Mount a CPFS for Lingjun file system on ACS

RDMA network

VPC network

Applies to CPU pods, standard GPUs (such as T4 or A10), and Lingjun GPU access across compute-storage affinity domains.

1. Create a file named cpfs-pv-pvc.yaml to define the PV and PVC:

   apiVersion: v1
   kind: PersistentVolume
   metadata:
     name: cpfs-test
     labels:
       alicloud-pvname: cpfs-test   # [PV label] Used by the PVC to bind to this PV
   spec:
     accessModes:
     - ReadWriteMany
     capacity:
       storage: 10Ti
     csi:
       driver: nasplugin.csi.alibabacloud.com
       volumeAttributes:
         mountProtocol: efc
         server: cpfs-***-vpc-***.cn-wulanchabu.cpfs.aliyuncs.com  # Replace with your VPC mount point domain name
         path: /
         autoSwitch: "true"    # Optional. Enables automatic VSC/VPC switching (recommended for Lingjun cross-AZ scenarios)
       volumeHandle: bmcpfs-*****   # Replace with your CPFS for Lingjun file system ID
   ---
   apiVersion: v1
   kind: PersistentVolumeClaim
   metadata:
     name: cpfs-test      # [PVC name] Referenced by the Deployment
   spec:
     accessModes:
     - ReadWriteMany
     selector:
       matchLabels:
         alicloud-pvname: cpfs-test    # [PVC selector] Must match the PV label
     resources:
       requests:
         storage: 10Ti

   Parameter description:

   • PV parameters:

     Parameter	Description
     labels	Sets a label so that the PVC can use selector to match and bind the PV.
     accessModes	The access mode of the PV.
     capacity.storage	The declared capacity of the volume.
     csi.driver	The driver type. Set the value to bmcpfsplugin.csi.alibabacloud.com.
     csi.volumeAttributes	The attributes of the CPFS volume. filesystemId: the ID of the CPFS for Lingjun file system. path: the default value / mounts the root directory of the CPFS file system. Optionally, specify a subdirectory, such as /dir. If the subdirectory does not exist, it is automatically created when the volume is mounted.
     csi.volumeHandle	The ID of the CPFS for Lingjun file system.

   • PVC parameters:

     Parameter	Description
     accessModes	The access mode that the PVC requests from the PV.
     selector	Matches and binds the PV based on the label on the PV.
     resources.requests.storage	The storage capacity allocated to the pod. The value must not exceed the capacity of the PV.

   • autoSwitch: An optional parameter. When set to "true", it enables automatic VSC/VPC switching. When Lingjun GPUs access across availability zones, the system automatically switches from RDMA to the VPC channel.

   Note

   The throughput of the VPC channel is limited by network bandwidth, up to approximately 3.2 GB/s (25 Gbps), which is lower than a direct RDMA connection. For performance-sensitive scenarios, we recommend prioritizing a CPFS instance in the same compute-storage affinity domain.

2. Run the create command:

   kubectl create -f cpfs-pv-pvc.yaml

3. Confirm that the PVC is bound to the PV.

   kubectl get pvc cpfs-test

   Example output (a STATUS of Bound indicates success):

   NAME        STATUS   VOLUME           CAPACITY   ACCESS MODES   STORAGECLASS    VOLUMEATTRIBUTESCLASS   AGE
   cpfs-test   Bound    cpfs-test        10Ti       RWX            <unset>         <unset>                 10s

kubectl

1. Create a file named cpfs-test.yaml with the following content.

   GPU application

   The following YAML example creates a Deployment with two Pods. Both Pods use the alibabacloud.com/compute-class: gpu label to request GPU-based computing power and a PersistentVolumeClaim (PVC) named cpfs-test to request storage resources. The mount path for both Pods is /data.

