Agent Connection

The component Tower of KubeSphere is used for agent connection. Tower is a tool for network connection between clusters through the agent. If the Host Cluster (H Cluster) cannot access the Member Cluster (M Cluster) directly, you can expose the proxy service address of the H cluster. This enables the M Cluster to connect to the H Cluster through the agent. This method is applicable when the M Cluster is in a private environment (for example, IDC) and the H Cluster is able to expose the proxy service. The agent connection is also applicable when your clusters are distributed across different cloud providers.

To use the multi-cluster feature using an agent, you must have at least two clusters serving as the H Cluster and the M Cluster respectively. A cluster can be defined as the H Cluster or the M Cluster either before or after you install KubeSphere. For more information about installing KubeSphere, refer to Installing on Linux and Installing on Kubernetes.

Video Demonstration

Prepare a Host Cluster

A host cluster provides you with the central control plane and you can only define one host cluster.

If you already have a standalone KubeSphere cluster installed, you can set the value of clusterRole to host by editing the cluster configuration.

  • Option A - Use the web console:

    Use the admin account to log in to the console and go to CRDs on the Cluster Management page. Enter the keyword ClusterConfiguration and go to its detail page. Edit the YAML of ks-installer, which is similar to Enable Pluggable Components.

  • Option B - Use Kubectl:

    kubectl edit cc ks-installer -n kubesphere-system
    

In the YAML file of ks-installer, navigate to multicluster, set the value of clusterRole to host, then click Update (if you use the web console) to make it effective:

multicluster:
  clusterRole: host

You need to wait for a while so that the change can take effect.

You can define a host cluster before you install KubeSphere either on Linux or on an existing Kubernetes cluster. If you want to install KubeSphere on Linux, you use a config-sample.yaml file. If you want to install KubeSphere on an existing Kubernetes cluster, you use two YAML files, one of which is cluster-configuration.yaml. To set a host cluster, change the value of clusterRole to host in config-sample.yaml or cluster-configuration.yaml accordingly before you install KubeSphere.

multicluster:
  clusterRole: host

Note

If you install KubeSphere on a single-node cluster (All-in-One), you do not need to create a config-sample.yaml file. In this case, you can set a host cluster after KubeSphere is installed.

You can use kubectl to retrieve the installation logs to verify the status by running the following command. Wait for a while, and you will be able to see the successful log return if the host cluster is ready.

kubectl logs -n kubesphere-system $(kubectl get pod -n kubesphere-system -l app=ks-install -o jsonpath='{.items[0].metadata.name}') -f

Set the Proxy Service Address

After the installation of the host cluster, a proxy service called tower will be created in kubesphere-system, whose type is LoadBalancer.

If a LoadBalancer plugin is available for the cluster, you can see a corresponding address for EXTERNAL-IP of tower, which will be acquired by KubeSphere. In this case, the proxy service is set automatically. That means you can skip the step to set the proxy. Execute the following command to verify if you have a LoadBalancer.

kubectl -n kubesphere-system get svc

The output is similar to this:

NAME       TYPE            CLUSTER-IP      EXTERNAL-IP     PORT(S)              AGE
tower      LoadBalancer    10.233.63.191   139.198.110.23  8080:30721/TCP       16h

Note

Generally, there is always a LoadBalancer solution in the public cloud, and the external IP can be allocated by the load balancer automatically. If your clusters are running in an on-premises environment, especially a bare metal environment, you can use PorterLB as the LB solution.
  1. If you cannot see a corresponding address displayed (EXTERNAL-IP is pending), you need to manually set the proxy address. For example, you have an available public IP address 139.198.120.120, and port 8080 of this IP address has been forwarded to port 30721 of the cluster. Execute the following command to check the service.

    kubectl -n kubesphere-system get svc
    

    The output is similar to this:

    NAME       TYPE            CLUSTER-IP      EXTERNAL-IP     PORT(S)              AGE
    tower      LoadBalancer    10.233.63.191   <pending>  8080:30721/TCP            16h
    
  2. Add the value of proxyPublishAddress to the configuration file of ks-installer and provide the public IP address (139.198.120.120 in this tutorial) and port number as follows.

