Merge branch 'main' into ci-add-area-labels

Author: hickeyma

.github/workflows/ci-lint.yaml

@@ -62,7 +62,7 @@ jobs:
         uses: actions/checkout@v7
 
       - name: Set up Go
-        uses: actions/setup-go@v6
+        uses: actions/setup-go@v7
         with:
           go-version-file: go.mod
           cache: true

.github/workflows/ci-pr-checks.yaml

@@ -1,4 +1,4 @@
-# Matrix jobs (e2e-image-common, e2e-image-router, e2e-router) are each
+# Matrix jobs (e2e-image-common, e2e-image-router) are each
 # followed by a *-status job that aggregates their result, since GitHub
 # required status checks cannot reference individual matrix legs.
 name: CI - Test
@@ -64,7 +64,7 @@ jobs:
         uses: actions/checkout@v7
 
       - name: Set up Go
-        uses: actions/setup-go@v6
+        uses: actions/setup-go@v7
         with:
           go-version-file: go.mod
 
@@ -348,29 +348,6 @@ jobs:
     if: ${{ needs.check-changes.outputs.src == 'true' }}
     runs-on: ubuntu-latest
     timeout-minutes: 45
-    strategy:
-      fail-fast: false
-      matrix:
-        suite:
-          - name: pd
-            label-filter: "!Disruptive && !Extended && !SharedStorage && !Metrics"
-            needs-renderer: "false"
-          - name: pd-shared-storage-deprecated
-            label-filter: "SharedStorage && DeprecatedPD"
-            needs-renderer: "false"
-          - name: pd-shared-storage-disagg
-            label-filter: "SharedStorage && Disagg"
-            needs-renderer: "false"
-          - name: pd-metrics
-            label-filter: "Metrics"
-            needs-renderer: "false"
-          - name: extended
-            label-filter: "Extended"
-            needs-renderer: "true"
-          - name: disruption
-            label-filter: "Disruptive"
-            needs-renderer: "false"
-    name: e2e-router (${{ matrix.suite.name }})
     steps:
       - name: Checkout source
         uses: actions/checkout@v7
@@ -404,7 +381,6 @@ jobs:
           path: ${{ runner.temp }}/e2e-images
 
       - name: Download renderer image
-        if: ${{ matrix.suite.needs-renderer == 'true' }}
         uses: actions/download-artifact@v8
         with:
           name: e2e-image-renderer
@@ -417,26 +393,11 @@ jobs:
             docker load --input "$image"
           done
 
-      - name: Run e2e-router (${{ matrix.suite.name }})
+      - name: Run e2e-router
         shell: bash
         env:
           GO_MOD_CACHE_VOL: ${{ steps.e2e-setup.outputs.go-mod-cache }}
           GO_BUILD_CACHE_VOL: ${{ steps.e2e-setup.outputs.go-build-cache }}
-          E2E_LABEL_FILTER: ${{ matrix.suite.label-filter }}
-          LOAD_VLLM_RENDER_IMAGE: ${{ matrix.suite.needs-renderer }}
           PULL_VLLM_RENDER_IMAGE: "false"
           HF_TOKEN: ${{ secrets.HF_TOKEN }}
         run: make test-e2e-run
-
-  e2e-router-status:
-    needs: e2e-router
-    if: always()
-    runs-on: ubuntu-latest
-    steps:
-      - name: Check e2e-router result
-        run: |
-          result="${{ needs.e2e-router.result }}"
-          if [[ "$result" != "success" && "$result" != "skipped" ]]; then
-            echo "::error::e2e-router failed for at least one matrix suite"
-            exit 1
-          fi

.github/workflows/ci-release.yaml

@@ -62,7 +62,7 @@ jobs:
         uses: actions/checkout@v7
 
       - name: Set up Go
-        uses: actions/setup-go@v6
+        uses: actions/setup-go@v7
         with:
           go-version-file: go.mod
 

.github/workflows/nightly-router-perf-test-optimized-baseline-10k-1k.yaml

@@ -67,21 +67,21 @@ jobs:
       HF_TOKEN: ${{ secrets.HF_TOKEN }}
     steps:
       - name: Checkout llm-d/llm-d-router
-        uses: actions/checkout@v6
+        uses: actions/checkout@v7
 
       - name: Checkout kubernetes-sigs/inference-perf
-        uses: actions/checkout@v6
+        uses: actions/checkout@v7
         with:
           repository: kubernetes-sigs/inference-perf
           path: inference-perf
 
       - name: Authenticate to Google Cloud
-        uses: google-github-actions/auth@v2
+        uses: google-github-actions/auth@v3
         with:
           credentials_json: ${{ secrets.GKE_SA_KEY }}
 
