fix: delayed MHC replacement of unreachable nodes

Co-authored-by: Michael Shen <[email protected]>

Author: typeid

internal/controllers/machine/machine_controller.go

@@ -670,8 +670,19 @@ func (r *Reconciler) drainNode(ctx context.Context, cluster *clusterv1.Cluster,
 	}
 
 	if noderefutil.IsNodeUnreachable(node) {
-		// When the node is unreachable and some pods are not evicted for as long as this timeout, we ignore them.
-		drainer.SkipWaitForDeleteTimeoutSeconds = 60 * 5 // 5 minutes
+		// Kubelet is unreachable, pods will never disappear.
+
+		// SkipWaitForDeleteTimeoutSeconds ensures the drain completes
+		// even if pod objects are not deleted.
+		drainer.SkipWaitForDeleteTimeoutSeconds = 1
+
+		// We don't need to wait for the grace period specified by pods
+		// as we know this won't happen.
+		// The grace period generally won't change anything here as
+		// the kubelet should not be reachable, but it doesn't hurt setting it.
+		drainer.GracePeriodSeconds = 1
+
+		log.V(5).Info("Node is unreachable, draining will ignore gracePeriod. PDBs are still honored.")
 	}
 
 	if err := kubedrain.RunCordonOrUncordon(drainer, node, true); err != nil {