k8s（kubernetes）拉取本地镜像部署节点

本人在虚拟机上部署了1master、2node的集群，想要使用本地打包的镜像hyperledger/fabric-orderer：1.4.3，一开始不清楚原因，只在master上安装了：

docker load -i fabric-orderer.tar

随后查看：

[root@k8s-master my-test]# docker images
REPOSITORY                                               TAG                 IMAGE ID            CREATED             SIZE
hyperledger/fabric-orderer                               1.4.3               b666a6ebbe09        17 months ago       173 MB

然后创建测试fabric-orderer.yaml文件：

---
apiVersion: extensions/v1beta1
kind: Deployment
metadata:
  name: my-test
spec:
  replicas: 2
  template:
    metadata:
      labels:
        k8s-app: fabric-orderer
    spec:
      containers:
        - name:  fabric-orderer-select-node1
          image: hyperledger/fabric-orderer:1.4.3
          imagePullPolicy: Never
          ports:
          - containerPort: 80

随后创建该deployment：

[root@k8s-master my-test]# kubectl create -f fabric-orderer.yaml 
deployment "my-test" created

但是查看dashboard发现一片飘红，于是用命令查看：

[root@k8s-master my-test]# kubectl get pods --all-namespaces
NAMESPACE     NAME                                   READY     STATUS              RESTARTS   AGE
default       my-test-66cf46b4d-cn7qr                0/1       ErrImageNeverPull   0          19s
default       my-test-66cf46b4d-xgkqd                0/1       ErrImageNeverPull   0          19s

再具体一点：

[root@k8s-master my-test]# kubectl describe pod my-test-66cf46b4d-cn7qr
Name:           my-test-66cf46b4d-cn7qr
Namespace:      default
Node:           k8s-node-2/192.168.203.203
Start Time:     Tue, 26 Jan 2021 15:42:28 +0800
Labels:         k8s-app=fabric-orderer
                pod-template-hash=227902608
Annotations:    <none>
Status:         Pending
IP:             
Controlled By:  ReplicaSet/my-test-66cf46b4d
Containers:
  fabric-orderer-demo:
    Container ID:   
    Image:          fabric-orderer
    Image ID:       
    Port:           <none>
    State:          Waiting
      Reason:       ErrImageNeverPull
    Ready:          False
    Restart Count:  0
    Environment:    <none>
    Mounts:
      /var/run/secrets/kubernetes.io/serviceaccount from default-token-bzrrv (ro)
Conditions:
  Type           Status
  Initialized    True 
  Ready          False 
  PodScheduled   True 
Volumes:
  default-token-bzrrv:
    Type:        Secret (a volume populated by a Secret)
    SecretName:  default-token-bzrrv
    Optional:    false
QoS Class:       BestEffort
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              52s                default-scheduler    Successfully assigned my-test-66cf46b4d-cn7qr to k8s-node-2
  Normal   SuccessfulMountVolume  51s                kubelet, k8s-node-2  MountVolume.SetUp succeeded for volume "default-token-bzrrv"
  Warning  ErrImageNeverPull      32s (x8 over 49s)  kubelet, k8s-node-2  Container image "fabric-orderer" is not present with pull policy of Never
  Warning  Failed                 32s (x8 over 49s)  kubelet, k8s-node-2  Error: ErrImageNeverPull
  Normal   SandboxChanged         31s (x8 over 49s)  kubelet, k8s-node-2  Pod sandbox changed, it will be killed and re-created.

可以从最下面的Events看到两个warning，尤其是第一个，发现我的fabric-orderer没有部署上，理所应当的拉取不了，经过一顿查询：参考，每一个要使用该镜像的节点上都要事先docker load上该镜像，于是在k8s-node-1和k8s-node-2上：

docker load -i fabric-orderer.tar

随后再次：

[root@k8s-master my-test]# kubectl create -f fabric-orderer.yaml 
deployment "my-test" created

 这次成功了，当然这只是临时的用法，因为这样意味着你要把所有用到的镜像都打包好，部署到每个节点上。

不怕麻烦而且条件允许的话，还是创建一个本地镜像仓库为好（一劳永逸）。