Merge remote-tracking branch 'upstream/incubating'

Author: spolti

config/runtimes/ovms-kserve-template.yaml

@@ -25,6 +25,7 @@ objects:
     spec:
       multiModel: false
       annotations:
+        opendatahub.io/kserve-runtime: 'ovms'
         prometheus.io/port: '8888'
         prometheus.io/path: /metrics
       supportedModelFormats:

config/runtimes/vllm-cpu-template.yaml

@@ -27,6 +27,7 @@ objects:
         opendatahub.io/dashboard: 'true'
     spec:
       annotations:
+        opendatahub.io/kserve-runtime: 'vllm'
         prometheus.io/port: '8080'
         prometheus.io/path: '/metrics'
       multiModel: false

config/runtimes/vllm-cuda-template.yaml

@@ -28,6 +28,7 @@ objects:
         opendatahub.io/dashboard: 'true'
     spec:
       annotations:
+        opendatahub.io/kserve-runtime: 'vllm'
         prometheus.io/port: '8080'
         prometheus.io/path: '/metrics'
       multiModel: false

config/runtimes/vllm-gaudi-template.yaml

@@ -28,6 +28,7 @@ objects:
         opendatahub.io/dashboard: 'true'
     spec:
       annotations:
+        opendatahub.io/kserve-runtime: 'vllm'
         prometheus.io/port: '8080'
         prometheus.io/path: '/metrics'
       multiModel: false

config/runtimes/vllm-multinode-template.yaml

@@ -28,6 +28,7 @@ objects:
         opendatahub.io/dashboard: 'false'
     spec:
       annotations:
+        opendatahub.io/kserve-runtime: 'vllm'
         prometheus.io/port: '8080'
         prometheus.io/path: '/metrics'
       multiModel: false

config/runtimes/vllm-rocm-template.yaml

@@ -28,6 +28,7 @@ objects:
         opendatahub.io/dashboard: 'true'
     spec:
       annotations:
+        opendatahub.io/kserve-runtime: 'vllm'
         prometheus.io/port: '8080'
         prometheus.io/path: '/metrics'
       multiModel: false

config/runtimes/vllm-spyre-s390x-template.yaml

@@ -28,6 +28,7 @@ objects:
         opendatahub.io/dashboard: 'true'
     spec:
       annotations:
+        opendatahub.io/kserve-runtime: 'vllm'
         prometheus.io/path: /metrics
         prometheus.io/port: '8080'
       containers:

config/runtimes/vllm-spyre-x86-template.yaml

@@ -28,6 +28,7 @@ objects:
         opendatahub.io/dashboard: 'true'
     spec:
       annotations:
+        opendatahub.io/kserve-runtime: 'vllm'
         prometheus.io/path: /metrics
         prometheus.io/port: '8080'
       containers:

internal/controller/constants/constants.go

@@ -31,6 +31,7 @@ const (
 	KserveNetworkVisibility = "networking.kserve.io/visibility"
 	KserveGroupAnnotation   = "serving.kserve.io/inferenceservice"
 	RhoaiObservabilityLabel = "monitoring.opendatahub.io/scrape"
+	RuntimesBaseAnnotation  = "opendatahub.io"
 
 	ODHManagedLabel         = "opendatahub.io/managed"
 	EnableAuthODHAnnotation = "security.opendatahub.io/enable-auth"
@@ -90,11 +91,11 @@ const (
 	IntervalValue                    = "1m"
 	RequestRateInterval              = "5m"
 	GPUKVCacheSamplingInterval       = "24h"
-	OvmsImageName                    = "openvino_model_server"
-	TgisImageName                    = "text-generation-inference"
-	VllmImageName                    = "vllm"
-	CaikitImageName                  = "caikit-nlp"
-	ServingRuntimeFallBackImageName  = "unsupported"
+
+	KServeRuntimeAnnotation = RuntimesBaseAnnotation + "/kserve-runtime"
+	OvmsRuntimeName         = "ovms"
+	TgisRuntimeName         = "tgis"
+	VllmRuntimeName         = "vllm"
 )
 
 // openshift

internal/controller/serving/inferenceservice_controller_test.go

@@ -40,21 +40,16 @@ import (
 	"sigs.k8s.io/controller-runtime/pkg/client"
 
 	"github.com/opendatahub-io/odh-model-controller/internal/controller/constants"
-	"github.com/opendatahub-io/odh-model-controller/internal/controller/utils"
 	. "github.com/opendatahub-io/odh-model-controller/test/matchers"
 	testutils "github.com/opendatahub-io/odh-model-controller/test/utils"
 )
 
 const (
 	KserveOvmsInferenceServiceName         = "example-onnx-mnist"
 	UnsupportedMetricsInferenceServiceName = "sklearn-v2-iris"
-	NilRuntimeInferenceServiceName         = "sklearn-v2-iris-no-runtime"
-	NilModelInferenceServiceName           = "custom-runtime"
 
