KubeFlow 1.2.0镜像缓存（持续更新）

KubeFlow 1.2.0镜像比较多，存在多个docker仓库中，其中部分访问不可达。这里介绍其缓存方法，可以加速KubeFlow部署的过程。本文实现了一个镜像名称的自动提取脚本。

更多参考：

1、准备配置文件

wget https://raw.githubusercontent.com/kubeflow/manifests/v1.2-branch/kfdef/kfctl_k8s_istio.v1.2.0.yaml

修改其中的 manifest source为本地文件（先下载到本地）：

  repos:
  - name: manifests
    uri: file:/home/supermap/openthings/kubeflow/v1.2.0.tar.gz
  version: v1.2-branch

改完后的kfctl_k8s_istio.v1.2.0.yaml长这样：

apiVersion: kfdef.apps.kubeflow.org/v1
kind: KfDef
metadata:
  clusterName: kubernetes
  creationTimestamp: null
  name: kubeflow
  namespace: kubeflow
spec:
  applications:
  - kustomizeConfig:
      repoRef:
        name: manifests
        path: namespaces/base
    name: namespaces
  - kustomizeConfig:
      repoRef:
        name: manifests
        path: application/v3
    name: application
  - kustomizeConfig:
      repoRef:
        name: manifests
        path: stacks/kubernetes/application/istio-1-3-1-stack
    name: istio-stack
  - kustomizeConfig:
      repoRef:
        name: manifests
        path: stacks/kubernetes/application/cluster-local-gateway-1-3-1
    name: cluster-local-gateway
  - kustomizeConfig:
      repoRef:
        name: manifests
        path: istio/istio/base
    name: istio
  - kustomizeConfig:
      repoRef:
        name: manifests
        path: stacks/kubernetes/application/cert-manager-crds
    name: cert-manager-crds
  - kustomizeConfig:
      repoRef:
        name: manifests
        path: stacks/kubernetes/application/cert-manager-kube-system-resources
    name: cert-manager-kube-system-resources
  - kustomizeConfig:
      repoRef:
        name: manifests
        path: stacks/kubernetes/application/cert-manager
    name: cert-manager
  - kustomizeConfig:
      repoRef:
        name: manifests
        path: stacks/kubernetes/application/add-anonymous-user-filter
    name: add-anonymous-user-filter
  - kustomizeConfig:
      repoRef:
        name: manifests
        path: metacontroller/base
    name: metacontroller
  - kustomizeConfig:
      repoRef:
        name: manifests
        path: admission-webhook/bootstrap/overlays/application
    name: bootstrap
  - kustomizeConfig:
      repoRef:
        name: manifests
        path: stacks/kubernetes/application/spark-operator
    name: spark-operator
  - kustomizeConfig:
      repoRef:
        name: manifests
        path: stacks/kubernetes
    name: kubeflow-apps
  - kustomizeConfig:
      repoRef:
        name: manifests
        path: knative/installs/generic
    name: knative
  - kustomizeConfig:
      repoRef:
        name: manifests
        path: kfserving/installs/generic
    name: kfserving
  - kustomizeConfig:
      repoRef:
        name: manifests
        path: stacks/kubernetes/application/spartakus
    name: spartakus
  repos:
  - name: manifests
    uri: file:/home/supermap/openthings/kubeflow/v1.2.0.tar.gz
  version: v1.2-branch
status:
  reposCache:
  - localPath: '".cache/manifests/manifests-1.2.0"'
    name: manifests

把其中的 manifest下载下来：

wget https://github.com/kubeflow/manifests/archive/v1.2.0.tar.gz

然后，执行安装到Kubernetes集群：

kubectl apply -f kfctl_k8s_istio.v1.2.0.yaml

安装完后，会发现kubectl get pod -n kubeflow得到的pod没有正常启动，主要是因为镜像下载失败。

2、创建镜像缓存脚本

下一步我们创建脚本，把镜像缓存到本地。下面是使用Jupyter Lab来完成的脚本。

提取deployment的镜像：

%%bash
#提取deployment的镜像
depl=$(kubectl get deployment -n kubeflow -oname)
echo "#deployment " > imgs_list.txt
for d in ${depl}
do
    dimgs=$(kubectl describe $d -n kubeflow | grep "Image")
    dimg=(${dimgs//: / }) #(${dimgs//:/})
    #echo $dimg
    image_name=${dimg[1]}
    echo "docker pull "$image_name >> imgs_list.txt
    echo "docker tag "$image_name >> imgs_list.txt
done

