[ML] Require basic licence for the Elastic Inference Service

docs/changelog/137434.yaml

@@ -0,0 +1,5 @@
+pr: 137434
+summary: Require basic licence for the Elastic Inference Service
+area: Machine Learning
+type: enhancement
+issues: []

x-pack/plugin/core/src/main/java/org/elasticsearch/xpack/core/XPackField.java

@@ -32,6 +32,8 @@ public final class XPackField {
     public static final String UPGRADE = "upgrade";
     // inside of YAML settings we still use xpack do not having handle issues with dashes
     public static final String SETTINGS_NAME = "xpack";
+    /** Name constant for the EIS feature. */
+    public static final String ELASTIC_INFERENCE_SERVICE = "Elastic Inference Service";
     /** Name constant for the eql feature. */
     public static final String EQL = "eql";
     /** Name constant for the esql feature. */

x-pack/plugin/inference/qa/inference-service-tests/src/javaRestTest/java/org/elasticsearch/xpack/inference/BaseMockEISAuthServerTest.java

@@ -34,7 +34,7 @@ public class BaseMockEISAuthServerTest extends ESRestTestCase {
 
     private static ElasticsearchCluster cluster = ElasticsearchCluster.local()
         .distribution(DistributionType.DEFAULT)
-        .setting("xpack.license.self_generated.type", "trial")
+        .setting("xpack.license.self_generated.type", "basic")
         .setting("xpack.security.enabled", "true")
         // Adding both settings unless one feature flag is disabled in a particular environment
         .setting("xpack.inference.elastic.url", mockEISServer::getUrl)

x-pack/plugin/inference/qa/inference-service-tests/src/javaRestTest/java/org/elasticsearch/xpack/inference/InferenceBasicLicenseIT.java

@@ -45,28 +45,6 @@ public void testPutModel_RestrictedWithBasicLicense() throws Exception {
         sendRestrictedRequest("PUT", endpoint, modelConfig);
     }
 
-    public void testUpdateModel_RestrictedWithBasicLicense() throws Exception {
-        var endpoint = Strings.format("_inference/%s/%s/_update?error_trace", TaskType.SPARSE_EMBEDDING, "endpoint-id");
-        var requestBody = """
-            {
-              "task_settings": {
-                "num_threads": 2
-              }
-            }
-            """;
-        sendRestrictedRequest("PUT", endpoint, requestBody);
-    }
-
-    public void testPerformInference_RestrictedWithBasicLicense() throws Exception {
-        var endpoint = Strings.format("_inference/%s/%s?error_trace", TaskType.SPARSE_EMBEDDING, "endpoint-id");
-        var requestBody = """
-            {
-              "input": ["washing", "machine"]
-            }
-            """;
-        sendRestrictedRequest("POST", endpoint, requestBody);
-    }
-
     public void testGetServices_NonRestrictedWithBasicLicense() throws Exception {
         var endpoint = "_inference/_services";
         sendNonRestrictedRequest("GET", endpoint, null, 200, false);

x-pack/plugin/inference/src/internalClusterTest/java/org/elasticsearch/xpack/inference/action/filter/ShardBulkInferenceActionFilterBasicLicenseIT.java

@@ -46,6 +46,8 @@
 @ESTestCase.WithoutEntitlements // due to dependency issue ES-12435
 public class ShardBulkInferenceActionFilterBasicLicenseIT extends ESIntegTestCase {
     public static final String INDEX_NAME = "test-index";
+    private static final String SPARSE_INFERENCE_ID = "sparse-endpoint";
+    private static final String DENSE_INFERENCE_ID = "dense-endpoint";
 
     private final boolean useLegacyFormat;
 