       - name: Set up gcloud CLI and kubectl
-        uses: google-github-actions/setup-gcloud@v2
+        uses: google-github-actions/setup-gcloud@v3
         with:
           project_id: ${{ env.GCP_PROJECT_ID }}
           install_components: 'kubectl,gke-gcloud-auth-plugin'
@@ -97,7 +97,7 @@ jobs:
           curl https://raw.githubusercontent.com/helm/helm/main/scripts/get-helm-3 | bash
 
       - name: Set up Python
-        uses: actions/setup-python@v5
+        uses: actions/setup-python@v7
         with:
           python-version: "3.10"
 
@@ -125,7 +125,7 @@ jobs:
             --gcs-bucket ${{ env.GCS_BUCKET }}
 
       - name: Upload Performance Results Artifact
-        uses: actions/upload-artifact@v4
+        uses: actions/upload-artifact@v7
         if: always()
         with:
           name: epp-perf-results

Makefile

@@ -126,7 +126,7 @@ E2E_NUM_PROCS ?= 5
 # Should we pass ALL env vars here?
 E2E_ENV_VARS = EPP_IMAGE VLLM_IMAGE SIDECAR_IMAGE VLLM_RENDER_IMAGE \
                E2E_KEEP_CLUSTER_ON_FAILURE E2E_PORT E2E_METRICS_PORT K8S_CONTEXT READY_TIMEOUT \
-               E2E_NUM_PROCS E2E_LABEL_FILTER LOAD_VLLM_RENDER_IMAGE HF_TOKEN
+               E2E_NUM_PROCS LOAD_VLLM_RENDER_IMAGE HF_TOKEN
 BUILDER_E2E_ENV_FLAGS = $(foreach v,$(E2E_ENV_VARS),$(if $($(v)),-e '$(v)=$($(v))'))
 ifneq ($(filter command line environment,$(origin NAMESPACE)),)
 BUILDER_E2E_ENV_FLAGS += -e NAMESPACE=$(NAMESPACE)

cmd/epp/runner/runner.go

@@ -60,13 +60,16 @@ import (
 	fwkdl "github.com/llm-d/llm-d-router/pkg/epp/framework/interface/datalayer"
 	fwkplugin "github.com/llm-d/llm-d-router/pkg/epp/framework/interface/plugin"
 	attrconcurrency "github.com/llm-d/llm-d-router/pkg/epp/framework/plugins/datalayer/attribute/concurrency"
+	attrgpu "github.com/llm-d/llm-d-router/pkg/epp/framework/plugins/datalayer/attribute/gpu"
 	attrlatency "github.com/llm-d/llm-d-router/pkg/epp/framework/plugins/datalayer/attribute/latency"
 	attrmodels "github.com/llm-d/llm-d-router/pkg/epp/framework/plugins/datalayer/attribute/models"
 	attrprefix "github.com/llm-d/llm-d-router/pkg/epp/framework/plugins/datalayer/attribute/prefix"
 	attrsession "github.com/llm-d/llm-d-router/pkg/epp/framework/plugins/datalayer/attribute/session"
 	discoveryfile "github.com/llm-d/llm-d-router/pkg/epp/framework/plugins/datalayer/discovery/file"
+	extdcgm "github.com/llm-d/llm-d-router/pkg/epp/framework/plugins/datalayer/extractor/dcgm"
 	extractormetrics "github.com/llm-d/llm-d-router/pkg/epp/framework/plugins/datalayer/extractor/metrics"
 	extmodels "github.com/llm-d/llm-d-router/pkg/epp/framework/plugins/datalayer/extractor/models"
+	srcdcgm "github.com/llm-d/llm-d-router/pkg/epp/framework/plugins/datalayer/source/dcgm"
 	sourcemetrics "github.com/llm-d/llm-d-router/pkg/epp/framework/plugins/datalayer/source/metrics"
 	srcmodels "github.com/llm-d/llm-d-router/pkg/epp/framework/plugins/datalayer/source/models"
 	sourcenotifications "github.com/llm-d/llm-d-router/pkg/epp/framework/plugins/datalayer/source/notifications"
@@ -564,6 +567,10 @@ func (r *Runner) registerInTreePlugins() {
 	fwkplugin.Register(srcmodels.ModelsDataSourceType, srcmodels.ModelDataSourceFactory)
 	fwkplugin.Register(attrmodels.ModelsExtractorType, extmodels.ModelServerExtractorFactory)
 
