Fix #7991: ensure kubelet_shutdown_grace_period and kubelet_shutdown_…

…grace_period_critical_pods are compared as integers

roles/kubernetes/preinstall/tasks/0020-verify-settings.yml

@@ -109,7 +109,7 @@
 
 - name: Stop when ShutdownGracePeriod less than ShutdownGracePeriodCriticalPods
   assert:
-    that: kubelet_shutdown_grace_period > kubelet_shutdown_grace_period_critical_pods
+    that: kubelet_shutdown_grace_period|int > kubelet_shutdown_grace_period_critical_pods|int
     msg: "ShutdownGracePeriod ({{ kubelet_shutdown_grace_period }}) needs to be greater than ShutdownGracePeriodCriticalPods ({{ kubelet_shutdown_grace_period_critical_pods }}) in order to give normal pods time to be evacuated, please see https://kubernetes.io/blog/2021/04/21/graceful-node-shutdown-beta/ for details"
   when:
     - kube_version is version('v1.21.0', '>=')

roles/kubespray-defaults/defaults/main.yaml

@@ -233,6 +233,10 @@ kube_profiling: false
 
 # Graceful Node Shutdown
 # This requires kubernetes >= 1.21.0
+#
+# Note, when using both kubelet_shutdown_grace_period and kubelet_shutdown_grace_period_critical_pods
+# ensure to use the same meaasurement unit (i.e. s=seconds, m=minutes) because ansible is not aware
+# of the variable type and can cause an assertion failure when validating your input
 kubelet_shutdown_grace_period: 60s
 # kubelet_shutdown_grace_period_critical_pods should be less than kubelet_shutdown_grace_period
 # to give normal pods time to be gracefully evacuated