fix: reduce rerun bootstrap pre-pull delays

.gitea/workflows/deploy.yml

@@ -195,36 +195,6 @@ jobs:
           PRIMARY_IP=$(terraform output -raw primary_control_plane_ip)
           sed -i "s#https://k8s-cluster-cp-1\.[^:]*:6443#https://${PRIMARY_IP}:6443#g" ../outputs/kubeconfig
 
-      - name: Pre-pull Flux controller images on primary control plane
-        working-directory: terraform
-        run: |
-          set -euo pipefail
-          PRIMARY_IP=$(terraform output -raw primary_control_plane_ip)
-          ssh -o StrictHostKeyChecking=no "ubuntu@${PRIMARY_IP}" 'bash -s' <<'EOF'
-          set -euo pipefail
-          images=(
-            ghcr.io/fluxcd/source-controller:v1.8.0
-            ghcr.io/fluxcd/kustomize-controller:v1.8.1
-            ghcr.io/fluxcd/helm-controller:v1.5.1
-            ghcr.io/fluxcd/notification-controller:v1.8.1
-          )
-
-          for image in "${images[@]}"; do
-            for attempt in $(seq 1 12); do
-              if timeout 180s sudo /usr/local/bin/ctr -n k8s.io images pull "${image}"; then
-                break
-              fi
-
-              if [ "${attempt}" -eq 12 ]; then
-                echo "Failed to pre-pull ${image} after ${attempt} attempts" >&2
-                exit 1
-              fi
-
-              sleep 20
-            done
-          done
-          EOF
-
       - name: Bootstrap Flux source and reconciliation graph
         env:
           KUBECONFIG: outputs/kubeconfig

ansible/roles/kube-vip-deploy/tasks/main.yml

@@ -1,12 +1,37 @@
 ---
 - name: Pre-pull kube-vip bootstrap images into containerd
-  command: /usr/local/bin/ctr -n k8s.io images pull {{ item }}
+  shell: |
+    if /usr/local/bin/ctr -n k8s.io images ls -q | grep -Fx -- "{{ item }}" >/dev/null; then
+      echo "already present"
+      exit 0
+    fi
+
+    for attempt in 1 2 3; do
+      if timeout 120s /usr/local/bin/ctr -n k8s.io images pull "{{ item }}"; then
+        echo "pulled image"
+        exit 0
+      fi
+
+      sleep 10
+    done
+
+    exit 1
+  args:
+    executable: /bin/bash
   register: kube_vip_image_pull
-  retries: 12
-  delay: 15
-  until: kube_vip_image_pull.rc == 0
   loop: "{{ kube_vip_prepull_images }}"
-  changed_when: true
+  changed_when: "'pulled image' in kube_vip_image_pull.stdout"
+  failed_when: false
+
+- name: Report kube-vip images that did not pre-pull after retries
+  debug:
+    msg: >-
+      Best-effort kube-vip image pre-pull did not complete for {{ item.item }} after
+      3 attempt(s): {{ item.stderr | default('no stderr') }}
+  loop: "{{ kube_vip_image_pull.results | default([]) }}"
+  loop_control:
+    label: "{{ item.item }}"
+  when: item.rc is defined and item.rc != 0
 
 - name: Render kube-vip control plane manifest
   template:

ansible/site.yml

@@ -110,11 +110,12 @@
     - bootstrap-image-prepull
 
 - name: Pre-pull Rancher bootstrap images
-  hosts: cluster
+  hosts: workers
   become: true
 
   roles:
-    - rancher-image-prepull
+    - role: rancher-image-prepull
+      when: rancher_image_prepull_enabled | default(false) | bool
 
 - name: Deploy observability stack
   hosts: control_plane[0]