[model server] Add inference service config tests

tests/model_serving/model_server/components/constants.py

@@ -0,0 +1,19 @@
+from utilities.constants import (
+    KServeDeploymentType,
+    ModelAndFormat,
+    ModelFormat,
+    ModelInferenceRuntime,
+    ModelStoragePath,
+    ModelVersion,
+)
+
+KSERVE_RUNTIME_PARAMS = {
+    "runtime-name": ModelInferenceRuntime.OPENVINO_KSERVE_RUNTIME,
+    "model-format": {ModelAndFormat.OPENVINO_IR: ModelVersion.OPSET1},
+}
+SERVERLESS_ISVC_PARAMS = {
+    "name": ModelFormat.OPENVINO,
+    "model-version": ModelVersion.OPSET1,
+    "model-dir": ModelStoragePath.KSERVE_OPENVINO_EXAMPLE_MODEL,
+    "deployment-mode": KServeDeploymentType.SERVERLESS,
+}

tests/model_serving/model_server/components/model_mesh_kserve_co_exist/test_model_mesh_kserve_inference_co_exist.py

@@ -1,62 +1,50 @@
 import pytest
 
+from tests.model_serving.model_server.components.constants import KSERVE_RUNTIME_PARAMS, SERVERLESS_ISVC_PARAMS
 from tests.model_serving.model_server.utils import verify_inference_response
 from utilities.constants import (
-    ModelAndFormat,
-    ModelFormat,
-    ModelInferenceRuntime,
     ModelStoragePath,
-    ModelVersion,
     Protocols,
 )
 from utilities.inference_utils import Inference
 from utilities.manifests.openvino import OPENVINO_INFERENCE_CONFIG, OPENVINO_KSERVE_INFERENCE_CONFIG
 
 pytestmark = [pytest.mark.serverless, pytest.mark.modelmesh, pytest.mark.sanity]
 
-KSERVE_RUNTIME_PARAMS = {
-    "runtime-name": ModelInferenceRuntime.OPENVINO_KSERVE_RUNTIME,
-    "model-format": {ModelAndFormat.OPENVINO_IR: ModelVersion.OPSET1},
-}
-KSERVE_ISVC_PARAMS = {
-    "name": ModelFormat.OPENVINO,
-    "model-version": ModelVersion.OPSET1,
-    "model-dir": ModelStoragePath.KSERVE_OPENVINO_EXAMPLE_MODEL,
-}
 MODELMESH_ISVC_PARAMS = {
     "model-path": ModelStoragePath.OPENVINO_EXAMPLE_MODEL,
     "modelmesh-enabled": True,
 }
 
 
 @pytest.mark.parametrize(
-    "model_namespace, openvino_kserve_serving_runtime, ovms_serverless_inference_service, "
+    "model_namespace, openvino_kserve_serving_runtime, ovms_kserve_inference_service, "
     "http_s3_openvino_model_mesh_inference_service",
     [
         pytest.param(
             {"name": "serverless-model-mesh-openvino", "modelmesh-enabled": True},
             KSERVE_RUNTIME_PARAMS,
-            KSERVE_ISVC_PARAMS,
+            SERVERLESS_ISVC_PARAMS,
             MODELMESH_ISVC_PARAMS,
         )
     ],
     indirect=True,
 )
 class TestOpenVINOServerlessModelMesh:
     def test_serverless_openvino_created_before_model_mesh_ns_rest_inference(
-        self, ovms_serverless_inference_service, http_s3_openvino_model_mesh_inference_service
+        self, ovms_kserve_inference_service, http_s3_openvino_model_mesh_inference_service
     ):
         """Verify that Serverless model can be queried when running with modelmesh inference service"""
         verify_inference_response(
-            inference_service=ovms_serverless_inference_service,
+            inference_service=ovms_kserve_inference_service,
             inference_config=OPENVINO_KSERVE_INFERENCE_CONFIG,
             inference_type=Inference.INFER,
             protocol=Protocols.HTTPS,
             use_default_query=True,
         )
 
     def test_model_mesh_openvino_created_after_serverless_in_namespace_rest_inference(
-        self, ovms_serverless_inference_service, http_s3_openvino_model_mesh_inference_service
+        self, ovms_kserve_inference_service, http_s3_openvino_model_mesh_inference_service
     ):
         """Verify that modelmesh model can be queried when running with kserve inference service"""
         verify_inference_response(
@@ -70,20 +58,20 @@ def test_model_mesh_openvino_created_after_serverless_in_namespace_rest_inferenc
 
