Skip to content

Latest commit

 

History

History
112 lines (76 loc) · 2.98 KB

hooks_and_plugins.md

File metadata and controls

112 lines (76 loc) · 2.98 KB
Raw

A plugin allow custom code to be run at key point (a hook) of the ansible_kubedm playbooks. It comes with great possibilities, hence use with caution is required.

Plugins

Plugins are discovered in kubeadm_plugins_dir which default to {{ inventory_dir }}/kubeadm.plugins.d.

Simply create a directory with the name of the plugin (whatever you like exept hidden one).

Inside this directory create a directory per hook name.

Finally put yaml with tasks list inside hook's directories.

Note: Those are tasks like in roles not playbooks.

Here is a sample layout, using the default settings for kubeadm_plugins_dir:

hosts.cfg # the inventory
group_vars/
host_vars/
kubeadm.plugins.d/
    upgrade_os/
        post_nodes_upgrade/          # Will be run on control plane and workers node when upgrade is launched
            subdir/
                clean_old_kernel.yml # will not be loaded, but could included
            upgrade_os.yml           # will be loaded
    .upgrade_cri/
        post_nodes_upgrade/
            cri_update.yml           # will not be loaded, because plugin name is an hidden directory

All the available hooks are listed bellow.

Hooks

The following sections present each playbook workflow, hooks are marked with []

Some hooks are present multiple times, so be careful choosing the hook name, when writing plugin.

00-apiserver-proxy.yml

  • find IP of control plane
  • [pre_apiserver_proxy]
  • deploy apiserver proxy
  • [post_apiserver_proxy]

In case of running an upgrade to haproxy proxy loadbalancer (first on each control plane then all workers):

  • [pre_proxy_upgrade_haproxy]
  • Upgrade mechanism to haproxy
  • [post_proxy_upgrade_haproxy]

01_site.yml

The main body execute:

  • [pre_run]
  • Do some checks on control plane nodes
  • [post_preflight_cp]
  • Do some checks on worker nodes
  • [post_preflight_nodes]
  • Do some initialization tasks on control plane nodes
  • [post_first_tasks_cp]

In case no cluster is found, will init one on a single control plane nodes:

  • [pre_init]
  • Init cluster
  • [post_init]

Then continue some indempotent tasks on control plane nodes:

  • [pre_config_update]
  • Create bootstrap token if required
  • Update kubeadm config
  • [post_config_update]

In case an upgrade of kubernetes is required, run it now:

  • [pre_kube_upgrade]
  • Upgrade kubernetes
  • [post_kube_upgrade]

Apply upgrade on control plane nodes (node by node by default):

  • [pre_cp_upgrade, pre_nodes_upgrade]
  • Apply upgrade on control plane nodes
  • [post_cp_upgrades, post_nodes_upgrade]

Apply updade on worker nodes (node-by-node by default):

  • [pre_workers_upgrade, pre_nodes_upgrade]
  • Apply upgrade on worker nodes
  • [post_workers_upgrade, post_nodes_upgrade]

Then join missing control plane nodes

  • [pre_cp_join, pre_nodes_join]
  • Join control plane nodes
  • [post_cp_join, post_nodes_join]

Finally join missing workers nodes:

  • [pre_workers_join, pre_nodes_join]
  • Join worker nodes
  • [post_workers_join, post_nodes_join]

Finally executing the last hook:

  • [post_run]