配置私有镜像仓库

Tags:

1. 镜像仓库的基本操作

1.1. 登录镜像仓库

docker login -u <username> -p <password> <registry-addr>

1.2. 拉取镜像

docker pull https://registry.xxx.com/dev/nginx:latest

1.3. 推送镜像

docker push https://registry.xxx.com/dev/nginx:latest

1.4. 重命名镜像

docker tag <old-image> <new-image>

2. docker.xxx.com镜像仓库

使用docker.xxx.com镜像仓库。

2.1. 所有节点配置insecure-registries

#cat /etc/docker/daemon.json
{
  "data-root": "/data/docker",
  "debug": false,
  "insecure-registries": [
	...
    "docker.xxx.com:8080"
  ],
  ...
}

2.2. 所有节点配置/var/lib/kubelet/config.json

具体参考：configuring-nodes-to-authenticate-to-a-private-registry

  1. 在某个节点登录docker.xxx.com:8080镜像仓库，会更新 $HOME/.docker/config.json
  2. 检查$HOME/.docker/config.json是否有该镜像仓库的auth信息。
#cat ~/.docker/config.json
{
	"auths": {
		"docker.xxx.com:8080": {
			"auth": "<此处为凭证信息>"
		}
	},
	"HttpHeaders": {
		"User-Agent": "Docker-Client/18.09.9 (linux)"
	}
}
  1. 将$HOME/.docker/config.json拷贝到所有的Node节点上的/var/lib/kubelet/config.json。
# 获取所有节点的IP
nodes=$(kubectl get nodes -o jsonpath='{range .items[*].status.addresses[?(@.type=="ExternalIP")]}{.address} {end}')
# 拷贝到所有节点
for n in $nodes; do scp ~/.docker/config.json root@$n:/var/lib/kubelet/config.json; done

2.3. 创建docker.xxx.com镜像的pod

指定镜像为：docker.xxx.com:8080/public/2048:latest

完整pod.yaml

apiVersion: apps/v1beta2
kind: Deployment
metadata:
  annotations:
    deployment.kubernetes.io/revision: "1"
  generation: 1
  labels:
    k8s-app: dockeroa-hub
    qcloud-app: dockeroa-hub
  name: dockeroa-hub
  namespace: test
spec:
  progressDeadlineSeconds: 600
  replicas: 3
  revisionHistoryLimit: 10
  selector:
    matchLabels:
      k8s-app: dockeroa-hub
      qcloud-app: dockeroa-hub
  strategy:
    rollingUpdate:
      maxSurge: 25%
      maxUnavailable: 25%
    type: RollingUpdate
  template:
    metadata:
      labels:
        k8s-app: dockeroa-hub
        qcloud-app: dockeroa-hub
    spec:
      containers:
      - image: docker.xxx.com:8080/public/2048:latest
        imagePullPolicy: Always
        name: game
        resources:
          limits:
            cpu: 500m
            memory: 1Gi
          requests:
            cpu: 250m
            memory: 256Mi
        terminationMessagePath: /dev/termination-log
        terminationMessagePolicy: File
      dnsPolicy: ClusterFirst
      restartPolicy: Always
      nodeName: 192.168.1.1
      schedulerName: default-scheduler
      securityContext: {}
      terminationGracePeriodSeconds: 30

查看pod状态

#kgpoowide -n game
NAME                                     READY   STATUS    RESTARTS   AGE     IP             NODE            NOMINATED NODE   READINESS GATES
docker-oa-757bbbddb5-h6j7m               1/1     Running   0          14m     192.168.2.51   192.168.1.1    <none>           <none>
docker-oa-757bbbddb5-jp5dw               1/1     Running   0          14m     192.168.1.32   192.168.1.2    <none>           <none>
docker-oa-757bbbddb5-nlw9f               1/1     Running   0          14m     192.168.0.43   192.168.1.3   <none>           <none>

参考：


最后修改 December 25, 2022: deploy by blog source (a162b04)