fix: retry transient registry pulls

ansible/roles/bootstrap-image-prepull/tasks/main.yml

@@ -6,13 +6,17 @@
       exit 0
     fi
 
-    for attempt in 1 2 3 4 5; do
-      if timeout 180s /usr/local/bin/ctr -n k8s.io images pull "{{ item }}"; then
+    echo "mtu=$(cat /sys/class/net/{{ ansible_default_ipv4.interface | default('eth0') }}/mtu 2>/dev/null || true)"
+    ip route get 1.1.1.1 || true
+
+    for attempt in 1 2 3 4 5 6 7 8 9 10; do
+      echo "pull attempt ${attempt}: {{ item }}"
+      if timeout 240s /usr/local/bin/ctr -n k8s.io images pull "{{ item }}"; then
         echo "pulled image"
         exit 0
       fi
 
-      sleep 10
+      sleep 15
     done
 
     exit 1

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

@@ -6,13 +6,17 @@
       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 "mtu=$(cat /sys/class/net/{{ ansible_default_ipv4.interface | default('eth0') }}/mtu 2>/dev/null || true)"
+    ip route get 1.1.1.1 || true
+
+    for attempt in 1 2 3 4 5; do
+      echo "pull attempt ${attempt}: {{ item }}"
+      if timeout 180s /usr/local/bin/ctr -n k8s.io images pull "{{ item }}"; then
         echo "pulled image"
         exit 0
       fi
 
-      sleep 10
+      sleep 15
     done
 
     exit 1

ansible/roles/rancher-image-prepull/tasks/main.yml

@@ -6,13 +6,17 @@
       exit 0
     fi
 
-    for attempt in 1 2 3 4 5; do
-      if timeout 180s /usr/local/bin/ctr -n k8s.io images pull "{{ item }}"; then
+    echo "mtu=$(cat /sys/class/net/{{ ansible_default_ipv4.interface | default('eth0') }}/mtu 2>/dev/null || true)"
+    ip route get 1.1.1.1 || true
+
+    for attempt in 1 2 3 4 5 6 7 8 9 10; do
+      echo "pull attempt ${attempt}: {{ item }}"
+      if timeout 240s /usr/local/bin/ctr -n k8s.io images pull "{{ item }}"; then
         echo "pulled image"
         exit 0
       fi
 
-      sleep 10
+      sleep 15
     done
 
     exit 1