Updated on 2024-01-04 GMT+08:00

Connecting to a Cluster Using kubectl

Scenario

This section uses a CCE standard cluster as an example to describe how to connect to a CCE cluster using kubectl.

Permissions

When you access a cluster using kubectl, CCE uses kubeconfig.json generated on the cluster for authentication. This file contains user information, based on which CCE determines which Kubernetes resources can be accessed by kubectl. The permissions recorded in a kubeconfig.json file vary from user to user.

For details about user permissions, see Cluster Permissions (IAM-based) and Namespace Permissions (Kubernetes RBAC-based).

Using kubectl

To connect to a Kubernetes cluster from a PC, you can use kubectl, a Kubernetes command line tool. You can log in to the CCE console and click the name of the target cluster to access the cluster console. On the Overview page, view the access address and kubectl connection procedure.

CCE allows you to access a cluster through a private network or a public network.
  • Intranet access: The client that accesses the cluster must be in the same VPC as the cluster.
  • Public access: The client that accesses the cluster must be able to access public networks and the cluster has been bound with a public network IP.

    To bind an EIP to the cluster, go to the Overview page and click Bind next to EIP in the Connection Information area. In a cluster with an EIP bound, kube-apiserver will be exposed to the Internet and may be attacked. To solve this problem, you can configure Advanced Anti-DDoS for the EIP of the node on which kube-apiserver runs.

Download kubectl and the configuration file. Copy the file to your client, and configure kubectl. After the configuration is complete, you can access your Kubernetes clusters. Procedure:

  1. Download kubectl.

    Prepare a computer that can access the public network and install kubectl in CLI mode. You can run the kubectl version command to check whether kubectl has been installed. If kubectl has been installed, skip this step.

    This section uses the Linux environment as an example to describe how to install and configure kubectl. For details, see Installing kubectl.

    1. Log in to your client and download kubectl.
      cd /home
      curl -LO https://dl.k8s.io/release/{v1.25.0}/bin/linux/amd64/kubectl

      {v1.25.0} specifies the version number. Replace it as required.

    2. Install kubectl.
      chmod +x kubectl
      mv -f kubectl /usr/local/bin

  2. Obtain the kubectl configuration file (kubeconfig).

    On the Overview page, locate the Connection Info area, click Configure next to kubectl. On the window displayed, download the configuration file.

    • The kubectl configuration file kubeconfig.json is used for cluster authentication. If the file is leaked, your clusters may be attacked.
    • The Kubernetes permissions assigned by the configuration file downloaded by IAM users are the same as those assigned to the IAM users on the CCE console.
    • If the KUBECONFIG environment variable is configured in the Linux OS, kubectl preferentially loads the KUBECONFIG environment variable instead of $home/.kube/config.

  3. Configure kubectl.

    Configure kubectl (A Linux OS is used).
    1. Log in to your client and copy the kubeconfig.json configuration file downloaded in 2 to the /home directory on your client.
    2. Configure the kubectl authentication file.
      cd /home
      mkdir -p $HOME/.kube
      mv -f kubeconfig.json $HOME/.kube/config
    3. Switch the kubectl access mode based on service scenarios.
      • Run this command to enable intra-VPC access:
        kubectl config use-context internal
      • Run this command to enable public access (EIP required):
        kubectl config use-context external
      • Run this command to enable public access and two-way authentication (EIP required):
        kubectl config use-context externalTLSVerify

        For details about the cluster two-way authentication, see Two-Way Authentication for Domain Names.

Two-Way Authentication for Domain Names

CCE supports two-way authentication for domain names.

  • After an EIP is bound to an API Server, two-way domain name authentication will be disabled by default if kubectl is used to connect to the cluster. You can run kubectl config use-context externalTLSVerify to switch to the externalTLSVerify context to enable the two-way domain name authentication.
  • When an EIP is bound to or unbound from a cluster, or a custom domain name is configured or updated, the cluster server certificate will be added the latest cluster access address (including the EIP bound to the cluster and all custom domain names configured for the cluster).
  • Asynchronous cluster synchronization takes about 5 to 10 minutes. You can view the synchronization result in Synchronize Certificate in Operation Records.
  • For a cluster that has been bound to an EIP, if the authentication fails (x509: certificate is valid) when two-way authentication is used, bind the EIP again and download kubeconfig.json again.
  • If the domain name two-way authentication is not supported, kubeconfig.json contains the "insecure-skip-tls-verify": true field, as shown in Figure 1. To use two-way authentication, you can download the kubeconfig.json file again and enable two-way authentication for the domain names.
    Figure 1 Two-way authentication disabled for domain names

FAQs

  • Error from server Forbidden

    When you use kubectl to create or query Kubernetes resources, the following output is returned:

    # kubectl get deploy Error from server (Forbidden): deployments.apps is forbidden: User "0c97ac3cb280f4d91fa7c0096739e1f8" cannot list resource "deployments" in API group "apps" in the namespace "default"

    The cause is that the user does not have the permissions to operate the Kubernetes resources. For details about how to assign permissions, see Namespace Permissions (Kubernetes RBAC-based).

  • The connection to the server localhost:8080 was refused

    When you use kubectl to create or query Kubernetes resources, the following output is returned:

    The connection to the server localhost:8080 was refused - did you specify the right host or port?

    The cause is that cluster authentication is not configured for the kubectl client. For details, see 3.