diff --git a/deploy/k8s/10-deployment.yaml b/deploy/k8s/10-deployment.yaml
index e96d1e2..7790f58 100644
--- a/deploy/k8s/10-deployment.yaml
+++ b/deploy/k8s/10-deployment.yaml
@@ -15,7 +15,8 @@ spec:
   # across a reschedule). The flip side: the HPA does not work on a StatefulSet — the
   # controller and the autoscaler fight over the replica count — so scaling
   # means adding nodes and replicas here, not CPU targets.
-  replicas: 1
+  # Keep DNS available if a node disappears and its pod cannot terminate.
+  replicas: 2
   podManagementPolicy: Parallel
   selector:
     matchLabels: