Migrate Cluster from another platform to VKS

To migrate a Cluster from the Cloud Provider or On-premise system to the VKS system, follow the steps in this document.

Prerequisites

Perform download helper bash script and grand execute permission for this file ( velero_helper.sh )

(Optional) Deploy some services to check the correctness of the migration. Suppose, at the source Cluster, I have deployed an nginx service as follows:

Deployment files:

apiVersion: v1
kind: Service
metadata:
  name: nginx
  namespace: mynamespace
  labels:
    app: nginx
spec:
  ports:
  - port: 80
    name: web
  selector:
    app: nginx
  type: NodePort
---
apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: web
  namespace: mynamespace
spec:
  selector:
    matchLabels:
      app: nginx
  serviceName: "nginx"
  replicas: 1
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
      - name: nginx
        image: nginx
        ports:
        - containerPort: 80
          name: web
        volumeMounts:
        - name: disk-ssd
          mountPath: /usr/share/nginx/html
  volumeClaimTemplates:
  - metadata:
      name: disk-ssd
      namespace: mynamespace
    spec:
      accessModes: [ "ReadWriteOnce" ]
      storageClassName: standard-rwo
      resources:
        requests:
          storage: 40Gi

kubectl exec -n mynamespace -it web-0 bash
cd /usr/share/nginx/html
echo -e "<html>\n<head>\n  <title>MyVNGCloud</title>\n</head>\n<body>\n  <h1>Hello, MyVNGCloud</h1>\n</body>\n</html>" > index.html

Now, when you access Node's Public IP, you will see "Hello, MyVNGCloud".

Prepare target cluster (Prepare target resource)

On the VKS system, you need to initialize a Cluster according to the instructions here . Make sure that the destination cluster's configuration is the same as the source cluster's configuration.

Attention:

For the migration to be successful, on the target Cluster, you need to ensure the following requirements:

The amount of resources needed such as number of nodes, node instance configuration,...
Node labels and node taints are the same as the old cluster.
Corresponding or alternative Storage Class.

[Optional] Migrate private resources outside cluster

Migrating private resources outside cluster (moving private resources outside the cluster) is the process of moving private resources outside the source Cluster to a place that the destination Cluster can use. For example, you may have private resources such as images, databases, etc. Now, before starting to migrate, you need to migrate these resources yourself. For example, if you need:

Migrate Container Images: you can migrate images to VNGCloud Container Registry through instructions here .
Migrate Databases: you can use Relational Database Service (RDS) and Object Storage Service (OBS) depending on your needs. After the migration is complete, remember to reconfigure the database for your applications on VKS Cluster.
Migrate Storage: you can use vServer's NFS Server .

Attention:

After you migrate resources outside the Cluster, you need to ensure the target Cluster can connect to these migrated resources.

Install Velero on both source and destination clusters (Install Velero tool)

After you have migrated private resources outside the cluster, you can use the migration tool to backup and restore the application on the source cluster and target cluster.

Create a vStorage Project, Container to receive the cluster's backup data according to instructions here .
Create an S3 key corresponding to this vStorage Project according to the instructions here .

For example, I have initialized a vStorage Project, Container with the following information: Region: HCM03, Container: mycontainer, Endpoint: https://hcm03.vstorage.vngcloud.vn .

On both Clusters (source and target)

Create file credentials-velero with the following content:

[default]
aws_access_key_id=________________________
aws_secret_access_key=________________________

Install Velero CLI:

curl -OL https://github.com/vmware-tanzu/velero/releases/download/v1.13.2/velero-v1.13.2-linux-amd64.tar.gz
tar -xvf velero-v1.13.2-linux-amd64.tar.gz
cp velero-v1.13.2-linux-amd64/velero /usr/local/bin

Install Velero on your 2 clusters with the command:

velero install \
    --provider aws \
    --plugins velero/velero-plugin-for-aws:v1.9.0 \
    --use-node-agent \
    --use-volume-snapshots=false \
    --secret-file ./credentials-velero \
    --bucket ______________________________ \
    --backup-location-config region=hcm03,s3ForcePathStyle="true",s3Url=https://hcm03.vstorage.vngcloud.vn

For Clusters on Amazon Elastic Kubernetes Service (EKS)

At the source Cluster

Annotate the Persistent Volumes that need to be backed up. By default, Velero will not backup volume. You can run the command below to annotate backup of all volumes.
```
./velero_helper.sh mark_volume -c
```
Additionally, you can mark not to backup system resources with the following command:
```
./velero_helper.sh mark_exclude -c
```

Perform backup according to the syntax:

velero backup create eks-cluster --include-namespaces "" \
  --include-cluster-resources=true \
  --wait

velero backup create eks-namespace --exclude-namespaces velero \
    --wait

Attention:

You must create 2 backup versions for Cluster Resource and Namespace Resource.

At the destination Cluster

Create a Storage Class mapping file between source and destination Cluster:

apiVersion: v1
kind: ConfigMap
metadata:
  name: change-storage-class-config
  namespace: velero
  labels:
    velero.io/plugin-config: ""
    velero.io/change-storage-class: RestoreItemAction
data:
  _______old_storage_class_______: _______new_storage_class_______  # <= Adjust here
  _______old_storage_class_______: _______new_storage_class_______  # <= Adjust here

Perform restore according to the command:

velero restore create --item-operation-timeout 1m --from-backup eks-cluster \
    --exclude-resources="MutatingWebhookConfiguration,ValidatingWebhookConfiguration"

velero restore create --item-operation-timeout 1m --from-backup eks-namespace

velero restore create --item-operation-timeout 1m --from-backup eks-cluster

For Cluster on Google Kubernetes Engine (GKE)

At the source Cluster

Annotate the Persistent Volumes and label resources that need to be excluded from the backup
```
./velero_helper.sh mark_volume -c
```
Additionally, you can mark not to backup system resources with the following command:
```
./velero_helper.sh mark_exclude -c
```

Perform backup according to the syntax:

velero backup create gke-cluster --include-namespaces "" \
  --include-cluster-resources=true \
  --wait

velero backup create gke-namespace --exclude-namespaces velero \
    --wait

Attention:

You must create 2 backup versions for Cluster Resource and Namespace Resource.

At the destination Cluster

Create a Storage Class mapping file between source and destination Cluster:

apiVersion: v1
kind: ConfigMap
metadata:
  name: change-storage-class-config
  namespace: velero
  labels:
    velero.io/plugin-config: ""
    velero.io/change-storage-class: RestoreItemAction
data:
  _______old_storage_class_______: _______new_storage_class_______  # <= Adjust here
  _______old_storage_class_______: _______new_storage_class_______  # <= Adjust here

Perform restore according to the command:

velero restore create --item-operation-timeout 1m --from-backup gke-cluster \
    --exclude-resources="MutatingWebhookConfiguration,ValidatingWebhookConfiguration"

velero restore create --item-operation-timeout 1m --from-backup gke-namespace

velero restore create --item-operation-timeout 1m --from-backup gke-cluster

Attention:

Google Kubernetes Engine (GKE) does not allow daemonset deployment on all nodes. However, Velero only needs to deploy the daemonset on the node with the PV mount. The solution to this problem is that you can adjust the taint and toleration of the daemonset to only deploy it on the node with the PV mount.
You can change the default resource request cpu:500mand mem:512Min the installation step or make adjustments when deploying the yaml.

PreviousMigration Cluster from vContainer to VKS NextMigrate Limitation

Last updated 10 months ago