Kubernetes 硬盘持久化之 PV 和 PVC

Persistent Volume 定义

持久卷（PersistentVolume，PV）是集群中的一块存储，可以由管理员事先制备，或者使用存储类（Storage Class）来动态制备。持久卷是集群资源，就像节点也是集群资源一样。PV 持久卷和普通的 Volume 一样，拥有独立于任何使用 PV 的 Pod 的生命周期。

Persistent Volume 访问权限

权限	缩写	描述
ReadWriteOnce	RWO	卷可以被一个节点以读写方式挂载。 ReadWriteOnce 访问模式仍然可以在同一节点上运行的多个 Pod 访问该卷。对于单个 Pod 的访问，请参考 ReadWriteOncePod 访问模式。
ReadOnlyMany	ROX	卷可以被多个节点以只读方式挂载。
ReadWriteMany	RWX	卷可以被多个节点以读写方式挂载。

Persistent Volume 回收机制

机制
Retain	回收策略 `Retain` 使得用户可以手动回收资源。当 PersistentVolumeClaim 对象被删除时，PersistentVolume 卷仍然存在，对应的数据卷被视为"已释放（released）"。由于卷上仍然存在这前一申领人的数据，该卷还不能用于其他申领。
Delete	对于支持 `Delete` 回收策略的卷插件，删除动作会将 PersistentVolume 对象从 Kubernetes 中移除，同时也会从外部基础设施中移除所关联的存储资产。
Recycle（废弃）	如果下层的卷插件支持，回收策略 `Recycle` 会在卷上执行一些基本的擦除（`rm -rf /thevolume/*`）操作，之后允许该卷用于新的 PVC 申领。

Persistent Volume Claim 定义

持久卷申领（PersistentVolumeClaim，PVC） 表达的是用户对存储的请求。概念上与 Pod 类似。 Pod 会耗用节点资源，而 PVC 申领会耗用 PV 资源。Pod 可以请求特定数量的资源（CPU 和内存）。同样 PVC 申领也可以请求特定的大小和访问模式。

PV 和 PVC 联合使用

申明 3 个 PV

yaml 复制代码

apiVersion: v1
kind: PersistentVolume
metadata:
  name: pv1
  labels:
    app: pv1
spec:
  nfs:
    server: 192.168.31.60
    path: /nfs-test/volume1
  accessModes: ["ReadWriteOnce"]
  persistentVolumeReclaimPolicy: Retain
  capacity:
    storage: 1Gi
---
apiVersion: v1
kind: PersistentVolume
metadata:
  name: pv2
  labels:
    app: pv2
spec:
  nfs:
    server: 192.168.31.60
    path: /nfs-test/volume2
  accessModes: ["ReadOnlyMany"]
  persistentVolumeReclaimPolicy: Delete
  capacity:
    storage: 1Gi
---
apiVersion: v1
kind: PersistentVolume
metadata:
  name: pv3
  labels:
    app: pv3
spec:
  nfs:
    server: 192.168.31.60
    path: /nfs-test/volume3
  accessModes: ["ReadWriteMany"]
  persistentVolumeReclaimPolicy: Delete
  capacity:
    storage: 1Gi

复制代码

root@k8s-master1:~# kubectl get pv
NAME   CAPACITY   ACCESS MODES   RECLAIM POLICY   STATUS      CLAIM   STORAGECLASS   VOLUMEATTRIBUTESCLASS   REASON   AGE
pv1    1Gi        RWO            Retain           Available                          <unset>                          8s
pv2    1Gi        ROX            Delete           Available                          <unset>                          8s
pv3    1Gi        RWX            Delete           Available                          <unset>                          8s

声明 3 个 PVC

yaml 复制代码

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: pvc1
spec:
  accessModes: ["ReadWriteOnce"]
  selector:
    matchLabels:
      app: pv1
  resources:
    requests:
      storage: 100Mi
---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: pvc2
spec:
  accessModes: ["ReadOnlyMany"]
  selector:
    matchLabels:
      app: pv2
  resources:
    requests:
      storage: 100Mi
---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: pvc3
spec:
  accessModes: ["ReadWriteMany"]
  selector:
    matchLabels:
      app: pv3
  resources:
    requests:
      storage: 100Mi

复制代码

root@k8s-master1:~# kubectl get pvc
NAME   STATUS   VOLUME   CAPACITY   ACCESS MODES   STORAGECLASS   VOLUMEATTRIBUTESCLASS   AGE
pvc1   Bound    pv1      1Gi        RWO                           <unset>                 2m35s
pvc2   Bound    pv2      1Gi        ROX                           <unset>                 2m35s
pvc3   Bound    pv3      1Gi        RWX                           <unset>                 2m35s

申明 3 个 Deployment 进行测试

复制代码

apiVersion: apps/v1
kind: Deployment
metadata:
  name: pv1-deploy
spec:
  replicas: 2
  selector:
    matchLabels:
      app: pv1-pod
  template:
    metadata:
      labels:
        app: pv1-pod
    spec:
      containers:
      - name: pv1-test
        image: busybox
        imagePullPolicy: IfNotPresent
        command: ["/bin/sh"]
        args: ["-c","sleep 3600"]
        volumeMounts:
        - name: pv1-test
          mountPath: /pv1-test
      volumes:
      - persistentVolumeClaim:
          claimName: pvc1
        name: pv1-test
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: pv2-deploy
spec:
  replicas: 2
  selector:
    matchLabels:
      app: pv2-pod
  template:
    metadata:
      labels:
        app: pv2-pod
    spec:
      containers:
        - name: pv2-test
          image: busybox
          imagePullPolicy: IfNotPresent
          command: ["/bin/sh"]
          args: ["-c","sleep 3600"]
          volumeMounts:
            - name: pv2-test
              mountPath: /pv2-test
      volumes:
        - persistentVolumeClaim:
            claimName: pvc2
          name: pv2-test
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: pv3-deploy
spec:
  replicas: 2
  selector:
    matchLabels:
      app: pv3-pod
  template:
    metadata:
      labels:
        app: pv3-pod
    spec:
      containers:
        - name: pv3-test
          image: busybox
          imagePullPolicy: IfNotPresent
          command: ["/bin/sh"]
          args: ["-c","sleep 3600"]
          volumeMounts:
            - name: pv3-test
              mountPath: /pv3-test
      volumes:
        - persistentVolumeClaim:
            claimName: pvc3
          name: pv3-test

NAME                          READY   STATUS    RESTARTS   AGE   IP              NODE          NOMINATED NODE   READINESS GATES
pv1-deploy-9b857996d-9mlwq    1/1     Running   0          5m    10.244.194.97   k8s-worker1   <none>           <none>
pv1-deploy-9b857996d-gs95v    1/1     Running   0          5m    10.244.126.40   k8s-worker2   <none>           <none>
pv2-deploy-6f76d7c88b-bgr8k   1/1     Running   0          5m    10.244.126.41   k8s-worker2   <none>           <none>
pv2-deploy-6f76d7c88b-grq7q   1/1     Running   0          5m    10.244.194.95   k8s-worker1   <none>           <none>
pv3-deploy-8696ff555b-hkwgm   1/1     Running   0          5m    10.244.194.96   k8s-worker1   <none>           <none>
pv3-deploy-8696ff555b-z72pp   1/1     Running   0          5m    10.244.126.42   k8s-worker2   <none>           <none>

测试过程中发现访问模式没有起到任何限制作用，可能 Kubernetes 对特定配置下 NFS 没有很好的支持。