Ansible Role - Kubernetes
Go to file
2022-10-26 17:26:42 -05:00
.github Try Kubernetes 1.24.4. 2022-10-26 09:52:36 -05:00
defaults Back to 1.25. 2022-10-26 09:52:57 -05:00
handlers Initial commit. 2018-05-09 09:19:11 -05:00
meta Update Kubernetes version to 1.25, see what happens. 2022-10-26 09:50:47 -05:00
molecule/default Attempt to get tests working with debug. 2022-10-26 12:38:20 -05:00
tasks Remove TODOs in code. 2022-10-26 17:26:42 -05:00
templates Fix kubelet and kubeadm configs. 2022-10-26 09:50:50 -05:00
vars PR #75 follow-up: Make sure proper package is present and only run on older Debian and other OSes. 2020-09-01 12:22:54 -05:00
.ansible-lint Make ansible-lint work again. 2021-02-22 16:30:07 -06:00
.gitignore Make ansible-lint work again. 2021-02-22 16:30:07 -06:00
.yamllint Fix linting issues. 2020-08-20 12:23:45 -05:00
LICENSE Initial commit. 2018-05-09 09:19:11 -05:00
README.md Attempt to get tests working with debug. 2022-10-26 12:38:20 -05:00

Ansible Role: Kubernetes

CI

An Ansible Role that installs Kubernetes on Linux.

Requirements

Requires a compatible Container Runtime; recommended role for CRI installation: geerlingguy.containerd.

Role Variables

Available variables are listed below, along with default values (see defaults/main.yml):

kubernetes_packages:
  - name: kubelet
    state: present
  - name: kubectl
    state: present
  - name: kubeadm
    state: present
  - name: kubernetes-cni
    state: present

Kubernetes packages to be installed on the server. You can either provide a list of package names, or set name and state to have more control over whether the package is present, absent, latest, etc.

kubernetes_version: '1.25'
kubernetes_version_rhel_package: '1.25.1'

The minor version of Kubernetes to install. The plain kubernetes_version is used to pin an apt package version on Debian, and as the Kubernetes version passed into the kubeadm init command (see kubernetes_version_kubeadm). The kubernetes_version_rhel_package variable must be a specific Kubernetes release, and is used to pin the version on Red Hat / CentOS servers.

kubernetes_role: control_plane

Whether the particular server will serve as a Kubernetes control_plane (default) or node. The control plane will have kubeadm init run on it to intialize the entire K8s control plane, while nodes will have kubeadm join run on them to join them to the control_plane.

With this role, kubeadm init will be run with --config <FILE>.

kubernetes_kubeadm_kubelet_config_file_path: '/etc/kubernetes/kubeadm-kubelet-config.yaml'

Path for <FILE>. If the directory does not exist, this role will create it.

The following variables are parsed as options to . To understand its syntax, see kubelet-integration and kubeadm-config-file . The skeleton (apiVersion, kind) of the config file will be created by this role, so do not define them within the variables. (See templates/kubeadm-kubelet-config.j2).

kubernetes_config_init_configuration:
  localAPIEndpoint:
    advertiseAddress: "{{ kubernetes_apiserver_advertise_address | default(ansible_default_ipv4.address, true) }}"

Defines the options under kind: InitConfiguration. Including kubernetes_apiserver_advertise_address here is for backward-compatibilty to older versions of this role, where kubernetes_apiserver_advertise_address was used with a command-line-option.

kubernetes_config_cluster_configuration:
  networking:
    podSubnet: "{{ kubernetes_pod_network.cidr }}"
  kubernetesVersion: "{{ kubernetes_version_kubeadm }}"

Options under kind: ClusterConfiguration. Including kubernetes_pod_network.cidr and kubernetes_version_kubeadm here are for backward-compatibilty to older versions of this role, where they were used with command-line-options.

kubernetes_config_kubelet_configuration:
  cgroupDriver: systemd

Options to configure kubelet on any nodes in your cluster through the kubeadm init process. For syntax options read the kubelet config file and kubelet integration documentation.

NOTE: This is the recommended way to do the kubelet-configuration. Most command-line-options are deprecated.

NOTE: The recommended cgroupDriver depends on your Container Runtime. When using this role with Docker instead of containerd, this value should be changed to cgroupfs.

kubernetes_config_kube_proxy_configuration: {}

Options to configure kubelet's proxy configuration in the KubeProxyConfiguration section of the kubelet configuration.

