Add Edge Nodes

KubeSphere leverages KubeEdge, to extend native containerized application orchestration capabilities to hosts at edge. With separate cloud and edge core modules, KubeEdge provides complete edge computing solutions while the installation may be complex and difficult.

kubeedge_arch

Note

For more information about different components of KubeEdge, see the KubeEdge documentation.

After an edge node joins your cluster, the native KubeEdge cloud component requires you to manually configure iptables so that you can use commands such as kubectl logs and kubectl exec. In this connection, KubeSphere features an efficient and convenient way to add edge nodes to a Kubernetes cluster. It uses supporting components (for example, EdgeWatcher) to automatically configure iptables.

edge-watcher

This tutorial demonstrates how to add an edge node to your cluster.

Prerequisites

  • You have enabled KubeEdge.
  • You have an available node to serve as an edge node. The node can run either Ubuntu (recommended) or CentOS. This tutorial uses Ubuntu 18.04 as an example.
  • Edge nodes, unlike Kubernetes cluster nodes, should work in a separate network.

Configure an Edge Node

You need to install a container runtime and configure EdgeMesh on your edge node.

Install a container runtime

KubeEdge supports several container runtimes including Docker, containerd, CRI-O and Virtlet. For more information, see the KubeEdge documentation.

Note

If you use Docker as the container runtime for your edge node, Docker v19.3.0 or later must be installed so that KubeSphere can get Pod metrics of it.

Configure EdgeMesh

Perform the following steps to configure EdgeMesh on your edge node.

  1. Edit /etc/nsswitch.conf.

    vi /etc/nsswitch.conf
    
  2. Add the following content to this file:

    hosts:          dns files mdns4_minimal [NOTFOUND=return]
    
  3. Save the file and run the following command to enable IP forwarding:

    sudo echo "net.ipv4.ip_forward = 1" >> /etc/sysctl.conf
    
  4. Verify your modification:

    sudo sysctl -p | grep ip_forward
    

    Expected result:

    net.ipv4.ip_forward = 1
    

Create Firewall Rules and Port Forwarding Rules

To make sure edge nodes can successfully talk to your cluster, you must forward ports for outside traffic to get into your network. Specifically, map an external port to the corresponding internal IP address (master node) and port based on the table below. Besides, you also need to create firewall rules to allow traffic to these ports (10000 to 10004).

Fields External Ports Fields Internal Ports
cloudhubPort 10000 cloudhubNodePort 30000
cloudhubQuicPort 10001 cloudhubQuicNodePort 30001
cloudhubHttpsPort 10002 cloudhubHttpsNodePort 30002
cloudstreamPort 10003 cloudstreamNodePort 30003
tunnelPort 10004 tunnelNodePort 30004

Add an Edge Node

  1. Log in to the console as admin and click Platform in the top-left corner.

  2. Select Cluster Management and navigate to Edge Nodes under Node Management.

    Note

    If you have enabled multi-cluster management, you need to select a cluster first.
  3. Click Add Node. In the dialog that appears, set a node name and enter an internal IP address of your edge node. Click Validate to continue.

    Note

    • The internal IP address is only used for inter-node communication and you do not necessarily need to use the actual internal IP address of the edge node. As long as the IP address is successfully validated, you can use it.
    • It is recommended that you check the box to add the default taint.
  4. Copy the command automatically created under Add Command and run it on your edge node.

    Note

    Make sure wget is installed on your edge node before you run the command.

    edge-node-dialog

  5. Close the dialog, refresh the page, and the edge node will appear in the list.

    edge-node-added

    Note

    After an edge node is added, if you cannot see CPU and memory resource usage on the Edge Nodes page, make sure Metrics Server 0.4.1 or later is installed in your cluster.
  6. After an edge node joins your cluster, some Pods may be scheduled to it while they remains in the Pending state on the edge node. Due to the tolerations some DaemonSets (for example, Calico) have, in the current version (KubeSphere v3.1.1), you need to manually patch some Pods so that they will not be schedule to the edge node.

    #!/bin/bash
       
    NodeSelectorPatchJson='{"spec":{"template":{"spec":{"nodeSelector":{"node-role.kubernetes.io/master": "","node-role.kubernetes.io/worker": ""}}}}}'
       
    NoShedulePatchJson='{"spec":{"template":{"spec":{"affinity":{"nodeAffinity":{"requiredDuringSchedulingIgnoredDuringExecution":{"nodeSelectorTerms":[{"matchExpressions":[{"key":"node-role.kubernetes.io/edge","operator":"DoesNotExist"}]}]}}}}}}}'
       
    edgenode="edgenode"
    if [ $1 ]; then
            edgenode="$1"
    fi
       
       
    namespaces=($(kubectl get pods -A -o wide |egrep -i $edgenode | awk '{print $1}' ))
    pods=($(kubectl get pods -A -o wide |egrep -i $edgenode | awk '{print $2}' ))
    length=${#namespaces[@]}
       
       
    for((i=0;i<$length;i++));  
    do
            ns=${namespaces[$i]}
            pod=${pods[$i]}
            resources=$(kubectl -n $ns describe pod $pod | grep "Controlled By" |awk '{print $3}')
            echo "Patching for ns:"${namespaces[$i]}",resources:"$resources
            kubectl -n $ns patch $resources --type merge --patch "$NoShedulePatchJson"
            sleep 1
    done
    

Custom Configurations

To customize some configurations of an edge node, such as download URL and KubeEdge version, create a ConfigMap as below:

apiVersion: v1
data:
  region: zh # Download region.
  version: v1.6.1 # The version of KubeEdge to be installed. Allowed values are v1.5.0, v1.6.0, v1.6.1 (default) and v1.6.2.
kind: ConfigMap
metadata:
  name: edge-watcher-config
  namespace: kubeedge

Note

  • You can specify zh or en for the field region. zh is the default value and the default download link is https://kubeedge.pek3b.qingstor.com/bin/v1.6.1/$arch/keadm-v1.6.1-linux-$arch.tar.gz. If you set region to en, the download link will be https://github.com/kubesphere/kubeedge/releases/download/v1.6.1-kubesphere/keadm-v1.6.1-linux-amd64.tar.gz.
  • The ConfigMap does not affect the configurations of exiting edge nodes in your cluster. It is only used to change the KubeEdge configurations to be used on a new edge node. More specifically, it decides the command automatically created by KubeSphere mentioned above which needs to be executed on the edge node.
  • While you can change the KubeEdge version to be installed on an edge node, it is recommended that the cloud and edge modules have the same KubeEdge version.

Remove an Edge Node

Before you remove an edge node, delete all your workloads running on it.

  1. On your edge node, run the following commands:

    ./keadm reset
    
    apt remove mosquitto
    
    rm -rf /var/lib/kubeedge /var/lib/edged /etc/kubeedge/ca /etc/kubeedge/certs
    

    Note

    If you cannot delete the tmpfs-mounted folder, restart the node or unmount the folder first.
  2. Run the following command to remove the edge node from your cluster:

    kubectl delete node <edgenode-name>
    
  3. To uninstall KubeEdge from your cluster, run the following commands:

    helm uninstall kubeedge -n kubeedge
    
    kubectl delete ns kubeedge
    

    Note

    After the uninstallation, you will not be able to add edge nodes to your cluster.

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.