提取replicaset的镜像：

%%bash
#提取replicaset的镜像
depl=$(kubectl get replicaset -n kubeflow -oname)
echo "#replicaset " >> imgs_list.txt
for d in ${depl}
do
    dimgs=$(kubectl describe $d -n kubeflow | grep "Image")
    dimg=(${dimgs//: / }) #(${dimgs//:/})
    #echo $dimg
    image_name=${dimg[1]}
    echo "docker pull "$image_name >> imgs_list.txt
    echo "docker tag "$image_name >> imgs_list.txt
done

提取statefulset的镜像：

%%bash
#提取statefulset的镜像
depl=$(kubectl get statefulset -n kubeflow -oname)
echo "#statefulset " >> imgs_list.txt
for d in ${depl}
do
    dimgs=$(kubectl describe $d -n kubeflow | grep "Image")
    dimg=(${dimgs//: / }) #(${dimgs//:/})
    #echo $dimg
    image_name=${dimg[1]}
    echo "docker pull "$image_name >> imgs_list.txt
    echo "docker tag "$image_name >> imgs_list.txt
done

通过上面的操作，所有kubeflow需要的镜像名称被提取到了imgs_list.txt文件中。

3、pull images list 以及 push 到 aliyun

修改脚本，把镜像pull下来，然后push到aliyun的镜像仓库，其它节点就可以从aliyun直接拉取了。

注意：以下脚本需要修改成自己的镜像仓库名称，有的仓库需要提前登录。

pull to local

下面的脚本需要在可以访问gcr.io的机器上执行。

#deployment 
docker pull gcr.io/kubeflow-images-public/admission-webhook:vmaster-ge5452b6f
docker pull argoproj/argoui:v2.3.0
docker pull gcr.io/ml-pipeline/cache-deployer:1.0.4
docker pull gcr.io/ml-pipeline/cache-server:1.0.4
docker pull gcr.io/kubeflow-images-public/centraldashboard:vmaster-g8097cfeb
docker pull gcr.io/kubeflow-images-public/jupyter-web-app:vmaster-g845af298
docker pull docker.io/kubeflowkatib/katib-controller:v1beta1-a96ff59
docker pull docker.io/kubeflowkatib/katib-db-manager:v1beta1-a96ff59
docker pull mysql:8
docker pull docker.io/kubeflowkatib/katib-ui:v1beta1-a96ff59
docker pull python:3.7
docker pull mysql:8.0.3
docker pull gcr.io/ml-pipeline/envoy:metadata-grpc
docker pull gcr.io/tfx-oss-public/ml_metadata_store_server:v0.21.1
docker pull gcr.io/ml-pipeline/metadata-writer:1.0.4
docker pull gcr.io/ml-pipeline/minio:RELEASE.2019-08-14T20-37-41Z-license-compliance
docker pull gcr.io/ml-pipeline/api-server:1.0.4
docker pull gcr.io/ml-pipeline/persistenceagent:1.0.4
docker pull gcr.io/ml-pipeline/scheduledworkflow:1.0.4
docker pull gcr.io/ml-pipeline/frontend:1.0.4
docker pull gcr.io/ml-pipeline/viewer-crd-controller:1.0.4
docker pull gcr.io/ml-pipeline/visualization-server:1.0.4
docker pull mpioperator/mpi-operator:latest
docker pull kubeflow/mxnet-operator:v1.0.0-20200625
docker pull gcr.io/ml-pipeline/mysql:5.6
docker pull gcr.io/kubeflow-images-public/notebook-controller:vmaster-g6eb007d0
docker pull gcr.io/kubeflow-images-public/profile-controller:vmaster-ga49f658f
docker pull gcr.io/kubeflow-images-public/pytorch-operator:vmaster-g518f9c76
docker pull docker.io/seldonio/seldon-core-operator:1.4.0
docker pull gcr.io/spark-operator/spark-operator:v1beta2-1.1.0-2.4.5
docker pull gcr.io/google_containers/spartakus-amd64:v1.1.0
docker pull gcr.io/kubeflow-images-public/tf_operator:vmaster-gda226016
docker pull argoproj/workflow-controller:v2.3.0

