fix: harden cluster rebuild determinism

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

@@ -62,6 +62,10 @@
     url: https://get.k3s.io
     dest: /tmp/install-k3s.sh
     mode: "0755"
+  register: k3s_install_script
+  until: k3s_install_script is succeeded
+  retries: 5
+  delay: 10
   when: k3s_install_needed
 
 - name: Install k3s server (primary)
@@ -82,7 +86,9 @@
     {% if k3s_disable_servicelb | bool %}--disable=servicelb{% endif %}
     {% if k3s_kubelet_cloud_provider_external | bool %}--kubelet-arg=cloud-provider=external{% endif %}
   register: primary_install
-  failed_when: false
+  until: primary_install.rc == 0
+  retries: 3
+  delay: 20
   when:
     - k3s_install_needed
     - k3s_primary | default(false)
@@ -106,7 +112,9 @@
         {% if k3s_disable_servicelb | bool %}--disable=servicelb{% endif %}
         {% if k3s_kubelet_cloud_provider_external | bool %}--kubelet-arg=cloud-provider=external{% endif %}
       register: secondary_install
-      failed_when: false
+      until: secondary_install.rc == 0
+      retries: 3
+      delay: 20
 
 - name: Wait for k3s to be ready
   command: "{{ (k3s_primary | default(false)) | ternary('kubectl get nodes', 'systemctl is-active k3s') }}"