Install a Multi-node Kubernetes and KubeSphere Cluster

In a production environment, a single-node cluster cannot satisfy most of the needs as the cluster has limited resources with insufficient compute capabilities. Thus, single-node clusters are not recommended for large-scale data processing. Besides, a cluster of this kind is not available with high availability as it only has one node. On the other hand, a multi-node architecture is the most common and preferred choice in terms of application deployment and distribution.

This section gives you an overview of a single-master multi-node installation, including the concept, KubeKey and steps. For information about HA installation, refer to High Availability Configurations, Installing on Public Cloud and Installing in On-premises Environment.

Video Demonstration

Concept

A multi-node cluster is composed of at least one control plane and one worker node. You can use any node as the taskbox to carry out the installation task. You can add additional nodes based on your needs (for example, for high availability) both before and after the installation.

• Control plane node. The control plane generally hosts the control plane and controls and manages the whole system.

• Worker node. Worker nodes run the actual applications deployed on them.

Step 1: Prepare Linux Hosts

Please see the requirements for hardware and operating system shown below. To get started with multi-node installation in this tutorial, you need to prepare at least three hosts according to the following requirements. It is possible to install the KubeSphere Container Platform on two nodes if they have sufficient resources.

System requirements

Node requirements

• All nodes must be accessible through SSH.
• Time synchronization for all nodes.
• sudo/curl/openssl/tar should be used in all nodes.

Container runtimes

Your cluster must have an available container runtime. If you use KubeKey to set up a cluster, KubeKey will install the latest version of Docker by default. Alternatively, you can install Docker or other container runtimes by yourself before you create a cluster.

Dependency requirements

KubeKey can install Kubernetes and KubeSphere together. The dependency that needs to be installed may be different based on the Kubernetes version to be installed. You can refer to the list below to see if you need to install relevant dependencies on your node in advance.

Network and DNS requirements

• Make sure the DNS address in /etc/resolv.conf is available. Otherwise, it may cause some issues of DNS in clusters.
• If your network configuration uses firewall rules or security groups, you must ensure infrastructure components can communicate with each other through specific ports. It's recommended that you turn off the firewall or follow the guide Port Requirements.
• Supported CNI plugins: Calico and Flannel. Others (such as Cilium and Kube-OVN) may also work but note that they have not been fully tested.

This example includes three hosts as below with the control plane serving as the taskbox.

Step 2: Download KubeKey

Follow the step below to download KubeKey.

curl -sfL https://get-kk.kubesphere.io | VERSION=v3.0.7 sh -

export KKZONE=cn

curl -sfL https://get-kk.kubesphere.io | VERSION=v3.0.7 sh -

Make kk executable:

chmod +x kk

Step 3: Create a Kubernetes Multi-node Cluster

For multi-node installation, you need to create a cluster by specifying a configuration file.

1. Create an example configuration file

Command:

./kk create config [--with-kubernetes version] [--with-kubesphere version] [(-f | --file) path]

Here are some examples for your reference:

• You can create an example configuration file with default configurations. You can also specify the file with a different filename, or in a different folder.

  ./kk create config [-f ~/myfolder/abc.yaml]
  

• You can specify a KubeSphere version that you want to install (for example, --with-kubesphere v3.3.2).

  ./kk create config --with-kubesphere [version]
  

2. Edit the configuration file of a Kubernetes multi-node cluster

A default file config-sample.yaml will be created if you do not change the name. Edit the file and here is an example of the configuration file of a multi-node cluster with the control plane.

spec:
  hosts:
  - {name: master, address: 192.168.0.2, internalAddress: 192.168.0.2, user: ubuntu, password: Testing123}
  - {name: node1, address: 192.168.0.3, internalAddress: 192.168.0.3, user: ubuntu, password: Testing123}
  - {name: node2, address: 192.168.0.4, internalAddress: 192.168.0.4, user: ubuntu, password: Testing123}
  roleGroups:
    etcd:
    - master
    control-plane:
    - master
    worker:
    - node1
    - node2
  controlPlaneEndpoint:
    domain: lb.kubesphere.local
    address: ""
    port: 6443

Hosts

List all your machines under hosts and add their detailed information as above.

name: The hostname of the instance.

address: The IP address you use for the connection between the taskbox and other instances through SSH. This can be either the public IP address or the private IP address depending on your environment. For example, some cloud platforms provide every instance with a public IP address which you use to access instances through SSH. In this case, you can provide the public IP address for this field.

internalAddress: The private IP address of the instance.

At the same time, you must provide the login information used to connect to each instance. Here are some examples:

• For password login:

  hosts:
    - {name: master, address: 192.168.0.2, internalAddress: 192.168.0.2, port: 8022, user: ubuntu, password: Testing123}
  

  Note

  In this tutorial, port 22 is the default port of SSH so you do not need to add it in the YAML file. Otherwise, you need to add the port number after the IP address as above.

• For the default root user:

  hosts:
    - {name: master, address: 192.168.0.2, internalAddress: 192.168.0.2, password: Testing123}
  

• For passwordless login with SSH keys:

  hosts:
    - {name: master, address: 192.168.0.2, internalAddress: 192.168.0.2, privateKeyPath: "~/.ssh/id_rsa"}
  

roleGroups

• etcd: etcd node names
• control-plane: Control plane node names
• worker: Worker node names

controlPlaneEndpoint (for HA installation only)

The controlPlaneEndpoint is where you provide your external load balancer information for an HA cluster. You need to prepare and configure the external load balancer if and only if you need to install multiple control plane nodes. Please note that the address and port should be indented by two spaces in config-sample.yaml, and address should be your load balancer's IP address. See HA Configurations for details.

addons

You can customize persistent storage plugins (for example, NFS Client, Ceph RBD, and GlusterFS) by specifying storage under the field addons in config-sample.yaml. For more information, see Persistent Storage Configurations.

KubeKey will install OpenEBS to provision LocalPV for development and testing environment by default, which is convenient for new users. In this example of multi-node installation, the default storage class (local volume) is used. For production, you can use Ceph/GlusterFS/CSI or commercial products as persistent storage solutions.

When you finish editing, save the file.

3. Create a cluster using the configuration file

./kk create cluster -f config-sample.yaml

The whole installation process may take 10-20 minutes, depending on your machine and network.

4. Verify the installation

When the installation finishes, you can see the content as follows:

#####################################################
###              Welcome to KubeSphere!           ###
#####################################################

Console: http://192.168.0.2:30880
Account: admin
Password: P@88w0rd

NOTES：
  1. After you log into the console, please check the
     monitoring status of service components in
     the "Cluster Management". If any service is not
     ready, please wait patiently until all components
     are up and running.
  2. Please change the default password after login.

#####################################################
https://kubesphere.io             20xx-xx-xx xx:xx:xx
#####################################################

Now, you will be able to access the web console of KubeSphere at <NodeIP>:30880 with the default account and password (admin/P@88w0rd).

Enable kubectl Autocompletion

KubeKey doesn't enable kubectl autocompletion. See the content below and turn it on:

# Install bash-completion
apt-get install bash-completion

# Source the completion script in your ~/.bashrc file
echo 'source <(kubectl completion bash)' >>~/.bashrc

# Add the completion script to the /etc/bash_completion.d directory
kubectl completion bash >/etc/bash_completion.d/kubectl

Detailed information can be found here.

Code Demonstration

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.