#===============================================
docker pull gcr.io/kubeflow-images-public/kfam:vmaster-g9f3bfd00
#docker save gcr.io/kubeflow-images-public/kfam:vmaster-g9f3bfd00 -o ./kfam.tar

docker pull gcr.io/kfserving/kfserving-controller:v0.4.1
#docker save gcr.io/kfserving/kfserving-controller:v0.4.1 -o ./kfserving-controller.tar
#===============================================

#statefulset 
docker pull gcr.io/kubeflow-images-public/ingress-setup:latest
docker pull gcr.io/kubeflow-images-public/kubernetes-sigs/application:1.0-beta
docker pull gcr.io/kubebuilder/kube-rbac-proxy:v0.4.0
docker pull metacontroller/metacontroller:v0.3.0

push到aliyun

MY_REGISTRY=registry.cn-hangzhou.aliyuncs.com/openthings/kfimgs1.2-

#######deployent ########################
docker tag gcr.io/kubeflow-images-public/admission-webhook:vmaster-ge5452b6f ${MY_REGISTRY}admission-webhook
docker push ${MY_REGISTRY}admission-webhook

docker tag argoproj/argoui:v2.3.0 ${MY_REGISTRY}argoui
docker push ${MY_REGISTRY}argoui 

docker tag gcr.io/ml-pipeline/cache-deployer:1.0.4 ${MY_REGISTRY}cache-deployer
docker push ${MY_REGISTRY}cache-deployer

docker tag gcr.io/ml-pipeline/cache-server:1.0.4 ${MY_REGISTRY}cache-server
docker push ${MY_REGISTRY}cache-server

docker tag gcr.io/kubeflow-images-public/centraldashboard:vmaster-g8097cfeb ${MY_REGISTRY}centraldashboard
docker push ${MY_REGISTRY}centraldashboard

docker tag gcr.io/kubeflow-images-public/jupyter-web-app:vmaster-g845af298 ${MY_REGISTRY}jupyter-web-app
docker push ${MY_REGISTRY}jupyter-web-app

docker tag docker.io/kubeflowkatib/katib-controller:v1beta1-a96ff59 ${MY_REGISTRY}katib-controller
docker push ${MY_REGISTRY}katib-controller

docker tag docker.io/kubeflowkatib/katib-db-manager:v1beta1-a96ff59 ${MY_REGISTRY}katib-db-manager
docker push ${MY_REGISTRY}katib-db-manager

docker tag mysql:8 ${MY_REGISTRY}mysql8
docker push ${MY_REGISTRY}mysql8

docker tag docker.io/kubeflowkatib/katib-ui:v1beta1-a96ff59 ${MY_REGISTRY}katib-ui
docker push ${MY_REGISTRY}katib-ui

docker tag python:3.7 ${MY_REGISTRY}python3.7
docker push ${MY_REGISTRY}python3.7

docker tag mysql:8.0.3 ${MY_REGISTRY}mysql8.0.3
docker push ${MY_REGISTRY}mysql8.0.3

docker tag gcr.io/ml-pipeline/envoy:metadata-grpc ${MY_REGISTRY}envoy
docker push ${MY_REGISTRY}envoy

docker tag gcr.io/tfx-oss-public/ml_metadata_store_server:v0.21.1 ${MY_REGISTRY}ml_metadata_store_server
docker push ${MY_REGISTRY}ml_metadata_store_server

docker tag gcr.io/ml-pipeline/metadata-writer:1.0.4 ${MY_REGISTRY}metadata-writer
docker push ${MY_REGISTRY}metadata-writer

docker tag gcr.io/ml-pipeline/minio:RELEASE.2019-08-14T20-37-41Z-license-compliance ${MY_REGISTRY}minio
docker push ${MY_REGISTRY}minio

docker tag gcr.io/ml-pipeline/api-server:1.0.4 ${MY_REGISTRY}api-server
docker push ${MY_REGISTRY}api-server

docker tag gcr.io/ml-pipeline/persistenceagent:1.0.4 ${MY_REGISTRY}persistenceagent
docker push ${MY_REGISTRY}persistenceagent

