Question

我试图将gcePersistentDisk用作ReadOnlyMany，以便多个节点上的Pod可以读取该磁盘上的数据。与文档here相同。

要创建gce永久磁盘并在以后对其进行格式化，请遵循文档here中的说明。遵循此文档之后，我进入了一个节点并格式化了磁盘。参见下面的完整错误以及其他Yaml文件。

kubectl描述pods -l podName

Name:               punk-fly-nodejs-deployment-5dbbd7b8b5-5cbfs
Namespace:          default
Priority:           0
PriorityClassName:  <none>
Node:               gke-mycluster-default-pool-b1c1d316-d016/10.160.0.12
Start Time:         Thu, 25 Apr 2019 23:55:38 +0530
Labels:             app.kubernetes.io/instance=punk-fly
                    app.kubernetes.io/name=nodejs
                    pod-template-hash=1866836461
Annotations:        kubernetes.io/limit-ranger=LimitRanger plugin set: cpu request for container nodejs
Status:             Pending
IP:
Controlled By:      ReplicaSet/punk-fly-nodejs-deployment-5dbbd7b8b5
Containers:
  nodejs:
    Container ID:
    Image:          rajesh12/smartserver:server
    Image ID:
    Port:           3002/TCP
    Host Port:      0/TCP
    State:          Waiting
      Reason:       ContainerCreating
    Ready:          False

    Restart Count:  0
    Requests:
      cpu:  100m
    Environment:
      MYSQL_HOST:           mysqlservice
      MYSQL_DATABASE:       app
      MYSQL_ROOT_PASSWORD:  password
    Mounts:
      /usr/src/ from helm-vol (rw)
      /var/run/secrets/kubernetes.io/serviceaccount from default-token-jpkzg (ro)
Conditions:
  Type              Status
  Initialized       True
  Ready             False
  ContainersReady   False
  PodScheduled      True
Volumes:
  helm-vol:
    Type:       PersistentVolumeClaim (a reference to a PersistentVolumeClaim in the same namespace)
    ClaimName:  my-readonly-pvc
    ReadOnly:   true
  default-token-jpkzg:
    Type:        Secret (a volume populated by a Secret)
    SecretName:  default-token-jpkzg
    Optional:    false
QoS Class:       Burstable
Node-Selectors:  <none>
Tolerations:     node.kubernetes.io/not-ready:NoExecute for 300s
                 node.kubernetes.io/unreachable:NoExecute for 300s
Events:
  Type     Reason                  Age               From                                               Message
  ----     ------                  ----              ----                                               -------
  Normal   Scheduled               2m                default-scheduler                                  Successfully assigned default/punk-fly-nodejs-deployment-5dbbd7b8b5-5cbfs to gke-mycluster-default-pool-b1c1d316-d016
  Normal   SuccessfulAttachVolume  1m                attachdetach-controller                            AttachVolume.Attach succeeded for volume "pvc-9c796180-677e-11e9-ad35-42010aa0000f"
  Warning  FailedMount             10s (x8 over 1m)  kubelet, gke-mycluster-default-pool-b1c1d316-d016  MountVolume.MountDevice failed for volume "pvc-9c796180-677e-11e9-ad35-42010aa0000f" : failed to mount unformatted volume as read only
  Warning  FailedMount             0s                kubelet, gke-mycluster-default-pool-b1c1d316-d016  Unable to mount volumes for pod "punk-fly-nodejs-deployment-5dbbd7b8b5-5cbfs_default(86293044-6787-11e9-ad35-42010aa0000f)": timeout expired waiting for volumes to attach or mount for pod "default"/"punk-fly-nodejs-deployment-5dbbd7b8b5-5cbfs". list of unmounted volumes=[helm-vol]. list of unattached volumes=[helm-vol default-token-jpkzg]

readonly_pv.yaml

apiVersion: v1
kind: PersistentVolume
metadata:
  name: my-readonly-pv
spec:
  storageClassName: ""
  capacity:
    storage: 1G
  accessModes:
    - ReadOnlyMany
  gcePersistentDisk:
    pdName: mydisk0
    fsType: ext4
---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: my-readonly-pvc
spec:
  accessModes:
    - ReadOnlyMany
  resources:
    requests:
      storage: 1G

deployment.yaml

  volumes:
    - name: helm-vol
      persistentVolumeClaim:
        claimName: my-readonly-pvc
        readOnly: true
  containers:
    - name: {{ .Values.app.backendName }}
      image: "{{ .Values.image.repository }}:{{ .Values.image.tagServer }}"
      imagePullPolicy: {{ .Values.image.pullPolicy }}
      env:
      - name: MYSQL_HOST
        value: mysqlservice
      - name: MYSQL_DATABASE
        value: app
      - name: MYSQL_ROOT_PASSWORD
        value: password
      ports:
        - name: http-backend
          containerPort: 3002
      volumeMounts:
        - name: helm-vol
          mountPath: /usr/src/

Answer 1

听起来您的PVC正在动态配置未使用default StorageClass格式化的新卷

可能是因为您的Pod的创建可用性与配置PV的可用性不同。对于gce卷，具有多个Pod读取器的问题是，Pods必须始终位于相同的可用区中。

一些选项：

只需在节点所在的同一可用区上创建PV并设置其格式。
定义PV时，可以指定Node Affinity以确保始终将其分配给特定节点。

定义用于指定文件系统的StorageClass

kind: StorageClass
apiVersion: storage.k8s.io/v1
metadata:
  name: mysc
provisioner: kubernetes.io/aws-ebs
parameters:
  type: gp2
  fsType: ext4

然后在您的PVC中使用它：

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: my-pvc
spec:
  accessModes:
    - ReadOnlyMany
  resources:
    requests:
      storage: 1G
  storageClassName: mysc

该卷将自动配置和格式化。

Answer 2

我遇到了相同的错误，并设法使用相关文章中有关使用现有磁盘的几行内容进行了修复 https://cloud.google.com/kubernetes-engine/docs/how-to/persistent-volumes/preexisting-pd

您需要将volumeName和apiVersion: v1 kind: PersistentVolumeClaim metadata: name: pv-claim-demo spec: # It's necessary to specify "" as the storageClassName # so that the default storage class won't be used, see # https://kubernetes.io/docs/concepts/storage/persistent-volumes/#class-1 storageClassName: "" volumeName: pv-demo accessModes: - ReadOnlyMany resources: requests: storage: 500G添加到您的永久批量声明中：

class subdivision(callbacks.Callback):
    def __init__(self,subdiv):
        super(subdivision,self).__init__()
        self.subdiv = subdiv
        self.weights = None
        self.index = 1
        self.gradient = None

    def on_train_begin(self,logs=None):
        self.weights = self.model.get_weights()

    def on_batch_begin(self,batch,logs=None):
        self.model.set_weights(self.weights)

    def on_batch_end(self,batch,logs=None):
        if (self.index % self.subdiv==0):
            self.model.set_weights(self.weights+self.gradient)
            self.weights = self.model.get_weights()
            self.gradient=None
        else:
            if self.gradient==None:
                self.gradient = k.gradients(self.model.output,self.model.weights)
            else:
                self.gradient += k.gradients(self.model.output,self.model.weights)
        self.index+=1
    def on_epoch_end(self,epoch,logs=None):
        self.index = 1

Kubernetes：无法将未格式化的卷挂载为只读

2 个答案: