refactor(ansible): use atomic direct binary replacement for aerodrome control plane upgrade

ansible/playbooks/upgrade_k3s_aerodrome.yml

@@ -16,32 +16,35 @@
       ansible.builtin.debug:
         msg: "Upgrading {{ inventory_hostname }} from {{ pre_k3s_ver.stdout_lines[0] }} to {{ target_k3s_version }}"
 
-    - name: Download and execute K3s binary upgrade
+    - name: Download target K3s release binary and replace executable
       ansible.builtin.shell: |
-        curl -sfL https://get.k3s.io | INSTALL_K3S_VERSION={{ target_k3s_version }} sh -
-      register: k3s_upgrade_cmd
+        set -e
+        ENCODED_VER=$(echo "{{ target_k3s_version }}" | sed 's/+/%2B/g')
+        curl -sfL "https://github.com/k3s-io/k3s/releases/download/${ENCODED_VER}/k3s-arm64" -o /usr/local/bin/k3s.new
+        chmod +x /usr/local/bin/k3s.new
+        mv /usr/local/bin/k3s.new /usr/local/bin/k3s
+      register: k3s_download
 
-    - name: Ensure K3s systemd unit is active
+    - name: Restart K3s systemd service
       ansible.builtin.systemd:
         name: k3s
-        state: started
-        enabled: true
+        state: restarted
 
-    - name: Wait for local K3s service to be active
+    - name: Wait for K3s systemd unit to report active
       ansible.builtin.command: systemctl is-active k3s
       register: k3s_active
       until: k3s_active.stdout == "active"
-      retries: 15
-      delay: 5
+      retries: 20
+      delay: 4
       changed_when: false
 
-    - name: Wait for node to report Ready in cluster
+    - name: Wait for upgraded node to report Ready in Kubernetes
       ansible.builtin.command: kubectl get node {{ inventory_hostname }} -o jsonpath='{.status.conditions[?(@.type=="Ready")].status}'
       delegate_to: sninfr2lnxctl01
       register: node_ready
       until: node_ready.stdout == "True"
-      retries: 20
-      delay: 6
+      retries: 30
+      delay: 5
       changed_when: false
 
     - name: Query post-upgrade K3s version