docker tag gcr.io/ml-pipeline/scheduledworkflow:1.0.4 ${MY_REGISTRY}scheduledworkflow
docker push ${MY_REGISTRY}scheduledworkflow

docker tag gcr.io/ml-pipeline/frontend:1.0.4 ${MY_REGISTRY}frontend
docker push ${MY_REGISTRY}frontend

docker tag gcr.io/ml-pipeline/viewer-crd-controller:1.0.4 ${MY_REGISTRY}viewer-crd-controller
docker push ${MY_REGISTRY}viewer-crd-controller

docker tag gcr.io/ml-pipeline/visualization-server:1.0.4 ${MY_REGISTRY}visualization-server
docker push ${MY_REGISTRY}visualization-server

docker tag mpioperator/mpi-operator:latest ${MY_REGISTRY}mpi-operator
docker push ${MY_REGISTRY}mpi-operator

docker tag kubeflow/mxnet-operator:v1.0.0-20200625 ${MY_REGISTRY}mxnet-operator
docker push ${MY_REGISTRY}mxnet-operator

docker tag gcr.io/ml-pipeline/mysql:5.6 ${MY_REGISTRY}mysql5.6
docker push ${MY_REGISTRY}mysql5.6

docker tag gcr.io/kubeflow-images-public/notebook-controller:vmaster-g6eb007d0 ${MY_REGISTRY}notebook-controller
docker push ${MY_REGISTRY}notebook-controller

docker tag gcr.io/kubeflow-images-public/profile-controller:vmaster-ga49f658f ${MY_REGISTRY}profile-controller
docker push ${MY_REGISTRY}profile-controller

docker tag gcr.io/kubeflow-images-public/pytorch-operator:vmaster-g518f9c76 ${MY_REGISTRY}pytorch-operator
docker push ${MY_REGISTRY}pytorch-operator

docker tag docker.io/seldonio/seldon-core-operator:1.4.0 ${MY_REGISTRY}seldon-core-operator
docker push ${MY_REGISTRY}seldon-core-operator

docker tag gcr.io/spark-operator/spark-operator:v1beta2-1.1.0-2.4.5 ${MY_REGISTRY}spark-operator
docker push ${MY_REGISTRY}spark-operator

docker tag gcr.io/google_containers/spartakus-amd64:v1.1.0 ${MY_REGISTRY}spartakus-amd64
docker push ${MY_REGISTRY}spartakus-amd64

docker tag gcr.io/kubeflow-images-public/tf_operator:vmaster-gda226016 ${MY_REGISTRY}tf_operator
docker push ${MY_REGISTRY}tf_operator

docker tag argoproj/workflow-controller:v2.3.0 ${MY_REGISTRY}workflow-controller
docker push ${MY_REGISTRY}workflow-controller

#===============================================
docker tag gcr.io/kubeflow-images-public/kfam:vmaster-g9f3bfd00 ${MY_REGISTRY}kfam
docker push ${MY_REGISTRY}kfam

docker tag gcr.io/kfserving/kfserving-controller:v0.4.1 ${MY_REGISTRY}kfserving-controller
docker push ${MY_REGISTRY}kfserving-controller
#===============================================

#statefulset 
docker tag gcr.io/kubeflow-images-public/ingress-setup:latest ${MY_REGISTRY}ingress-setup
docker push ${MY_REGISTRY}ingress-setup

docker tag gcr.io/kubeflow-images-public/kubernetes-sigs/application:1.0-beta ${MY_REGISTRY}application
docker push ${MY_REGISTRY}application

docker tag gcr.io/kubebuilder/kube-rbac-proxy:v0.4.0 ${MY_REGISTRY}kube-rbac-proxy
docker push ${MY_REGISTRY}kube-rbac-proxy

docker tag metacontroller/metacontroller:v0.3.0 ${MY_REGISTRY}metacontroller
docker push ${MY_REGISTRY}metacontroller

