI. You can run the oceanctl get backend -n huawei-csi command to obtain the backend name.
metadata.labels	PVC object labels.	No	-	Format: provisioner: Driver name specified during installation Example: provisioner: csi.huawei.com This parameter takes effect when a PVC is created. It is used to listen to PVC resources and obtain information about metadata.annotations.
metadata.name	User-defined name of a PVC object.	Yes	-	Take Kubernetes v1.22.1 as an example. The value can contain digits, lowercase letters, hyphens (-), and periods (.), and must start and end with a letter or digit.
spec.volumeMode	Volume mode. This parameter is optional. When LUN volumes are used, the following types are supported: Filesystem: local file system. Block: raw device. NOTE: This parameter takes effect when a PV is mounted. The use method of this parameter must be the same as that of the managed volume. If a volume is used as a raw volume before being managed, volumeMode must be set to Block. If a volume is used in ext2, ext3, or ext4 mode before being managed, volumeMode must be set to Filesystem and fsType in the StorageClass must be set to ext2, ext3, or ext4. If a volume is used in XFS mode before being managed, volumeMode must be set to Filesystem and fsType in the StorageClass must be set to xfs.	No	Filesystem	This parameter takes effect when a PV is mounted. Filesystem indicates that a container accesses a PV using a local file system. The local file system type is specified by the fsType field in the specified StorageClass. Block indicates that a PV is accessed in raw volume mode.
spec.storageClassName	Name of the StorageClass object.	Yes	-	The configuration of the StorageClass must be the same as that of the managed volume.
spec.resources.requests.storage	Size of the volume to be created. The format is ***Gi and the unit is GiB.	Yes	-	The PVC capacity depends on storage specifications and host specifications. For example, OceanStor Dorado 6.1.2 or OceanStor Pacific series 8.1.0 is connected to CentOS 7. If ext4 file systems are used, see Table 2. If XFS file systems are used, see Table 3. If NFS or raw devices are used, the capacity must meet the specifications of the used Huawei storage device model and version. If the PVC capacity does not meet the specifications, a PVC or Pod may fail to be created due to the limitations of storage specifications or host file system specifications.
spec.accessModes	Access mode of the volume. RWO (ReadWriteOnce): A volume can be mounted to a node in read/write mode. This mode also allows multiple Pods running on the same node to access the volume. ROX (ReadOnlyMany): A volume can be mounted to multiple nodes in read-only mode. RWX (ReadWriteMany): A volume can be mounted to multiple nodes in read/write mode. RWOP (ReadWriteOncePod): A volume can only be mounted to a single Pod in read/write mode. Kubernetes 1.22 and later versions support this feature.	Yes	ReadWriteOnce	RWO/ROX/RWOP: supported by all types of volumes. RWOP is supported only by Kubernetes 1.22 and later versions. For versions earlier than Kubernetes 1.29, you need to enable this feature by following the instructions in Enabling the ReadWriteOncePod Feature Gate. The support for RWX is as follows: NAS storage: supported by all volumes SAN storage: supported only by volumes whose volumeMode is set to Block

Using a PVC

The procedure is the same as that for Using a PVC for dynamic volume provisioning.