 @pytest.mark.parametrize(
     "model_namespace, http_s3_openvino_model_mesh_inference_service, openvino_kserve_serving_runtime, "
-    "ovms_serverless_inference_service, ",
+    "ovms_kserve_inference_service, ",
     [
         pytest.param(
             {"name": "model-mesh-serverless-openvino", "modelmesh-enabled": True},
             MODELMESH_ISVC_PARAMS,
             KSERVE_RUNTIME_PARAMS,
-            KSERVE_ISVC_PARAMS,
+            SERVERLESS_ISVC_PARAMS,
         )
     ],
     indirect=True,
 )
 class TestOpenVINOModelMeshServerless:
     def test_model_mesh_openvino_created_before_serverless_in_namespace_rest_inference(
-        self, http_s3_openvino_model_mesh_inference_service, ovms_serverless_inference_service
+        self, http_s3_openvino_model_mesh_inference_service, ovms_kserve_inference_service
     ):
         """Verify that modelmesh model can be queried when running with kserve inference service"""
         verify_inference_response(
@@ -95,11 +83,11 @@ def test_model_mesh_openvino_created_before_serverless_in_namespace_rest_inferen
         )
 
     def test_serverless_openvino_created_after_model_mesh_ns_rest_inference(
-        self, http_s3_openvino_model_mesh_inference_service, ovms_serverless_inference_service
+        self, http_s3_openvino_model_mesh_inference_service, ovms_kserve_inference_service
     ):
         """Verify that Serverless model can be queried when running with modelmesh inference service"""
         verify_inference_response(
-            inference_service=ovms_serverless_inference_service,
+            inference_service=ovms_kserve_inference_service,
             inference_config=OPENVINO_KSERVE_INFERENCE_CONFIG,
             inference_type=Inference.INFER,
             protocol=Protocols.HTTPS,

tests/model_serving/model_server/components/raw_deployment_serverless_co_exist/test_raw_deployment_serverless_inference_co_exist.py

@@ -1,14 +1,13 @@
 import pytest
 
+from tests.model_serving.model_server.components.constants import KSERVE_RUNTIME_PARAMS, SERVERLESS_ISVC_PARAMS
 from tests.model_serving.model_server.utils import verify_inference_response
 from utilities.constants import (
     KServeDeploymentType,
-    ModelAndFormat,
     ModelFormat,
     ModelInferenceRuntime,
     ModelName,
     ModelStoragePath,
-    ModelVersion,
     Protocols,
     RuntimeTemplates,
 )
@@ -20,16 +19,6 @@
 
 pytestmark = [pytest.mark.serverless, pytest.mark.rawdeployment, pytest.mark.sanity]
 
-
-SERVERLESS_RUNTIME_PARAMS = {
-    "runtime-name": ModelInferenceRuntime.OPENVINO_KSERVE_RUNTIME,
-    "model-format": {ModelAndFormat.OPENVINO_IR: ModelVersion.OPSET1},
-}
-SERVERLESS_ISVC_PARAMS = {
-    "name": ModelFormat.OPENVINO,
-    "model-version": ModelVersion.OPSET1,
-    "model-dir": ModelStoragePath.KSERVE_OPENVINO_EXAMPLE_MODEL,
-}
 RAW_RUNTIME_PARAMS = {
     "name": f"{Protocols.HTTP}-{ModelInferenceRuntime.CAIKIT_STANDALONE_RUNTIME}".lower(),
     "template-name": RuntimeTemplates.CAIKIT_STANDALONE_SERVING,
@@ -44,12 +33,12 @@
 
 
 @pytest.mark.parametrize(
-    "model_namespace, openvino_kserve_serving_runtime, ovms_serverless_inference_service, "
+    "model_namespace, openvino_kserve_serving_runtime, ovms_kserve_inference_service, "
     "serving_runtime_from_template, s3_models_inference_service",
     [
         pytest.param(
             {"name": "serverless-raw-deployment"},
-            SERVERLESS_RUNTIME_PARAMS,
+            KSERVE_RUNTIME_PARAMS,
             SERVERLESS_ISVC_PARAMS,
             RAW_RUNTIME_PARAMS,
             RAW_ISVC_PARAMS,
@@ -59,11 +48,11 @@
 )
 class TestServerlessRawInternalDeploymentInferenceCoExist:
     def test_serverless_openvino_created_before_raw_internal_deployment_caikit_inference(
-        self, ovms_serverless_inference_service, s3_models_inference_service
+        self, ovms_kserve_inference_service, s3_models_inference_service
     ):
         """Verify that Serverless model can be queried when running with raw deployment inference service"""
         verify_inference_response(
-            inference_service=ovms_serverless_inference_service,
+            inference_service=ovms_kserve_inference_service,
             inference_config=OPENVINO_KSERVE_INFERENCE_CONFIG,
             inference_type=Inference.INFER,
             protocol=Protocols.HTTPS,
@@ -72,7 +61,7 @@ def test_serverless_openvino_created_before_raw_internal_deployment_caikit_infer
 