4、从aliyun拉取到本地

从aliyun拉取到本地，然后tag为原来的名称。

注意：改为自己的镜像仓库名称。

MY_REGISTRY=registry.cn-hangzhou.aliyuncs.com/openthings/kfimgs1.2-

#######deployent ########################
docker pull ${MY_REGISTRY}admission-webhook
docker tag ${MY_REGISTRY}admission-webhook gcr.io/kubeflow-images-public/admission-webhook:vmaster-ge5452b6f 

docker pull ${MY_REGISTRY}argoui 
docker tag ${MY_REGISTRY}argoui argoproj/argoui:v2.3.0

docker pull ${MY_REGISTRY}cache-deployer
docker tag ${MY_REGISTRY}cache-deployer gcr.io/ml-pipeline/cache-deployer:1.0.4

docker pull ${MY_REGISTRY}cache-server
docker tag ${MY_REGISTRY}cache-server gcr.io/ml-pipeline/cache-server:1.0.4

docker pull ${MY_REGISTRY}centraldashboard
docker tag ${MY_REGISTRY}centraldashboard gcr.io/kubeflow-images-public/centraldashboard:vmaster-g8097cfeb

docker pull ${MY_REGISTRY}jupyter-web-app
docker tag ${MY_REGISTRY}jupyter-web-app gcr.io/kubeflow-images-public/jupyter-web-app:vmaster-g845af298

###======
docker pull ${MY_REGISTRY}katib-controller
docker tag ${MY_REGISTRY}katib-controller docker.io/kubeflowkatib/katib-controller:v1beta1-a96ff59

docker pull ${MY_REGISTRY}katib-db-manager
docker tag ${MY_REGISTRY}katib-db-manager docker.io/kubeflowkatib/katib-db-manager:v1beta1-a96ff59

docker pull ${MY_REGISTRY}mysql8
docker tag ${MY_REGISTRY}mysql8 mysql:8

docker pull ${MY_REGISTRY}katib-ui
docker tag ${MY_REGISTRY}katib-ui docker.io/kubeflowkatib/katib-ui:v1beta1-a96ff59

docker pull ${MY_REGISTRY}python3.7
docker tag ${MY_REGISTRY}python3.7 python:3.7

docker pull ${MY_REGISTRY}mysql8.0.3
docker tag ${MY_REGISTRY}mysql8.0.3 mysql:8.0.3

docker pull ${MY_REGISTRY}envoy
docker tag ${MY_REGISTRY}envoy gcr.io/ml-pipeline/envoy:metadata-grpc

###
docker pull ${MY_REGISTRY}ml_metadata_store_server
docker tag ${MY_REGISTRY}ml_metadata_store_server gcr.io/tfx-oss-public/ml_metadata_store_server:v0.21.1 

docker pull ${MY_REGISTRY}metadata-writer
docker tag ${MY_REGISTRY}metadata-writer gcr.io/ml-pipeline/metadata-writer:1.0.4

docker pull ${MY_REGISTRY}minio
docker tag ${MY_REGISTRY}minio gcr.io/ml-pipeline/minio:RELEASE.2019-08-14T20-37-41Z-license-compliance

docker pull ${MY_REGISTRY}api-server
docker tag ${MY_REGISTRY}api-server gcr.io/ml-pipeline/api-server:1.0.4

docker pull ${MY_REGISTRY}persistenceagent
docker tag ${MY_REGISTRY}persistenceagent gcr.io/ml-pipeline/persistenceagent:1.0.4
##
docker pull ${MY_REGISTRY}scheduledworkflow
docker tag ${MY_REGISTRY}scheduledworkflow gcr.io/ml-pipeline/scheduledworkflow:1.0.4

docker pull ${MY_REGISTRY}frontend
docker tag ${MY_REGISTRY}frontend gcr.io/ml-pipeline/frontend:1.0.4

docker pull ${MY_REGISTRY}viewer-crd-controller
docker tag ${MY_REGISTRY}viewer-crd-controller gcr.io/ml-pipeline/viewer-crd-controller:1.0.4

docker pull ${MY_REGISTRY}visualization-server
docker tag ${MY_REGISTRY}visualization-server gcr.io/ml-pipeline/visualization-server:1.0.4

docker pull ${MY_REGISTRY}mpi-operator
docker tag ${MY_REGISTRY}mpi-operator mpioperator/mpi-operator:latest

docker pull ${MY_REGISTRY}mxnet-operator
docker tag ${MY_REGISTRY}mxnet-operator kubeflow/mxnet-operator:v1.0.0-20200625

docker pull ${MY_REGISTRY}mysql5.6
docker tag ${MY_REGISTRY}mysql5.6 gcr.io/ml-pipeline/mysql:5.6

