wait for control plane node to become ready after joining (#12794)

When joining a control plane node and "upgrading" the cluster setup (for
example, to update etcd addresses after adding a new etcd) in the same
playbook run, the node can take a bit of time to become ready after
joining.
This triggers a kubeadm preflight check (ControlPlaneNodesReady) in
kubeadm upgrade, which is run directly after the join tasks.

Add a configurable wait for the control plane node to become Ready to
fix this race condition.

roles/kubernetes/control-plane/defaults/main/main.yml

@@ -2,6 +2,9 @@
 # disable upgrade cluster
 upgrade_cluster_setup: false
 
+# Number of retries (with 5 seconds interval) to check that new control plane nodes
+# are in Ready condition after joining
+control_plane_node_become_ready_tries: 24
 # By default the external API listens on all interfaces, this can be changed to
 # listen on a specific address/interface.
 # NOTE: If you specific address/interface and use loadbalancer_apiserver_localhost

roles/kubernetes/control-plane/tasks/kubeadm-secondary.yml

@@ -98,3 +98,18 @@
   when:
     - inventory_hostname != first_kube_control_plane
     - kubeadm_already_run is not defined or not kubeadm_already_run.stat.exists
+
+- name: Wait for new control plane nodes to be Ready
+  when: kubeadm_already_run.stat.exists
+  run_once: true
+  command: >
+    {{ kubectl }} get nodes --selector node-role.kubernetes.io/control-plane
+    -o jsonpath-as-json="{.items[*].status.conditions[?(@.type == 'Ready')]}"
+  register: control_plane_node_ready_conditions
+  retries: "{{ control_plane_node_become_ready_tries }}"
+  delay: 5
+  delegate_to: "{{ groups['kube_control_plane'][0] }}"
+  until: >
+    control_plane_node_ready_conditions.stdout
+    | from_json | selectattr('status', '==', 'True')
+    | length == (groups['kube_control_plane'] | length)