+	// data layer DCGM source/extractor
+	fwkplugin.Register(srcdcgm.DCGMDataSourceType, srcdcgm.DCGMDataSourceFactory)
+	fwkplugin.Register(attrgpu.DCGMExtractorType, extdcgm.DCGMExtractorFactory)
+
 	fwkplugin.Register(prefix.PrefixCacheScorerPluginType, prefix.PrefixCachePluginFactory)
 	fwkplugin.Register(maxscore.MaxScorePickerType, maxscore.MaxScorePickerFactory)
 	fwkplugin.Register(random.RandomPickerType, random.RandomPickerFactory)

pkg/epp/datastore/datastore.go

@@ -315,6 +315,7 @@ func (ds *datastore) podUpdateOrAddIfNotExist(ctx context.Context, pod *corev1.P
 				NamespacedName: createEndpointNamespacedName(pod, idx),
 				PodName:        pod.Name,
 				Address:        pod.Status.PodIP,
+				NodeAddress:    pod.Status.HostIP,
 				Port:           strconv.Itoa(port),
 				MetricsHost:    net.JoinHostPort(pod.Status.PodIP, strconv.Itoa(port)),
 				Labels:         labels,

pkg/epp/datastore/datastore_test.go

@@ -305,6 +305,10 @@ var (
 		ObjectMeta: metav1.ObjectMeta{
 			Name: "pod1",
 		},
+		Status: corev1.PodStatus{
+			PodIP:  "10.0.0.1",
+			HostIP: "192.168.1.10",
+		},
 	}
 	pod1Metrics = &fwkdl.Metrics{
 		WaitingQueueSize:    0,
@@ -320,6 +324,10 @@ var (
 		ObjectMeta: metav1.ObjectMeta{
 			Name: "pod2",
 		},
+		Status: corev1.PodStatus{
+			PodIP:  "10.0.0.2",
+			HostIP: "192.168.1.10",
+		},
 	}
 	pod2Metrics = &fwkdl.Metrics{
 		WaitingQueueSize:    1,
@@ -506,7 +514,10 @@ func TestPods(t *testing.T) {
 				for idx, pm := range podList {
 					gotPods[idx] = &corev1.Pod{
 						ObjectMeta: metav1.ObjectMeta{Name: pm.GetMetadata().PodName, Namespace: pm.GetMetadata().NamespacedName.Namespace},
-						Status:     corev1.PodStatus{PodIP: pm.GetMetadata().GetIPAddress()},
+						Status: corev1.PodStatus{
+							PodIP:  pm.GetMetadata().GetIPAddress(),
+							HostIP: pm.GetMetadata().GetNodeAddress(),
+						},
 					}
 				}
 				if !cmp.Equal(gotPods, test.wantPods, cmpopts.SortSlices(func(a, b *corev1.Pod) bool { return a.Name < b.Name })) {
@@ -649,6 +660,7 @@ func TestEndpointMetadata(t *testing.T) {
 
 					PodName:     pod1.Name,
 					Address:     pod1.Status.PodIP,
+					NodeAddress: pod1.Status.HostIP,
 					Port:        inferencePoolTargetPort,
 					MetricsHost: net.JoinHostPort(pod1.Status.PodIP, inferencePoolTargetPort),
 					Labels:      map[string]string{},
@@ -671,6 +683,7 @@ func TestEndpointMetadata(t *testing.T) {
 
 					PodName:     pod1.Name,
 					Address:     pod1.Status.PodIP,
+					NodeAddress: pod1.Status.HostIP,
 					Port:        inferencePoolMultiTargetPort0,
 					MetricsHost: net.JoinHostPort(pod1.Status.PodIP, inferencePoolMultiTargetPort0),
 					Labels:      map[string]string{},
@@ -683,6 +696,7 @@ func TestEndpointMetadata(t *testing.T) {
 
 					PodName:     pod1.Name,
 					Address:     pod1.Status.PodIP,
+					NodeAddress: pod1.Status.HostIP,
 					Port:        inferencePoolMultiTargetPort1,
 					MetricsHost: net.JoinHostPort(pod1.Status.PodIP, inferencePoolMultiTargetPort1),
 					Labels:      map[string]string{},
@@ -706,6 +720,7 @@ func TestEndpointMetadata(t *testing.T) {
 
 					PodName:     pod1.Name,
 					Address:     pod1.Status.PodIP,
+					NodeAddress: pod1.Status.HostIP,
 					Port:        inferencePoolMultiTargetPort0,
 					MetricsHost: net.JoinHostPort(pod1.Status.PodIP, inferencePoolMultiTargetPort0),
 					Labels:      map[string]string{},
@@ -718,6 +733,7 @@ func TestEndpointMetadata(t *testing.T) {
 
 					PodName:     pod1.Name,
 					Address:     pod1.Status.PodIP,
+					NodeAddress: pod1.Status.HostIP,
 					Port:        inferencePoolMultiTargetPort1,
 					MetricsHost: net.JoinHostPort(pod1.Status.PodIP, inferencePoolMultiTargetPort1),
 					Labels:      map[string]string{},
@@ -731,8 +747,9 @@ func TestEndpointMetadata(t *testing.T) {
 
