Air-Gapped Installation

Edit

The air-gapped installation is almost the same as the online installation except it creates a local registry to host the Docker images. We will demonstrate how to install KubeSphere and Kubernetes on air-gapped environment.

Note: The dependencies in different operating systems may cause upexpected problems. If you encounter any installation problems on air-gapped environment, please describe your OS information and error logs on GitHub.

Prerequisites

  • If your machine is behind a firewall, you need to open the ports by following the document Ports Requirements for more information.

    • Installer will use /var/lib/docker as the default directory where all Docker related files, including the images, are stored. We recommend you to add additional storage to a disk with at least 100G mounted at /var/lib/docker and /mnt/registry respectively, use the fdisk command for reference.

  • Installer uses Local volume based on OpenEBS to provide storage service with dynamic provisioning. It is convenient for testing and development. For production, please configure supported persistent storage service and prepare high availability configuration before installation.

  • Since the air-gapped machines cannot connect to apt or yum source, please use clean Linux machine to avoid this problem.

Step 1: Prepare Linux Hosts

The following describes the requirements of hardware and operating system. To get started with multi-node installation, you need to prepare at least three hosts according to the following requirements.

  • Supported OSes: CentOS 7.4 ~ 7.7 (64-bit), Ubuntu 16.04.5/16.04.6/18.04.1/18.04.2/18.04.3 LTS (64-bit)
  • Time synchronization is required across all nodes, otherwise the installation may not succeed;
  • For Ubuntu 16.04 OS, it is recommended to select 16.04.5;
  • If you are using Ubuntu 18.04, you need to use the user root.
  • Ensure your disk of each node is at least 100G.
  • CPU and memory in total of all machines: 2 cores and 4 GB for minimal installation; 8 cores and 16 GB for complete installation.

The following section describes an example to introduce multi-node installation. This example shows three hosts installation by taking the master serving as the taskbox to execute the installation. The following cluster consists of one Master and two Nodes.

Note: KubeSphere supports the high-availability configuration of the Masters and Etcd nodes. Please refer to Creating High Availability KubeSphere Cluster for guide.

Host IPHost NameRole
192.168.0.1mastermaster, etcd
192.168.0.2node1node
192.168.0.3node2node

Cluster Architecture

Single Master, Single Etcd, Two Nodes

Architecture

Step 2: Download Installer Package

Download KubeSphere 2.1.1 to your taskbox machine, then unpack it and go to the folder conf.

  1. curl -L https://kubesphere.io/download/offline/latest > kubesphere-all-offline-v2.1.1.tar.gz \
  2. && tar -zxf kubesphere-all-offline-v2.1.1.tar.gz && cd kubesphere-all-offline-v2.1.1/conf

Step 3: Configure Host Template

This step is only for multi-node installation, you can skip this step if you choose all-in-one installation.

Please refer to the following sample to configure all hosts in hosts.ini. It is recommended to install KubeSphere using root user. The following is an example configuration for CentOS 7.5 using root user. Note do not manually wrap any line in the file.

Note:

  • If you use non-root user with sudo access to install KubeSphere, you need to refer to the example block that is commented out in conf/hosts.ini.
  • If the root user of that taskbox machine cannot establish SSH connection with the rest of machines, you need to refer to the non-root user example at the top of the conf/hosts.ini, but it is recommended to switch root user when executing install.sh.
  • master, node1 and node2 are the host names of each node and all host names should be in lowercase.

hosts.ini

  1. [all]
  2. master ansible_connection=local  ip=192.168.0.1
  3. node1  ansible_host=192.168.0.2  ip=192.168.0.2  ansible_ssh_pass=PASSWORD
  4. node2  ansible_host=192.168.0.3  ip=192.168.0.3  ansible_ssh_pass=PASSWORD
  6. [local-registry]
  7. master
  9. [kube-master]
  10. master
  12. [kube-node]
  13. node1
  14. node2
  16. [etcd]
  17. master
  19. [k8s-cluster:children]
  20. kube-node
  21. kube-master

Note:

  • You need to replace each node information such as IP, password with real values in the group [all]. The master node is the taskbox so you do not need to add password field here.
  • Installer will use a node as the local registry for docker images, defaults to “master” in the group [local-registry].
  • The “master” node also takes the role of master and etcd, so “master” is filled under the group[kube-master] and the group [etcd] respectively.
  • “node1” and “node2” both serve the role of Node, so they are filled under the group [kube-node].

Parameters Specification:

  • ansible_connection: Connection type to the host, “local” in the example above means local connection.
  • ansible_host: The name of the host to be connected.
  • ip: The ip of the host to be connected.
  • ansible_user: The default ssh user name to use.
  • ansible_become_pass: Allows you to set the privilege escalation password.
  • ansible_ssh_pass: The password of the host to be connected using root.

Step 4: Enable All Components

This is step is complete installation. You can skip this step if you choose a minimal installation.

Edit conf/common.yaml, reference the following changes with values being true which are false by default.