Variables to configure kubeadm and kubelet through command-line-options

kubernetes_kubelet_extra_args: ""
kubernetes_kubelet_extra_args_config_file: /etc/default/kubelet

Extra args to pass to kubelet during startup. E.g. to allow kubelet to start up even if there is swap is enabled on your server, set this to: "--fail-swap-on=false". Or to specify the node-ip advertised by kubelet, set this to "--node-ip={{ ansible_host }}". This option is deprecated. Please use kubernetes_config_kubelet_configuration instead.

kubernetes_kubeadm_init_extra_opts: ""

Extra args to pass to kubeadm init during K8s control plane initialization. E.g. to specify extra Subject Alternative Names for API server certificate, set this to: "--apiserver-cert-extra-sans my-custom.host"

kubernetes_join_command_extra_opts: ""

Extra args to pass to the generated kubeadm join command during K8s node initialization. E.g. to ignore certain preflight errors like swap being enabled, set this to: --ignore-preflight-errors=Swap

Additional variables

kubernetes_allow_pods_on_control_plane: true

Whether to remove the taint that denies pods from being deployed to the Kubernetes control plane. If you have a single-node cluster, this should definitely be True. Otherwise, set to False if you want a dedicated Kubernetes control plane which doesn't run any other pods.

kubernetes_pod_network:
  # Flannel CNI.
  cni: 'flannel'
  cidr: '10.244.0.0/16'
  #
  # Calico CNI.
  # cni: 'calico'
  # cidr: '192.168.0.0/16'
  #
  # Weave CNI.
  # cni: 'weave'
  # cidr: '192.168.0.0/16'

This role currently supports flannel (default), calico or weave for cluster pod networking. Choose only one for your cluster; converting between them is not done automatically and could result in broken networking; if you need to switch from one to another, it should be done outside of this role.

kubernetes_apiserver_advertise_address: ''`
kubernetes_version_kubeadm: 'stable-{{ kubernetes_version }}'`
kubernetes_ignore_preflight_errors: 'all'

Options passed to kubeadm init when initializing the Kubernetes control plane. The kubernetes_apiserver_advertise_address defaults to ansible_default_ipv4.address if it's left empty.

kubernetes_apt_release_channel: main
kubernetes_apt_repository: "deb http://apt.kubernetes.io/ kubernetes-xenial {{ kubernetes_apt_release_channel }}"
kubernetes_apt_ignore_key_error: false

Apt repository options for Kubernetes installation.

kubernetes_yum_arch: x86_64
kubernetes_yum_base_url: "https://packages.cloud.google.com/yum/repos/kubernetes-el7-{{ kubernetes_yum_arch }}"
kubernetes_yum_gpg_key:
  - https://packages.cloud.google.com/yum/doc/yum-key.gpg
  - https://packages.cloud.google.com/yum/doc/rpm-package-key.gpg
kubernetes_yum_gpg_check: true
kubernetes_yum_repo_gpg_check: true

Yum repository options for Kubernetes installation. You can change kubernete_yum_gpg_key to a different url if you are behind a firewall or provide a trustworthy mirror. Usually in combination with changing kubernetes_yum_base_url as well.

kubernetes_flannel_manifest_file: https://raw.githubusercontent.com/coreos/flannel/master/Documentation/kube-flannel.yml

Flannel manifest file to apply to the Kubernetes cluster to enable networking. You can copy your own files to your server and apply them instead, if you need to customize the Flannel networking configuration.

kubernetes_calico_manifest_file: https://projectcalico.docs.tigera.io/manifests/calico.yaml

Calico manifest file to apply to the Kubernetes cluster (if using Calico instead of Flannel).

Dependencies

None.

Example Playbooks

Single node (control-plane-only) cluster

- hosts: all

  vars:
    kubernetes_allow_pods_on_control_plane: true

  roles:
    - geerlingguy.docker
    - geerlingguy.kubernetes

Two or more nodes (single control-plane) cluster

Control plane inventory vars:

kubernetes_role: "control_plane"

Node(s) inventory vars:

kubernetes_role: "node"

Playbook:

- hosts: all

  vars:
    kubernetes_allow_pods_on_control_plane: true

  roles:
    - geerlingguy.docker
    - geerlingguy.kubernetes

Then, log into the Kubernetes control plane, and run kubectl get nodes as root, and you should see a list of all the servers.

License

MIT / BSD

Author Information

This role was created in 2018 by Jeff Geerling, author of Ansible for DevOps.