 					PodName:     pod2.Name,
 					Address:     pod2.Status.PodIP,
+					NodeAddress: pod2.Status.HostIP,
 					Port:        inferencePoolMultiTargetPort0,
-					MetricsHost: net.JoinHostPort(pod1.Status.PodIP, inferencePoolMultiTargetPort0),
+					MetricsHost: net.JoinHostPort(pod2.Status.PodIP, inferencePoolMultiTargetPort0),
 					Labels:      map[string]string{},
 				},
 				{
@@ -743,8 +760,9 @@ func TestEndpointMetadata(t *testing.T) {
 
 					PodName:     pod2.Name,
 					Address:     pod2.Status.PodIP,
+					NodeAddress: pod2.Status.HostIP,
 					Port:        inferencePoolMultiTargetPort1,
-					MetricsHost: net.JoinHostPort(pod1.Status.PodIP, inferencePoolMultiTargetPort1),
+					MetricsHost: net.JoinHostPort(pod2.Status.PodIP, inferencePoolMultiTargetPort1),
 					Labels:      map[string]string{},
 					RankIndex:   1,
 				},
@@ -766,6 +784,7 @@ func TestEndpointMetadata(t *testing.T) {
 
 					PodName:     pod1.Name,
 					Address:     pod1.Status.PodIP,
+					NodeAddress: pod1.Status.HostIP,
 					Port:        inferencePoolMultiTargetPort0,
 					MetricsHost: net.JoinHostPort(pod1.Status.PodIP, inferencePoolMultiTargetPort0),
 					Labels:      map[string]string{},
@@ -778,6 +797,7 @@ func TestEndpointMetadata(t *testing.T) {
 
 					PodName:     pod1.Name,
 					Address:     pod1.Status.PodIP,
+					NodeAddress: pod1.Status.HostIP,
 					Port:        inferencePoolMultiTargetPort1,
 					MetricsHost: net.JoinHostPort(pod1.Status.PodIP, inferencePoolMultiTargetPort1),
 					Labels:      map[string]string{},

pkg/epp/framework/interface/datalayer/endpoint_metadata.go

@@ -28,9 +28,12 @@ type EndpointMetadata struct {
 	NamespacedName types.NamespacedName
 	PodName        string
 	Address        string
-	Port           string
-	MetricsHost    string
-	Labels         map[string]string
+	// NodeAddress is the node IP hosting this pod (pod.Status.HostIP).
+	// Empty for non-Kubernetes discovery sources (e.g. file discovery).
+	NodeAddress string
+	Port        string
+	MetricsHost string
+	Labels      map[string]string
 	// RankIndex is this endpoint's position in the pool's TargetPorts,
 	// identifying the pod-local rank in multi-port deployments.
 	RankIndex int
@@ -59,6 +62,7 @@ func (epm *EndpointMetadata) Clone() *EndpointMetadata {
 		},
 		PodName:     epm.PodName,
 		Address:     epm.Address,
+		NodeAddress: epm.NodeAddress,
 		Port:        epm.Port,
 		MetricsHost: epm.MetricsHost,
 		Labels:      clonedLabels,
@@ -75,6 +79,7 @@ func (epm *EndpointMetadata) Equal(other *EndpointMetadata) bool {
 	return epm.NamespacedName == other.NamespacedName &&
 		epm.PodName == other.PodName &&
 		epm.Address == other.Address &&
+		epm.NodeAddress == other.NodeAddress &&
 		epm.Port == other.Port &&
 		epm.MetricsHost == other.MetricsHost &&
 		epm.RankIndex == other.RankIndex &&
@@ -100,6 +105,14 @@ func (epm *EndpointMetadata) GetIPAddress() string {
 	return epm.Address
 }
 
+// GetNodeAddress returns the IP of the node hosting this endpoint.
+func (epm *EndpointMetadata) GetNodeAddress() string {
+	if epm == nil {
+		return ""
+	}
+	return epm.NodeAddress
+}
+
 // GetPort returns the Endpoint's inference port.
 func (epm *EndpointMetadata) GetPort() string {
 	return epm.Port

pkg/epp/framework/plugins/datalayer/attribute/gpu/README.md

@@ -0,0 +1,15 @@
+# GPU Attributes
+
+This package defines the data structures for GPU hardware metrics collected from NVIDIA DCGM Exporter.
+
+## `GPUUtilization`
+
+A normalized GPU compute utilization value in [0.0, 1.0], derived from `DCGM_FI_DEV_GPU_UTIL` (which reports 0-100). For multi-GPU pods the extractor aggregates across visible devices using `max`.
+
+- **Key**: `GPUUtilizationDataKey` (`GPUUtilization/dcgm-extractor`)
+
+## Producers
+
+The following plugins produce this attribute:
+
+- **`dcgm-extractor`** (Data Layer): Extracts GPU utilization from the DCGM Exporter Prometheus endpoint.