1. 发现问题

• PV

apiVersion: v1
kind: PersistentVolume
metadata:
  name: pv-ghost
  namespace: ghost
spec:
  capacity:
    storage: 10Gi
  accessModes:
  - ReadWriteOnce
  persistentVolumeReclaimPolicy: Retain
  storageClassName: local-storage
  local:
    path: /mnt/mydrive/ghost-data/
  nodeAffinity:
    required:
      nodeSelectorTerms:
      - matchExpressions:
        - key: kubernetes.io/hostname
          operator: In
          values:
          - rpi-mon-k8-worker

• PVC

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: pvc-ghost
  namespace: ghost
  labels:
    pv: pv-ghost
spec:
  accessModes:
    - ReadWriteOnce
  resources:
    requests:
      storage: 10Gi
  storageClassName: local-storage
  selector:
    matchLabels:
      name: pv-ghost

• StorageClass

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

• Deployment

apiVersion: apps/v1
kind: Deployment
metadata:
  name: deployment-ghost
  namespace: ghost
  labels:
    env: prod
    app: ghost-app
spec:
  template:
    metadata:
      name: ghost-app-pod
      labels:
        app:  ghost-app
        env:  production
    spec:
      containers:
        - name: ghost
          image: arm32v7/ghost
          imagePullPolicy: IfNotPresent
          volumeMounts:
           - mountPath: /var/lib/ghost/content
             name: ghost-blog-data
          securityContext:
            privileged: True
 volumes:
      - name: ghost-blog-data
        persistentVolumeClaim:
          claimName: pvc-ghost
      nodeSelector:
        beta.kubernetes.io/arch: arm
  replicas: 2
  selector:
    matchLabels:
      app: ghost-app

# kubectl get nodes -o wide

NAME                  STATUS   ROLES    AGE   VERSION   INTERNAL-IP      EXTERNAL-IP   OS-IMAGE                         KERNEL-VERSION      CONTAINER-RUNTIME
rpi-k8-workernode-2   Ready    <none>   92d   v1.15.0   192.168.100.50   <none>        Raspbian GNU/Linux 9 (stretch)   4.19.42-v7+         docker://18.9.0
rpi-mon-k8-worker     Ready    <none>   91d   v1.15.0   192.168.100.22   <none>        Raspbian GNU/Linux 9 (stretch)   4.19.42-v7+         docker://18.9.0
udubuntu              Ready    master   92d   v1.15.1   192.168.100.24   <none>        Ubuntu 18.04.3 LTS               4.15.0-55-generic   docker://19.3.4


# kubectl get nodes --show-labels

NAME                  STATUS   ROLES    AGE   VERSION   LABELS
rpi-k8-workernode-2   Ready    <none>   93d   v1.15.0   beta.kubernetes.io/arch=arm,beta.kubernetes.io/os=linux,kubernetes.io/arch=arm,kubernetes.io/hostname=rpi-k8-workernode-2,kubernetes.io/os=linux
rpi-mon-k8-worker     Ready    <none>   93d   v1.15.0   beta.kubernetes.io/arch=arm,beta.kubernetes.io/os=linux,kubernetes.io/arch=arm,kubernetes.io/hostname=rpi-mon-k8-worker,kubernetes.io/os=linux
udubuntu              Ready    master   93d   v1.15.1   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=udubuntu,kubernetes.io/os=linux,node-role.kubernetes.io/master=


# kubectl describe pvc pvc-ghost -n ghost

Name:          pvc-ghost
Namespace:     ghost
StorageClass:  manual
Status:        Pending
Volume:
Labels:        pv=pv-ghost
Annotations:   kubectl.kubernetes.io/last-applied-configuration:
                 {"apiVersion":"v1","kind":"PersistentVolumeClaim","metadata":{"annotations":{},"labels":{"pv":"pv-ghost"},"name":"pvc-ghost","namespace":"...
Finalizers:    [kubernetes.io/pvc-protection]
Capacity:
Access Modes:
VolumeMode:    Filesystem
Mounted By:    <none>
Events:
  Type    Reason                Age               From                         Message
  ----    ------                ----              ----                         -------
  Normal  WaitForFirstConsumer  6s (x2 over 21s)  persistentvolume-controller  waiting for first consumer to be created before binding

2. 分析问题

在deployment-ghost.yaml中，有人设置了一个nodeSelector，但是有一个worker节点没有匹配上这个nodeSelector。如果从deployment-ghost.yaml中删除这个nodeSelector，这样，pod就会部署到创建PV的节点上。据我所知，K8s无法把pod部署到PV与PVC的selector-matchLabels不同的工作节点。

在PVC中，我们可以指定一个label-selector去绑定包含目标label的PV。只有那些label匹配label-selector的PV，才能被绑定到PVC。PVC的label-selector包含以下两个字段：

matchLabels，PV必须含有一个与PVC的matchLabel相同的label。

matchExpressions，由指定key、values和相关算子组成的列表。有效的算子包括In, NotIn, Exists, and DoesNotExist。

3. 解决问题

从deployment-ghost.yaml中删除这个nodeSelector。

apiVersion: apps/v1
kind: Deployment
metadata:
  name: deployment-ghost
  namespace: ghost
  labels:
    env: prod
    app: ghost-app
spec:
  template:
    metadata:
      name: ghost-app-pod
      labels:
        app:  ghost-app
        env:  production
    spec:
      containers:
        - name: ghost
          image: arm32v7/ghost
          imagePullPolicy: IfNotPresent
          volumeMounts:
           - mountPath: /var/lib/ghost/content
             name: ghost-blog-data
          securityContext:
            privileged: True
 volumes:
      - name: ghost-blog-data
        persistentVolumeClaim:
          claimName: pvc-ghost
  replicas: 2
  selector:
    matchLabels:
      app: ghost-app

在PV的yaml文件中，我们需要指定一个label name: pv-ghost，这样能使对应的PVC匹配到。

apiVersion: v1
kind: PersistentVolume
metadata:
  name: pv-ghost
  labels:
    name: pv-ghost
spec:
  capacity:
    storage: 10Gi
  accessModes:
  - ReadWriteOnce
  persistentVolumeReclaimPolicy: Retain
  storageClassName: local-storage
  local:
    path: /mnt/mydrive/ghost-data/
  nodeAffinity:
    required:
      nodeSelectorTerms:
      - matchExpressions:
        - key: kubernetes.io/hostname
          operator: In
          values:
          - rpi-mon-k8-worker

在PV的yaml文件中，我们没有必要指定metadata.namespace字段，因为PV绑定是独享的，可以不用namespace。然而，在PVC的yaml中，我们可以使用用namespace。

# kubectl get pv

# kubectl get pvc -n <namespace>


# kubectl describe pv <pv_name> 

# kubectl describe pv <pv_name> -n <namespace>