##
docker pull ${MY_REGISTRY}notebook-controller
docker tag ${MY_REGISTRY}notebook-controller gcr.io/kubeflow-images-public/notebook-controller:vmaster-g6eb007d0

docker pull ${MY_REGISTRY}profile-controller
docker tag ${MY_REGISTRY}profile-controller gcr.io/kubeflow-images-public/profile-controller:vmaster-ga49f658f

docker pull ${MY_REGISTRY}pytorch-operator
docker tag ${MY_REGISTRY}pytorch-operator gcr.io/kubeflow-images-public/pytorch-operator:vmaster-g518f9c76

docker pull ${MY_REGISTRY}seldon-core-operator
docker tag ${MY_REGISTRY}seldon-core-operator docker.io/seldonio/seldon-core-operator:1.4.0

docker pull ${MY_REGISTRY}spark-operator
docker tag ${MY_REGISTRY}spark-operator gcr.io/spark-operator/spark-operator:v1beta2-1.1.0-2.4.5

docker pull ${MY_REGISTRY}spartakus-amd64
docker tag ${MY_REGISTRY}spartakus-amd64 gcr.io/google_containers/spartakus-amd64:v1.1.0

docker pull ${MY_REGISTRY}tf_operator
docker tag ${MY_REGISTRY}tf_operator gcr.io/kubeflow-images-public/tf_operator:vmaster-gda226016

docker pull ${MY_REGISTRY}workflow-controller
docker tag ${MY_REGISTRY}workflow-controller argoproj/workflow-controller:v2.3.0

#===============================================
docker pull ${MY_REGISTRY}kfam
docker tag ${MY_REGISTRY}kfam gcr.io/kubeflow-images-public/kfam:vmaster-g9f3bfd00 

docker pull ${MY_REGISTRY}kfserving-controller
docker tag ${MY_REGISTRY}kfserving-controller gcr.io/kfserving/kfserving-controller:v0.4.1
#===============================================

#statefulset 
docker pull ${MY_REGISTRY}ingress-setup
docker tag ${MY_REGISTRY}ingress-setup gcr.io/kubeflow-images-public/ingress-setup:latest

docker pull ${MY_REGISTRY}application
docker tag ${MY_REGISTRY}application gcr.io/kubeflow-images-public/kubernetes-sigs/application:1.0-beta

docker pull ${MY_REGISTRY}kube-rbac-proxy
docker tag ${MY_REGISTRY}kube-rbac-proxy gcr.io/kubebuilder/kube-rbac-proxy:v0.4.0

docker pull ${MY_REGISTRY}metacontroller
docker tag ${MY_REGISTRY}metacontroller metacontroller/metacontroller:v0.3.0

5、save 到tar文件

将所有的镜像保存为tar文件，就可以在各个kubernetes节点下载然后load进去。

脚本如下：

#######deployent ########################
docker save gcr.io/kubeflow-images-public/admission-webhook:vmaster-ge5452b6f -o ./kfimgs1.2/admission-webhook.tar
docker save argoproj/argoui:v2.3.0 -o ./kfimgs1.2/argoui.tar
docker save gcr.io/ml-pipeline/cache-deployer:1.0.4 -o ./kfimgs1.2/cache-deployer.tar
docker save gcr.io/ml-pipeline/cache-server:1.0.4 -o ./kfimgs1.2/cache-server.tar
docker save gcr.io/kubeflow-images-public/centraldashboard:vmaster-g8097cfeb -o ./kfimgs1.2/centraldashboard.tar

docker save gcr.io/kubeflow-images-public/jupyter-web-app:vmaster-g845af298 -o ./kfimgs1.2/jupyter-web-app.tar
docker save docker.io/kubeflowkatib/katib-controller:v1beta1-a96ff59 -o ./kfimgs1.2/katib-controller.tar
docker save docker.io/kubeflowkatib/katib-db-manager:v1beta1-a96ff59 -o ./kfimgs1.2/katib-db-manager.tar