     def test_raw_internal_deployment_caikit_created_after_serverless_in_namespace_rest_inference(
         self,
-        ovms_serverless_inference_service,
+        ovms_kserve_inference_service,
         s3_models_inference_service,
     ):
         """Verify that raw deployment model can be queried when running with kserve inference service"""
@@ -87,12 +76,12 @@ def test_raw_internal_deployment_caikit_created_after_serverless_in_namespace_re
 
 
 @pytest.mark.parametrize(
-    "model_namespace, openvino_kserve_serving_runtime, ovms_serverless_inference_service, "
+    "model_namespace, openvino_kserve_serving_runtime, ovms_kserve_inference_service, "
     "serving_runtime_from_template, s3_models_inference_service",
     [
         pytest.param(
             {"name": "serverless-raw-ext-deployment"},
-            SERVERLESS_RUNTIME_PARAMS,
+            KSERVE_RUNTIME_PARAMS,
             SERVERLESS_ISVC_PARAMS,
             RAW_RUNTIME_PARAMS,
             {**RAW_ISVC_PARAMS, "external-route": True},
@@ -102,11 +91,11 @@ def test_raw_internal_deployment_caikit_created_after_serverless_in_namespace_re
 )
 class TestServerlessRawExternalDeploymentInferenceCoExist:
     def test_serverless_openvino_created_before_raw_external_deployment_caikit_inference(
-        self, ovms_serverless_inference_service, s3_models_inference_service
+        self, ovms_kserve_inference_service, s3_models_inference_service
     ):
         """Verify that Serverless model can be queried when running with raw deployment inference service"""
         verify_inference_response(
-            inference_service=ovms_serverless_inference_service,
+            inference_service=ovms_kserve_inference_service,
             inference_config=OPENVINO_KSERVE_INFERENCE_CONFIG,
             inference_type=Inference.INFER,
             protocol=Protocols.HTTPS,
@@ -115,7 +104,7 @@ def test_serverless_openvino_created_before_raw_external_deployment_caikit_infer
 
     def test_raw_external_deployment_caikit_created_after_serverless_in_namespace_rest_inference(
         self,
-        ovms_serverless_inference_service,
+        ovms_kserve_inference_service,
         s3_models_inference_service,
     ):
         """Verify that raw deployment model can be queried when running with kserve inference service"""
@@ -131,13 +120,13 @@ def test_raw_external_deployment_caikit_created_after_serverless_in_namespace_re
 
 @pytest.mark.parametrize(
     "model_namespace, serving_runtime_from_template, s3_models_inference_service,"
-    "openvino_kserve_serving_runtime, ovms_serverless_inference_service",
+    "openvino_kserve_serving_runtime, ovms_kserve_inference_service",
     [
         pytest.param(
             {"name": "raw-deployment-serverless"},
             RAW_RUNTIME_PARAMS,
             RAW_ISVC_PARAMS,
-            SERVERLESS_RUNTIME_PARAMS,
+            KSERVE_RUNTIME_PARAMS,
             SERVERLESS_ISVC_PARAMS,
         )
     ],
@@ -147,7 +136,7 @@ class TestRawInternalDeploymentServerlessInferenceCoExist:
     def test_raw_internal_deployment_caikit_created_before_serverless_openvino_in_namespace_rest_inference(
         self,
         s3_models_inference_service,
-        ovms_serverless_inference_service,
+        ovms_kserve_inference_service,
     ):
         """Verify that raw deployment model can be queried when running with kserve inference service"""
         verify_inference_response(
@@ -162,11 +151,11 @@ def test_raw_internal_deployment_caikit_created_before_serverless_openvino_in_na
     def test_serverless_openvino_created_after_raw_internal_deployment_caikit_ns_rest_inference(
         self,
         s3_models_inference_service,
-        ovms_serverless_inference_service,
+        ovms_kserve_inference_service,
     ):
         """Verify that Serverless model can be queried when running with raw deployment exists"""
         verify_inference_response(
-            inference_service=ovms_serverless_inference_service,
+            inference_service=ovms_kserve_inference_service,
             inference_config=OPENVINO_KSERVE_INFERENCE_CONFIG,
             inference_type=Inference.INFER,
             protocol=Protocols.HTTPS,
@@ -176,13 +165,13 @@ def test_serverless_openvino_created_after_raw_internal_deployment_caikit_ns_res
 