   Note

   For more information about specific GPU models, see Specify GPU models and driver versions for ACS GPU Pods.

   apiVersion: apps/v1
   kind: Deployment
   metadata:
     name: cpfs-test
     labels:
       app: cpfs-test
   spec:
     replicas: 2
     selector:
       matchLabels:
         app: cpfs-test
     template:
       metadata:
         labels:
           app: cpfs-test
           # Specify the compute type as GPU.
           alibabacloud.com/compute-class: gpu
           # Specify the GPU model, for example, T4.
           alibabacloud.com/gpu-model-series: T4
           alibabacloud.com/compute-qos: default
       spec:
         containers:
         - name: nginx
           image: registry.cn-hangzhou.aliyuncs.com/acs-sample/nginx:latest
           ports:
           - containerPort: 80
           volumeMounts:
             - name: pvc-cpfs
               mountPath: /data
         volumes:
           - name: pvc-cpfs
             persistentVolumeClaim:
               claimName: cpfs-test

   CPU application

   The following YAML example creates a Deployment with 2 pods, which request storage resources by using a PVC named cpfs-test and are both mounted to the /data path.

   apiVersion: apps/v1
   kind: Deployment
   metadata:
     name: cpfs-test
     labels:
       app: cpfs-test
   spec:
     replicas: 2
     selector:
       matchLabels:
         app: cpfs-test
     template:
       metadata:
         labels:
           app: cpfs-test
       spec:
         containers:
         - name: nginx
           image: registry.cn-hangzhou.aliyuncs.com/acs-sample/nginx:latest
           ports:
           - containerPort: 80
           volumeMounts:
             - name: pvc-cpfs
               mountPath: /data
         volumes:
           - name: pvc-cpfs
             persistentVolumeClaim:
               claimName: cpfs-test

2. Create the Deployment and mount the CPFS volume.

   kubectl create -f cpfs-test.yaml

3. Check the status of the Pods in the Deployment.

   kubectl get pod | grep cpfs-test

   The expected output shows that two Pods are in the 'Running' state.

   cpfs-test-****-***a   1/1     Running   0          45s
   cpfs-test-****-***b   1/1     Running   0          45s

4. Check the mount path.

   This command lists the contents of the mounted directory (/data), which contains data from the CPFS for LINGJUN file system. The directory is empty by default.

   kubectl exec cpfs-test-****-***a -- ls /data

Console

1. In the left-side navigation pane of the cluster management page, choose Workloads > Deployments.

2. On the Deployments page, click Create from Image.

3. Configure the parameters for the Deployment and click Create.

   The key parameters are described below. You can keep the default values for other parameters. For more information, see Create a stateless application by using a Deployment.

   GPU application

   Section	Parameter	Description	Example
   Basic Information	Application Name	Enter a unique name for the Deployment. The name must conform to the format requirements shown on the page.	cpfs-test
   	Replicas:	The number of Pod replicas for the Deployment.	2
   	Instance Type	Specifies the hardware resources for the Pods, such as CPU or a specific GPU model.	GPU, T4
   Container	Image Name	The container image used to deploy the application.	registry.cn-hangzhou.aliyuncs.com/acs-sample/nginx:latest
   	Required Resources	The amount of GPU, vCPU, and memory resources to allocate to each Pod.	GPU: 1 CPU: 2 vCPU Memory: 2 GiB
   	Volume	Click Add PVC and then configure the parameters. Mount Source: Select the existing PersistentVolumeClaim. Container Path: Enter the path within the container where the CPFS file system will be mounted.	Mount source: pvc-cpfs Container path: /data

   CPU application

   Section	Parameter	Description	Example
   Basic Information	Application Name	Enter a unique name for the Deployment. The name must conform to the format requirements shown on the page.	cpfs-test
   	Replicas:	The number of Pod replicas for the Deployment.	2
   	Instance Type	The compute type of the Pod.	CPU, General Purpose
   Container	Image Name	The container image used to deploy the application.	registry.cn-hangzhou.aliyuncs.com/acs-sample/nginx:latest
   	Required Resources	The amount of vCPU and memory resources to allocate to each Pod.	CPU: 0.25 vCPU Memory: 0.5 GiB
   	Volume	Click Add PVC and then configure the parameters. Mount Source: Select the existing PersistentVolumeClaim. Container Path: Enter the path within the container where the CPFS file system will be mounted.	Mount source: pvc-cpfs Container path: /data

4. Check the application's status.

   1. On the Deployments page, click the application name.

   2. On the Pods tab, check that the Pods are in the 'Running' state.