 	UnsupportedMetricsInferenceServicePath = "./testdata/deploy/kserve-unsupported-metrics-inference-service.yaml"
 	UnsupprtedMetricsServingRuntimePath    = "./testdata/deploy/kserve-unsupported-metrics-serving-runtime.yaml"
-	NilRuntimeInferenceServicePath         = "./testdata/deploy/kserve-nil-runtime-inference-service.yaml"
-	NilModelInferenceServicePath           = "./testdata/deploy/kserve-nil-model-inference-service.yaml"
 )
 
 var _ = Describe("InferenceService Controller", func() {
@@ -94,38 +89,6 @@ var _ = Describe("InferenceService Controller", func() {
 			return inferenceService
 		}
 
-		verifyConfigMap := func(isvcName string, namespace string, supported bool, metricsData string) {
-			metricsConfigMap, err := testutils.WaitForConfigMap(k8sClient, namespace, isvcName+constants.KserveMetricsConfigMapNameSuffix, 30, 1*time.Second)
-			Expect(err).NotTo(HaveOccurred())
-			Expect(metricsConfigMap).NotTo(BeNil())
-			var expectedMetricsConfigMap *corev1.ConfigMap
-			if supported {
-				finaldata := utils.SubstituteVariablesInQueries(metricsData, namespace, isvcName)
-				expectedMetricsConfigMap = &corev1.ConfigMap{
-					ObjectMeta: metav1.ObjectMeta{
-						Name:      isvcName + constants.KserveMetricsConfigMapNameSuffix,
-						Namespace: namespace,
-					},
-					Data: map[string]string{
-						"supported": "true",
-						"metrics":   finaldata,
-					},
-				}
-			} else {
-				expectedMetricsConfigMap = &corev1.ConfigMap{
-					ObjectMeta: metav1.ObjectMeta{
-						Name:      UnsupportedMetricsInferenceServiceName + constants.KserveMetricsConfigMapNameSuffix,
-						Namespace: testNs,
-					},
-					Data: map[string]string{
-						"supported": "false",
-					},
-				}
-			}
-			Expect(testutils.CompareConfigMap(metricsConfigMap, expectedMetricsConfigMap)).Should(BeTrue())
-			Expect(expectedMetricsConfigMap.Data).NotTo(HaveKeyWithValue("metrics", ContainSubstring("${REQUEST_RATE_INTERVAL}")))
-		}
-
 		BeforeEach(func() {
 			testNs = testutils.Namespaces.Create(ctx, k8sClient).Name
 
@@ -136,34 +99,6 @@ var _ = Describe("InferenceService Controller", func() {
 			}
 		})
 
-		When("deploying a Kserve model", func() {
-			It("[raw] if the runtime is supported for metrics, it should create a configmap with prometheus queries", func() {
-				_ = createServingRuntime(testNs, KserveServingRuntimePath1)
-				_ = createInferenceService(testNs, KserveOvmsInferenceServiceName, KserveInferenceServicePath1, true)
-
-				verifyConfigMap(KserveOvmsInferenceServiceName, testNs, true, constants.OvmsMetricsData)
-			})
-
-			It("[raw] if the runtime is not supported for metrics, it should create a configmap with the unsupported config", func() {
-				_ = createServingRuntime(testNs, UnsupprtedMetricsServingRuntimePath)
-				_ = createInferenceService(testNs, UnsupportedMetricsInferenceServiceName, UnsupportedMetricsInferenceServicePath, true)
-
-				verifyConfigMap(UnsupportedMetricsInferenceServiceName, testNs, false, "")
-			})
-
-			It("[raw] if the isvc does not have a runtime specified and there is no supported runtime, an unsupported metrics configmap should be created", func() {
-				_ = createInferenceService(testNs, NilRuntimeInferenceServiceName, NilRuntimeInferenceServicePath, true)
-
-				verifyConfigMap(NilRuntimeInferenceServiceName, testNs, false, "")
-			})
-
-			It("[raw] if the isvc does not have the model field specified, an unsupported metrics configmap should be created", func() {
-				_ = createInferenceService(testNs, NilModelInferenceServiceName, NilModelInferenceServicePath, true)
-
-				verifyConfigMap(NilModelInferenceServiceName, testNs, false, "")
-			})
-		})
-
 		When("deleting the deployed models", func() {
 			timeout10s := time.Second * 10
 			interval4s := time.Second * 4