Revert start model deployment being in TransportPutInferenceModelAction

Author: dan-rubinstein

x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/action/TransportPutInferenceModelAction.java

@@ -191,23 +191,19 @@ private void parseAndStoreModel(
         ActionListener<Model> storeModelListener = listener.delegateFailureAndWrap(
             (delegate, verifiedModel) -> modelRegistry.storeModel(
                 verifiedModel,
-                ActionListener.wrap(
-                    r -> listener.onResponse(new PutInferenceModelAction.Response(verifiedModel.getConfigurations())),
-                    e -> {
-                        if (e.getCause() instanceof StrictDynamicMappingException
-                            && e.getCause().getMessage().contains("chunking_settings")) {
-                            delegate.onFailure(
-                                new ElasticsearchStatusException(
-                                    "One or more nodes in your cluster does not support chunking_settings. "
-                                        + "Please update all nodes in your cluster to the latest version to use chunking_settings.",
-                                    RestStatus.BAD_REQUEST
-                                )
-                            );
-                        } else {
-                            delegate.onFailure(e);
-                        }
+                ActionListener.wrap(r -> startInferenceEndpoint(service, timeout, verifiedModel, delegate), e -> {
+                    if (e.getCause() instanceof StrictDynamicMappingException && e.getCause().getMessage().contains("chunking_settings")) {
+                        delegate.onFailure(
+                            new ElasticsearchStatusException(
+                                "One or more nodes in your cluster does not support chunking_settings. "
+                                    + "Please update all nodes in your cluster to the latest version to use chunking_settings.",
+                                RestStatus.BAD_REQUEST
+                            )
+                        );
+                    } else {
+                        delegate.onFailure(e);
                     }
-                ),
+                }),
                 timeout
             )
         );
@@ -224,6 +220,19 @@ private void parseAndStoreModel(
         service.parseRequestConfig(inferenceEntityId, taskType, config, parsedModelListener);
     }
 
+    private void startInferenceEndpoint(
+        InferenceService service,
+        TimeValue timeout,
+        Model model,
+        ActionListener<PutInferenceModelAction.Response> listener
+    ) {
+        if (skipValidationAndStart) {
+            listener.onResponse(new PutInferenceModelAction.Response(model.getConfigurations()));
+        } else {
+            service.start(model, timeout, listener.map(started -> new PutInferenceModelAction.Response(model.getConfigurations())));
+        }
+    }
+
     private Map<String, Object> requestToMap(PutInferenceModelAction.Request request) throws IOException {
         try (
             XContentParser parser = XContentHelper.createParser(

x-pack/plugin/inference/src/test/java/org/elasticsearch/xpack/inference/services/validation/ElasticsearchInternalServiceModelValidatorTests.java

@@ -178,6 +178,33 @@ public void testValidate_ElandTextEmbeddingModelValidationSucceedsAndDimensionsS
         );
     }
 
+    public void testValidate_ElandTextEmbeddingAndValidationReturnsInvalidResultsType() {
+        var dimensions = randomIntBetween(1, 10);
+        var mockInferenceServiceResults = mock(InferenceServiceResults.class);
+        when(mockInferenceServiceResults.getWriteableName()).thenReturn(randomAlphaOfLength(10));
+        CustomElandEmbeddingModel customElandEmbeddingModel = createCustomElandEmbeddingModel(true, dimensions);
+
+        doAnswer(ans -> {
+            ActionListener<InferenceServiceResults> responseListener = ans.getArgument(3);
+            responseListener.onResponse(mockInferenceServiceResults);
+            return null;
+        }).when(mockServiceIntegrationValidator).validate(eq(mockInferenceService), any(), eq(TIMEOUT), any());
+
+        underTest.validate(mockInferenceService, customElandEmbeddingModel, TIMEOUT, mockActionListener);
+
+        verify(mockServiceIntegrationValidator).validate(eq(mockInferenceService), any(), eq(TIMEOUT), any());
+        verify(mockActionListener).delegateFailureAndWrap(any());
+        verify(mockActionListener).onFailure(any(ElasticsearchStatusException.class));
+        verify(mockInferenceServiceResults).getWriteableName();
+        verifyNoMoreInteractions(
+            mockServiceIntegrationValidator,
+            mockInferenceService,
+            mockCustomElandEmbeddingModel,
+            mockActionListener,
+            mockInferenceServiceResults
+        );
+    }
+
     public void testValidate_ElandTextEmbeddingModelDimensionsNotSetByUser() {
         var dimensions = randomIntBetween(1, 10);
         var mockInferenceServiceResults = mock(TextEmbeddingResults.class);
@@ -211,6 +238,32 @@ public void testValidate_ElandTextEmbeddingModelDimensionsNotSetByUser() {
         );
     }
 
+    public void testValidate_ElandTextEmbeddingModelAndEmbeddingSizeRetrievalThrowsException() {
+        var mockInferenceServiceResults = mock(TextEmbeddingResults.class);
+        when(mockInferenceServiceResults.getFirstEmbeddingSize()).thenThrow(ElasticsearchStatusException.class);
+        CustomElandEmbeddingModel customElandEmbeddingModel = createCustomElandEmbeddingModel(false, null);
+
+        doAnswer(ans -> {
+            ActionListener<InferenceServiceResults> responseListener = ans.getArgument(3);
+            responseListener.onResponse(mockInferenceServiceResults);
+            return null;
+        }).when(mockServiceIntegrationValidator).validate(eq(mockInferenceService), any(), eq(TIMEOUT), any());
+
+        underTest.validate(mockInferenceService, customElandEmbeddingModel, TIMEOUT, mockActionListener);
+
+        verify(mockServiceIntegrationValidator).validate(eq(mockInferenceService), any(), eq(TIMEOUT), any());
+        verify(mockActionListener).delegateFailureAndWrap(any());
+        verify(mockActionListener).onFailure(any(ElasticsearchStatusException.class));
+        verify(mockInferenceServiceResults).getFirstEmbeddingSize();
+        verifyNoMoreInteractions(
+            mockServiceIntegrationValidator,
+            mockInferenceService,
+            mockCustomElandEmbeddingModel,
+            mockActionListener,
+            mockInferenceServiceResults
+        );
+    }
+
     private CustomElandEmbeddingModel createCustomElandEmbeddingModel(boolean areDimensionsSetByUser, Integer dimensions) {
         var mockServiceSettings = mock(CustomElandInternalTextEmbeddingServiceSettings.class);
         when(mockServiceSettings.modelId()).thenReturn(randomAlphaOfLength(10));