fix: speed up ansible bootstrap

ansible/ansible.cfg

@@ -6,6 +6,11 @@ roles_path = roles
 stdout_callback = default
 result_format = yaml
 interpreter_python = auto_silent
+forks = 20
+
+[ssh_connection]
+pipelining = True
+ssh_args = -o ControlMaster=auto -o ControlPersist=60s -o ServerAliveInterval=30 -o ServerAliveCountMax=6
 
 [privilege_escalation]
 become = True

ansible/site.yml

@@ -16,11 +16,14 @@
   pre_tasks:
     - name: Wait for SSH
       wait_for_connection:
-        delay: 10
+        delay: 0
         timeout: 600
 
     - name: Gather facts after SSH is reachable
       setup:
+        gather_subset:
+          - "!all"
+          - network
       register: initial_setup
       ignore_errors: true
       ignore_unreachable: true
@@ -31,12 +34,15 @@
 
     - name: Wait for SSH after transient first-boot disconnect
       wait_for_connection:
-        delay: 10
+        delay: 5
         timeout: 300
       when: initial_setup.unreachable | default(false) or initial_setup.failed | default(false)
 
     - name: Gather facts after transient first-boot disconnect
       setup:
+        gather_subset:
+          - "!all"
+          - network
       when: initial_setup.unreachable | default(false) or initial_setup.failed | default(false)
 
   roles:
@@ -173,17 +179,9 @@
       when: kube_vip_rollout.rc != 0
 
     - name: Wait for control plane node readiness
-      shell: |
-        set -euo pipefail
-        kubectl get node/{{ item }} -o jsonpath='{.status.conditions[?(@.type=="Ready")].status}' | grep -qx True
-      args:
-        executable: /bin/bash
+      command: kubectl wait --for=condition=Ready node -l node-role.kubernetes.io/control-plane --timeout=900s
       register: control_plane_ready
-      until: control_plane_ready.rc == 0
-      retries: 90
-      delay: 10
       changed_when: false
-      loop: "{{ groups['control_plane'] }}"
 
     - name: Wait for Kubernetes readyz before worker joins
       command: kubectl --server=https://{{ kube_api_endpoint }}:6443 get --raw=/readyz