fix: harden cluster rebuild determinism

ansible/roles/k3s-agent/tasks/main.yml

@@ -32,11 +32,22 @@
     url: https://get.k3s.io
     dest: /tmp/install-k3s.sh
     mode: "0755"
+  register: k3s_agent_install_script
+  until: k3s_agent_install_script is succeeded
+  retries: 5
+  delay: 10
   when: k3s_agent_install_needed
 
 - name: Install k3s agent
   when: k3s_agent_install_needed
   block:
+    - name: Wait for Kubernetes API endpoint before agent join
+      wait_for:
+        host: "{{ k3s_server_url | regex_replace('^https?://([^:/]+).*$', '\\1') }}"
+        port: 6443
+        state: started
+        timeout: 180
+
     - name: Run k3s agent install
       environment:
         INSTALL_K3S_VERSION: "{{ k3s_version if k3s_version != 'latest' else '' }}"
@@ -48,7 +59,9 @@
         --flannel-iface={{ k3s_flannel_iface }}
         {% if k3s_kubelet_cloud_provider_external | bool %}--kubelet-arg=cloud-provider=external{% endif %}
       register: k3s_agent_install
-      failed_when: false
+      until: k3s_agent_install.rc == 0
+      retries: 3
+      delay: 20
 
 - name: Wait for k3s agent to be ready
   command: systemctl is-active k3s-agent