From 7586e2e35c7bf70a82cf90041789b4499477b12a Mon Sep 17 00:00:00 2001
From: Joey <569475269@qq.com>
Date: Thu, 27 Nov 2025 09:36:01 +0800
Subject: [PATCH] fix: avoid dummy device causing node scale up issue

---
 internal/webhook/v1/pod_webhook.go | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/internal/webhook/v1/pod_webhook.go b/internal/webhook/v1/pod_webhook.go
index 4c6d9322..150c71a6 100644
--- a/internal/webhook/v1/pod_webhook.go
+++ b/internal/webhook/v1/pod_webhook.go
@@ -29,6 +29,7 @@ import (
 	corev1 "k8s.io/api/core/v1"
 	"k8s.io/apimachinery/pkg/api/equality"
 	"k8s.io/apimachinery/pkg/api/errors"
+	"k8s.io/apimachinery/pkg/api/resource"
 	"k8s.io/apimachinery/pkg/util/strategicpatch"
 	ctrl "sigs.k8s.io/controller-runtime"
 	"sigs.k8s.io/controller-runtime/pkg/client"
@@ -331,7 +332,9 @@ func (m *TensorFusionPodMutator) patchTFClient(
 			container.Resources.Limits = make(corev1.ResourceList)
 		}
 		// Limit is set to actual index value (1-512) for Device Plugin to match Pod
-		// container.Resources.Limits[constants.PodIndexAnnotation] = resource.MustParse(strconv.Itoa(index))
+		// ResourceFit of dummy device already ignored in TF scheduler
+		container.Resources.Limits[constants.PodIndexAnnotation] = resource.MustParse(strconv.Itoa(index))
+		container.Resources.Requests[constants.PodIndexAnnotation] = resource.MustParse("0")
 
 		if !isLocalGPU {
 			addConnectionForRemoteFixedReplicaVirtualGPU(pod, container, clientConfig)