docker save mysql:8 -o ./kfimgs1.2/mysql8.tar
docker save docker.io/kubeflowkatib/katib-ui:v1beta1-a96ff59 -o ./kfimgs1.2/katib-ui.tar
docker save python:3.7 -o ./kfimgs1.2/python3.7.tar
docker save mysql:8.0.3 -o ./kfimgs1.2/mysql8.0.3.tar

docker save gcr.io/ml-pipeline/envoy:metadata-grpc -o ./kfimgs1.2/envoy.tar
docker save gcr.io/tfx-oss-public/ml_metadata_store_server:v0.21.1 -o ./kfimgs1.2/ml_metadata_store_server.tar
docker save gcr.io/ml-pipeline/metadata-writer:1.0.4 -o ./kfimgs1.2/metadata-writer.tar
docker save gcr.io/ml-pipeline/minio:RELEASE.2019-08-14T20-37-41Z-license-compliance -o ./kfimgs1.2/minio.tar
docker save gcr.io/ml-pipeline/api-server:1.0.4 -o ./kfimgs1.2/api-server.tar
docker save gcr.io/ml-pipeline/persistenceagent:1.0.4 -o ./kfimgs1.2/persistenceagent.tar
docker save gcr.io/ml-pipeline/scheduledworkflow:1.0.4 -o ./kfimgs1.2/scheduledworkflow.tar
docker save gcr.io/ml-pipeline/frontend:1.0.4 -o ./kfimgs1.2/frontend.tar
docker save gcr.io/ml-pipeline/viewer-crd-controller:1.0.4 -o ./kfimgs1.2/viewer-crd-controller.tar

docker save gcr.io/ml-pipeline/visualization-server:1.0.4 -o ./kfimgs1.2/visualization-server.tar
docker save mpioperator/mpi-operator:latest -o ./kfimgs1.2/mpi-operator.tar
docker save kubeflow/mxnet-operator:v1.0.0-20200625 -o ./kfimgs1.2/mxnet-operator.tar
docker save gcr.io/ml-pipeline/mysql:5.6 -o ./kfimgs1.2/mysql5.6.tar
docker save gcr.io/kubeflow-images-public/notebook-controller:vmaster-g6eb007d0 -o ./kfimgs1.2/notebook-controller.tar

docker save gcr.io/kubeflow-images-public/profile-controller:vmaster-ga49f658f -o ./kfimgs1.2/profile-controller.tar
docker save gcr.io/kubeflow-images-public/pytorch-operator:vmaster-g518f9c76 -o ./kfimgs1.2/pytorch-operator.tar
docker save docker.io/seldonio/seldon-core-operator:1.4.0 -o ./kfimgs1.2/seldon-core-operator.tar
docker save gcr.io/spark-operator/spark-operator:v1beta2-1.1.0-2.4.5 -o ./kfimgs1.2/spark-operator.tar

docker save gcr.io/google_containers/spartakus-amd64:v1.1.0 -o ./kfimgs1.2/spartakus-amd64.tar
docker save gcr.io/kubeflow-images-public/tf_operator:vmaster-gda226016 -o ./kfimgs1.2/tf_operator.tar
docker save argoproj/workflow-controller:v2.3.0 -o ./kfimgs1.2/workflow-controller.tar

#===============================================
docker save gcr.io/kubeflow-images-public/kfam:vmaster-g9f3bfd00 -o ./kfam.tar
docker save gcr.io/kfserving/kfserving-controller:v0.4.1 -o ./kfserving-controller.tar
#===============================================

#statefulset 
docker save gcr.io/kubeflow-images-public/ingress-setup:latest -o ./kfimgs1.2/ingress-setup.tar
docker save gcr.io/kubeflow-images-public/kubernetes-sigs/application:1.0-beta -o ./kfimgs1.2/application.tar
df -h
docker save gcr.io/kubebuilder/kube-rbac-proxy:v0.4.0 -o ./kfimgs1.2/kube-rbac-proxy.tar
docker save metacontroller/metacontroller:v0.3.0 -o ./kfimgs1.2/metacontroller.tar

6、load到docker images

从*.tar载入到docker系统中，在microk8s上使用juju进行安装（使用containerd），不需要使用这里的方法。如果安装失败，是访问juju服务器出现问题。

