Baidu AI Cloud
中国站

百度智能云

Cloud Container Engine

Scale Out Kubernetes Cluster at the Edges

Scale Out CCE Kubernetes Cluster at the Edges

Prerequisites

  • The account has enabled the BEC whitelist;
  • The account has enabled the BEC high-performance network whitelist;
  • Prepare the edge cloud server used to deploy the cluster Node through the BEC console. The instance configuration requirements are as follows:

    • The instances must be in the same region and the same operator as the Master instance;
    • The instances must be bound to the public IP;
    • The same instance cannot be used as the cluster master and node at the same time;
    • The operating system uses Centos 7.6;
    • For the instance, it is recommended to create a data disk, format, and mount it in advance;
    • If the instance has a GPU, you need to install the GPU driver yourself in advance;

Cluster Scale-out

  1. Find the cluster to be scaled out from the CCE cluster list page and click the button “Add node” on the right.

image.png

  1. Select the lower right corner of the custom parameters to scale out the cluster.

image.png

  1. Add the node information you need to add.

image.png

A complete example of scaling out 2 nodes:

[
    {
        "instanceSpec":{
            "machineType": "BEC",
            "existed":true,
            "existedOption":{
                "existedInstanceID":"vm-xxxxxxxxxxxxxxxxxxxxx"
            },
            "deployCustomConfig": {
                "kubeletRootDir": "/home/cce/kubelet",
                "dockerConfig": {
                    "dockerDataRoot": "/home/cce/docker"
                }
            },
            "becOption": {
                "becRegion": "CENTRAL_CHINA",
                "becCity": "ZHUZHOU",
                "becServiceProvider": "CHINA_UNICOM"
            },
            "adminPassword":"",
            "needGPU": false
        }
    },
    {
        "instanceSpec":{
            "machineType": "BEC",
            "existed":true,
            "existedOption":{
                "existedInstanceID":"vm-xxxxxxxxxxxxxxxxxxxxx"
            },
            "deployCustomConfig": {
                "kubeletRootDir": "/home/cce/kubelet",
                "dockerConfig": {
                    "dockerDataRoot": "/home/cce/docker"
                }
            },
            "becOption": {
                "becRegion": "CENTRAL_CHINA",
                "becCity": "ZHUZHOU",
                "becServiceProvider": "CHINA_UNICOM"
            },
            "adminPassword":"",
            "needGPU": false
        }
    }
]

Description of parameter configuration

Node Configuration Items

Parameter Name Description Example
instanceSpec.machineType Host type BEC
instanceSpec.existed Created or not. true
instanceSpec.existedOption.existedInstanceID The BEC instance ID vm-xxxxxxxxxxxxxxxxxxxxx
instanceSpec.deployCustomConfig.kubeletRootDir The kubelet data directory. If the data disk is mounted, recommend you select the data disk mounting directory. /var/lib/kubelet
instanceSpec.deployCustomConfig.dockerConfig.dockerDataRoot The docker data directory. If the data disk is mounted, recommend you select the data disk mounting directory. /var/lib/docker
instanceSpec.becOption.becRegion Region where the BEC virtual machine resides. CENTRAL_CHINA
instanceSpec.becOption.becCity City where the BEC virtual machine resides. ZHUZHOU
instanceSpec.becOption.becServiceProvider The service provider of the BEC virtual machine. CHINA_UNICOM
instanceSpec.adminPassword The password of the root account of the BEC instance. In the deployment phase, the cloud control plane needs to log in to the virtual machine through the root account for deployment. It can be changed after the deployment is complete.
instanceSpec.needGPU Whether to install a GPU-related container dependency package at the node. You do not need to install a GPU-related container dependency package at a non-GPU node. false
  1. Click Scale-out to start cluster scale-out after entering the configuration information.
Previous
Create Kubernetes Cluster at the Edges
Next
Kubernetes Nodes at the Edges