fix: append all pod endpoints from profile results

Signed-off-by: kyanokashi <kyanokashi2@gmail.com>

Author: kyanokashi

pkg/plugins/scorer/active_request.go

@@ -176,25 +176,24 @@ func (s *ActiveRequest) PreRequest(
 ) {
 	debugLogger := log.FromContext(ctx).V(logutil.DEBUG)
 
-	targetProfiles := []string{schedulingResult.PrimaryProfileName}
-	if schedulingResult.PrimaryProfileName != "prefill" {
-		targetProfiles = append(targetProfiles, "prefill")
-	}
-
-	podNames := make([]string, 0, len(targetProfiles))
-	for _, profileName := range targetProfiles {
-		profileResult := schedulingResult.ProfileResults[profileName]
+	podNames := make([]string, 0, len(schedulingResult.ProfileResults))
+	for profileName, profileResult := range schedulingResult.ProfileResults {
 		if profileResult == nil || len(profileResult.TargetPods) == 0 {
 			continue
 		}
 
 		podName := profileResult.TargetPods[0].GetPod().NamespacedName.String()
 		podNames = append(podNames, podName)
 		s.incrementPodCount(podName)
-		debugLogger.Info("Added request to cache", "requestId", request.RequestId, "podName", podName)
+		debugLogger.Info(
+			"Added request to cache",
+			"requestId", request.RequestId,
+			"podName", podName,
+			"profileName", profileName,
+		)
 	}
 
-	// add to request cache with TTL
+	// add to request cache
 	s.requestCache.Set(request.RequestId, &requestEntry{PodNames: podNames, RequestID: request.RequestId}, 0) // Use default TTL
 }
 

pkg/plugins/scorer/active_request_test.go

@@ -32,16 +32,15 @@ func newTestRequest(id string) *types.LLMRequest {
 	}
 }
 
-func newTestSchedulingResult(primaryProfile string, profilePods map[string]types.Pod) *types.SchedulingResult {
+func newTestSchedulingResult(profilePods map[string]types.Pod) *types.SchedulingResult {
 	profileResults := make(map[string]*types.ProfileRunResult)
 	for profile, pod := range profilePods {
 		profileResults[profile] = &types.ProfileRunResult{
 			TargetPods: []types.Pod{pod},
 		}
 	}
 	return &types.SchedulingResult{
-		ProfileResults:     profileResults,
-		PrimaryProfileName: primaryProfile,
+		ProfileResults: profileResults,
 	}
 }
 
@@ -140,7 +139,7 @@ func TestActiveRequestScorer_PreRequest(t *testing.T) {
 
 	t.Run("First request", func(t *testing.T) {
 		request := newTestRequest("test-request-1")
-		schedulingResult := newTestSchedulingResult(testProfile, map[string]types.Pod{
+		schedulingResult := newTestSchedulingResult(map[string]types.Pod{
 			testProfile: podA,
 		})
 
@@ -152,7 +151,7 @@ func TestActiveRequestScorer_PreRequest(t *testing.T) {
 
 	t.Run("Second request to multiple pods", func(t *testing.T) {
 		request := newTestRequest("test-request-2")
-		schedulingResult := newTestSchedulingResult(testProfile, map[string]types.Pod{
+		schedulingResult := newTestSchedulingResult(map[string]types.Pod{
 			testProfile: podA,
 			"prefill":   podB,
 		})
@@ -171,11 +170,10 @@ func TestActiveRequestScorer_ResponseComplete(t *testing.T) {
 
 	podA := newTestPod("pod-a", 2)
 	request := newTestRequest("test-request-1")
-	testProfile := "test-profile"
 
 	// Setup initial state: add request through PreRequest
-	schedulingResult := newTestSchedulingResult(testProfile, map[string]types.Pod{
-		testProfile: podA,
+	schedulingResult := newTestSchedulingResult(map[string]types.Pod{
+		"test-profile": podA,
 	})
 	scorer.PreRequest(ctx, request, schedulingResult)
 
@@ -196,7 +194,7 @@ func TestActiveRequestScorer_TTLExpiration(t *testing.T) {
 
 	podA := newTestPod("pod-a", 0)
 	request := newTestRequest("test-request-ttl")
-	schedulingResult := newTestSchedulingResult("test-profile", map[string]types.Pod{
+	schedulingResult := newTestSchedulingResult(map[string]types.Pod{
 		"test-profile": podA,
 	})