  1. # LOGGING CONFIGURATION
  2. # logging is an optional component when installing KubeSphere, and
  3. # Kubernetes builtin logging APIs will be used if logging_enabled is set to false.
  4. # Builtin logging only provides limited functions, so recommend to enable logging.
  5. logging_enabled: true # Whether to install logging system
  6. elasticsearch_master_replica: 1  # total number of master nodes, it's not allowed to use even number
  7. elasticsearch_data_replica: 2  # total number of data nodes
  8. elasticsearch_volume_size: 20Gi # Elasticsearch volume size
  9. log_max_age: 7 # Log retention time in built-in Elasticsearch, it is 7 days by default.
  10. elk_prefix: logstash # the string making up index names. The index name will be formatted as ks-<elk_prefix>-log
  11. kibana_enabled: false # Kibana Whether to install built-in Grafana
  12. #external_es_url: SHOULD_BE_REPLACED # External Elasticsearch address, KubeSphere supports integrate with Elasticsearch outside the cluster, which can reduce the resource consumption.
  13. #external_es_port: SHOULD_BE_REPLACED # External Elasticsearch service port
  15. #DevOps Configuration
  16. devops_enabled: true # Whether to install built-in DevOps system (Supports CI/CD pipeline, Source/Binary to image)
  17. jenkins_memory_lim: 8Gi # Jenkins memory limit, it is 8 Gi by default
  18. jenkins_memory_req: 4Gi # Jenkins memory request, it is 4 Gi by default
  19. jenkins_volume_size: 8Gi # Jenkins volume size, it is 8 Gi by default
  20. jenkinsJavaOpts_Xms: 3g # Following three are JVM parameters
  21. jenkinsJavaOpts_Xmx: 6g
  22. jenkinsJavaOpts_MaxRAM: 8g
  23. sonarqube_enabled: true # Whether to install built-in SonarQube
  24. #sonar_server_url: SHOULD_BE_REPLACED # External SonarQube address, KubeSphere supports integrate with SonarQube outside the cluster, which can reduce the resource consumption.
  25. #sonar_server_token: SHOULD_BE_REPLACED  # SonarQube token
  27. # Following components are all optional for KubeSphere,
  28. # Which could be turned on to install it before installation or later by updating its value to true
  29. openpitrix_enabled: true       # KubeSphere application store
  30. metrics_server_enabled: true   # For KubeSphere HPA to use
  31. servicemesh_enabled: true      # KubeSphere service mesh system(Istio-based)
  32. notification_enabled: true     # KubeSphere notification system
  33. alerting_enabled: true         # KubeSphere alerting system

Step 5: Install KubeSphere to Linux Machines

Note:

  • Generally, you can install KubeSphere without any modification, it will start with minimal installation by default.
  • If you want to enable pluggable feature components installation, modify common.yaml and refer to Enable Pluggable Components Installation for instructions.
  • Installer uses Local volume based on openEBS to provide storage service with dynamic provisioning. For production environment, please configure supported persistent storage service before installation.
  • Since the default subnet for Cluster IPs is 10.233.0.0/18, and the default subnet for Pod IPs is 10.233.64.0/18, the node IPs must not use the two IP range. You can modify the default subnets kube_service_addresses or kube_pods_subnet in the file conf/common.yaml to avoid conflicts.

1. Enter scripts folder, and execute install.sh using root user:

  1. cd ../cripts
  2. ./install.sh

2. Type 2 to select multi-node mode to start the installation. The installer will ask you if you have set up persistent storage service or not. Just type yes since we are going to use local volume.

  1. ################################################
  2.          KubeSphere Installer Menu
  3. ################################################
  4. *   1) All-in-one
  5. *   2) Multi-node
  6. *   3) Quit
  7. ################################################
  8. https://kubesphere.io/               2020-02-24
  9. ################################################
  10. Please input an option: 2

3. Verify the multi-node installation：

(1). If “Successful” it returned after install.sh process completed, then congratulation! you are ready to go.

  1. successsful!
  2. #####################################################
  3. ###              Welcome to KubeSphere!           ###
  4. #####################################################
  6. Console: http://192.168.0.1:30880
  7. Account: admin
  8. Password: P@88w0rd
  10. NOTE：Please modify the default password after login.
  11. #####################################################

Note: The information above is saved in a log file that you can view by following the guide.

(2). You will be able to use default account and password admin / P@88w0rd to log in the console http://{$IP}:30880 to take a tour of KubeSphere. Please change the default password after logging in.

Login

Note: After log in console, please verify the monitoring status of service components in the “Cluster Status”. If any service is not ready, please wait patiently untill all components get running up.

Landing Page

Enable Pluggable Components

If you already have set up minimal installation, you still can edit the ConfigMap of ks-installer using the following command. Please make sure there is enough resource in your machines, see Pluggable Components Overview.

  1. kubectl edit cm -n kubesphere-system ks-installer

FAQ

If you have further questions please do not hesitate to raise issues on GitHub.