    • Option A - Use the web console:

      Use the admin account to log in to the console and go to CRDs on the Cluster Management page. Enter the keyword ClusterConfiguration and go to its detail page. Edit the YAML of ks-installer, which is similar to Enable Pluggable Components.

    • Option B - Use Kubectl:

      kubectl -n kubesphere-system edit clusterconfiguration ks-installer
      

    Navigate to multicluster and add a new line for proxyPublishAddress to define the IP address to access tower.

    multicluster:
        clusterRole: host
        proxyPublishAddress: http://139.198.120.120:8080 # Add this line to set the address to access tower
    
  3. Save the configuration and wait for a while, or you can manually restart ks-apiserver to make the change effective immediately using the following command.

    kubectl -n kubesphere-system rollout restart deployment ks-apiserver
    

Prepare a Member Cluster

In order to manage the member cluster from the host cluster, you need to make jwtSecret the same between them. Therefore, get it first by excuting the following command on the host cluster.

kubectl -n kubesphere-system get cm kubesphere-config -o yaml | grep -v "apiVersion" | grep jwtSecret

The output may look like this:

jwtSecret: "gfIwilcc0WjNGKJ5DLeksf2JKfcLgTZU"

If you already have a standalone KubeSphere cluster installed, you can set the value of clusterRole to member by editing the cluster configuration.

  • Option A - Use the web console:

    Use the admin account to log in to the console and go to CRDs on the Cluster Management page. Enter the keyword ClusterConfiguration and go to its detail page. Edit the YAML of ks-installer, which is similar to Enable Pluggable Components.

  • Option B - Use Kubectl:

    kubectl edit cc ks-installer -n kubesphere-system
    

In the YAML file of ks-installer, enter the corresponding jwtSecret shown above:

authentication:
  jwtSecret: gfIwilcc0WjNGKJ5DLeksf2JKfcLgTZU

Scroll down and set the value of clusterRole to member, then click Update (if you use the web console) to make it effective:

multicluster:
  clusterRole: member

You need to wait for a while so that the change can take effect.

You can define a member cluster before you install KubeSphere either on Linux or on an existing Kubernetes cluster. If you want to install KubeSphere on Linux, you use a config-sample.yaml file. If you want to install KubeSphere on an existing Kubernetes cluster, you use two YAML files, one of which is cluster-configuration.yaml. To set a member cluster, enter the value of jwtSecret shown above and change the value of clusterRole to member in config-sample.yaml or cluster-configuration.yaml accordingly before you install KubeSphere.

authentication:
  jwtSecret: gfIwilcc0WjNGKJ5DLeksf2JKfcLgTZU
multicluster:
  clusterRole: member

Note

If you install KubeSphere on a single-node cluster (All-in-One), you do not need to create a config-sample.yaml file. In this case, you can set a member cluster after KubeSphere is installed.

You can use kubectl to retrieve the installation logs to verify the status by running the following command. Wait for a while, and you will be able to see the successful log return if the member cluster is ready.

kubectl logs -n kubesphere-system $(kubectl get pod -n kubesphere-system -l app=ks-install -o jsonpath='{.items[0].metadata.name}') -f

Import a Member Cluster

  1. Log in to the KubeSphere console as admin and click Add Cluster on the Cluster Management page.

    add-cluster

  2. Enter the basic information of the cluster to be imported on the Import Cluster page. You can also click Edit Mode in the top-right corner to view and edit the basic information in YAML format. After you finish editing, click Next.

    cluster-info

  3. In Connection Method, select Agent Connection and click Create. It will show the YAML configuration file for the agent Deployment generated by the H Cluster on the console.

    select-agent-connection

  4. Create an agent.yaml file on the M Cluster based on the instruction, then copy and paste the agent deployment to the file. Execute kubectl create -f agent.yaml on the node and wait for the agent to be up and running. Please make sure the proxy address is accessible to the M Cluster.

  5. You can see the cluster you have imported in the H Cluster when the cluster agent is up and running.

    cluster-imported


Thanks for the feedback. If you have a specific question about how to use KubeSphere, ask it on Slack. Open an issue in the GitHub repo if you want to report a problem or suggest an improvement.