 @pytest.mark.parametrize(
     "model_namespace, serving_runtime_from_template, s3_models_inference_service,"
-    "openvino_kserve_serving_runtime, ovms_serverless_inference_service",
+    "openvino_kserve_serving_runtime, ovms_kserve_inference_service",
     [
         pytest.param(
             {"name": "raw-etx-deployment-serverless"},
             RAW_RUNTIME_PARAMS,
             {**RAW_ISVC_PARAMS, "external-route": True},
-            SERVERLESS_RUNTIME_PARAMS,
+            KSERVE_RUNTIME_PARAMS,
             SERVERLESS_ISVC_PARAMS,
         ),
     ],
@@ -192,7 +181,7 @@ class TestRawExternalDeploymentServerlessInferenceCoExist:
     def test_raw_external_deployment_caikit_created_before_serverless_openvino_in_namespace_rest_inference(
         self,
         s3_models_inference_service,
-        ovms_serverless_inference_service,
+        ovms_kserve_inference_service,
     ):
         """Verify that raw deployment model can be queried when running with kserve inference service"""
         verify_inference_response(
@@ -205,11 +194,11 @@ def test_raw_external_deployment_caikit_created_before_serverless_openvino_in_na
         )
 
     def test_serverless_openvino_created_after_raw_external_deployment_caikit_ns_rest_inference(
-        self, s3_models_inference_service, ovms_serverless_inference_service
+        self, s3_models_inference_service, ovms_kserve_inference_service
     ):
         """Verify that Serverless model can be queried when running with raw deployment exists"""
         verify_inference_response(
-            inference_service=ovms_serverless_inference_service,
+            inference_service=ovms_kserve_inference_service,
             inference_config=OPENVINO_KSERVE_INFERENCE_CONFIG,
             inference_type=Inference.INFER,
             protocol=Protocols.HTTPS,

tests/model_serving/model_server/conftest.py

@@ -342,24 +342,36 @@ def ci_service_account(
 
 
 @pytest.fixture(scope="class")
-def ovms_serverless_inference_service(
+def ovms_kserve_inference_service(
     request: FixtureRequest,
     admin_client: DynamicClient,
     model_namespace: Namespace,
     openvino_kserve_serving_runtime: ServingRuntime,
     ci_endpoint_s3_secret: Secret,
 ) -> Generator[InferenceService, Any, Any]:
-    with create_isvc(
-        client=admin_client,
-        name=f"{request.param['name']}-serverless",
-        namespace=model_namespace.name,
-        runtime=openvino_kserve_serving_runtime.name,
-        storage_path=request.param["model-dir"],
-        storage_key=ci_endpoint_s3_secret.name,
-        model_format=ModelAndFormat.OPENVINO_IR,
-        deployment_mode=KServeDeploymentType.SERVERLESS,
-        model_version=request.param["model-version"],
-    ) as isvc:
+    deployment_mode = request.param["deployment-mode"]
+    isvc_kwargs = {
+        "client": admin_client,
+        "name": f"{request.param['name']}-{deployment_mode.lower()}",
+        "namespace": model_namespace.name,
+        "runtime": openvino_kserve_serving_runtime.name,
+        "storage_path": request.param["model-dir"],
+        "storage_key": ci_endpoint_s3_secret.name,
+        "model_format": ModelAndFormat.OPENVINO_IR,
+        "deployment_mode": deployment_mode,
+        "model_version": request.param["model-version"],
+    }
+
+    if env_vars := request.param.get("env-vars"):
+        isvc_kwargs["model_env_variables"] = env_vars
+
+    if min_replicas := request.param.get("min-replicas"):
+        isvc_kwargs["min_replicas"] = min_replicas
+
+    if max_replicas := request.param.get("max-replicas"):
+        isvc_kwargs["max_replicas"] = max_replicas
+
+    with create_isvc(**isvc_kwargs) as isvc:
         yield isvc