#######deployent ########################
docker load -i ./kfimgs1.2/admission-webhook.tar
docker load -i ./kfimgs1.2/argoui.tar
docker load -i ./kfimgs1.2/cache-deployer.tar
docker load -i ./kfimgs1.2/cache-server.tar
docker load -i ./kfimgs1.2/centraldashboard.tar
docker load -i ./kfimgs1.2/jupyter-web-app.tar
docker load -i ./kfimgs1.2/katib-controller.tar
docker load -i ./kfimgs1.2/katib-db-manager.tar
docker load -i ./kfimgs1.2/mysql8.tar
docker load -i ./kfimgs1.2/katib-ui.tar
docker load -i ./kfimgs1.2/python3.7.tar
docker load -i ./kfimgs1.2/mysql8.0.3.tar
docker load -i ./kfimgs1.2/envoy.tar
docker load -i ./kfimgs1.2/ml_metadata_store_server.tar
docker load -i ./kfimgs1.2/metadata-writer.tar
docker load -i ./kfimgs1.2/minio.tar
docker load -i ./kfimgs1.2/api-server.tar
docker load -i ./kfimgs1.2/persistenceagent.tar
docker load -i ./kfimgs1.2/scheduledworkflow.tar
docker load -i ./kfimgs1.2/frontend.tar
docker load -i ./kfimgs1.2/viewer-crd-controller.tar
docker load -i ./kfimgs1.2/visualization-server.tar
docker load -i ./kfimgs1.2/mpi-operator.tar
docker load -i ./kfimgs1.2/mxnet-operator.tar
docker load -i ./kfimgs1.2/mysql5.6.tar
docker load -i ./kfimgs1.2/notebook-controller.tar
docker load -i ./kfimgs1.2/profile-controller.tar
docker load -i ./kfimgs1.2/pytorch-operator.tar
docker load -i ./kfimgs1.2/seldon-core-operator.tar
docker load -i ./kfimgs1.2/spark-operator.tar
docker load -i ./kfimgs1.2/spartakus-amd64.tar
docker load -i ./kfimgs1.2/tf_operator.tar
docker load -i ./kfimgs1.2/workflow-controller.tar

#==================================================
docker load -i ./kfimgs1.2/kfam.tar
docker load -i ./kfimgs1.2/kfserving-controller.tar
#==================================================

#statefulset 
docker load -i ./kfimgs1.2/ingress-setup.tar
docker load -i ./kfimgs1.2/application.tar
docker load -i ./kfimgs1.2/kube-rbac-proxy.tar
docker load -i ./kfimgs1.2/metacontroller.tar

查看一下镜像是否已经进来：

#使用docker
docker images

#或者使用microk8s
microk8s.ctr images list

过一定时间后， kubernetes的pod将会自动重新创建，逐步进入正常运行。

更多参考：
- 基于国内阿里云镜像解决kubeflow一键安装，https://my.oschina.net/Kanonpy/blog/3147705
- kubeflow系列(一)：基于国内阿里云镜像解决kubeflow一键安装
  - https://zhuanlan.zhihu.com/p/102563248
  - https://github.com/shikanon/kubeflow-manifests
- kubeflow 1.0 安装教程，https://blog.flyfox.top/2020/04/04/kubeflow%E5%AE%89%E8%A3%85%E6%95%99%E7%A8%8B/

1、准备配置文件

2、创建镜像缓存脚本

3、pull images list 以及 push 到 aliyun

pull to local

push到aliyun

4、从aliyun拉取到本地

5、save 到tar文件

6、load到docker images

关于作者

作者的专辑

OSCHINA 社区

在线工具

攻略

QQ群

公众号

视频号

OSCHINA 小程序

OSCHINA 小程序 —— 关注技术领域的头条文章

KubeFlow 1.2.0镜像缓存（持续更新）

1、准备配置文件

2、创建镜像缓存脚本

3、pull images list 以及 push 到 aliyun

pull to local

push到aliyun

4、从aliyun拉取到本地

5、save 到tar文件

6、load到docker images

作者的其它热门文章

关于作者

作者的专辑

推荐关注

OSCHINA 社区

在线工具

攻略

QQ群

公众号

视频号

OSCHINA 小程序

OSCHINA 小程序 ——
关注技术领域的头条文章