@@ -61,9 +63,9 @@ public static Iterable<Object[]> parameters() {
     @Before
     public void setup() throws Exception {
         ModelRegistry modelRegistry = internalCluster().getCurrentMasterNodeInstance(ModelRegistry.class);
-        Utils.storeSparseModel("sparse-endpoint", modelRegistry);
+        Utils.storeSparseModel(SPARSE_INFERENCE_ID, modelRegistry);
         Utils.storeDenseModel(
-            "dense-endpoint",
+            DENSE_INFERENCE_ID,
             modelRegistry,
             randomIntBetween(1, 100),
             // dot product means that we need normalized vectors; it's not worth doing that in this test
@@ -92,27 +94,20 @@ public Settings indexSettings() {
     }
 
     public void testLicenseInvalidForInference() {
-        prepareCreate(INDEX_NAME).setMapping(
-            String.format(
-                Locale.ROOT,
-                """
-                    {
-                        "properties": {
-                            "sparse_field": {
-                                "type": "semantic_text",
-                                "inference_id": "%s"
-                            },
-                            "dense_field": {
-                                "type": "semantic_text",
-                                "inference_id": "%s"
-                            }
-                        }
+        prepareCreate(INDEX_NAME).setMapping(String.format(Locale.ROOT, """
+            {
+                "properties": {
+                    "sparse_field": {
+                        "type": "semantic_text",
+                        "inference_id": "%s"
+                    },
+                    "dense_field": {
+                        "type": "semantic_text",
+                        "inference_id": "%s"
                     }
-                    """,
-                TestSparseInferenceServiceExtension.TestInferenceService.NAME,
-                TestDenseInferenceServiceExtension.TestInferenceService.NAME
-            )
-        ).get();
+                }
+            }
+            """, SPARSE_INFERENCE_ID, DENSE_INFERENCE_ID)).get();
 
         BulkRequestBuilder bulkRequest = client().prepareBulk();
         int totalBulkReqs = randomIntBetween(2, 100);

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

@@ -0,0 +1,38 @@
+/*
+ * Copyright Elasticsearch B.V. and/or licensed to Elasticsearch B.V. under one
+ * or more contributor license agreements. Licensed under the Elastic License
+ * 2.0; you may not use this file except in compliance with the Elastic License
+ * 2.0.
+ */
+
+package org.elasticsearch.xpack.inference;
+
+import org.elasticsearch.ElasticsearchSecurityException;
+import org.elasticsearch.license.LicenseUtils;
+import org.elasticsearch.license.XPackLicenseState;
+import org.elasticsearch.xpack.core.XPackField;
+import org.elasticsearch.xpack.inference.services.elastic.ElasticInferenceService;
+
+import static org.elasticsearch.xpack.inference.InferencePlugin.EIS_INFERENCE_FEATURE;
+import static org.elasticsearch.xpack.inference.InferencePlugin.INFERENCE_API_FEATURE;
+
+public class InferenceLicenceCheck {
+
+    private InferenceLicenceCheck() {}
+
+    public static boolean isServiceLicenced(String serviceName, XPackLicenseState licenseState) {
+        if (ElasticInferenceService.NAME.equals(serviceName)) {
+            return EIS_INFERENCE_FEATURE.check(licenseState);
+        } else {
+            return INFERENCE_API_FEATURE.check(licenseState);
+        }
+    }
+
+    public static ElasticsearchSecurityException complianceException(String serviceName) {
+        if (ElasticInferenceService.NAME.equals(serviceName)) {
+            return LicenseUtils.newComplianceException(XPackField.ELASTIC_INFERENCE_SERVICE);
+        } else {
+            return LicenseUtils.newComplianceException(XPackField.INFERENCE);
+        }
+    }
+}

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

@@ -203,6 +203,12 @@ public class InferencePlugin extends Plugin
         License.OperationMode.ENTERPRISE
     );
 
+    public static final LicensedFeature.Momentary EIS_INFERENCE_FEATURE = LicensedFeature.momentary(
+        "inference",
+        "Elastic Inference Service",
+        License.OperationMode.BASIC
+    );
+
     public static final String X_ELASTIC_PRODUCT_USE_CASE_HTTP_HEADER = "X-elastic-product-use-case";
 
     public static final String NAME = "inference";

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

@@ -25,15 +25,14 @@
 import org.elasticsearch.inference.Model;
 import org.elasticsearch.inference.TaskType;
 import org.elasticsearch.inference.telemetry.InferenceStats;
-import org.elasticsearch.license.LicenseUtils;
 import org.elasticsearch.license.XPackLicenseState;
 import org.elasticsearch.rest.RestStatus;
 import org.elasticsearch.tasks.Task;
 import org.elasticsearch.threadpool.ThreadPool;
 import org.elasticsearch.transport.TransportService;
-import org.elasticsearch.xpack.core.XPackField;
 import org.elasticsearch.xpack.core.inference.action.BaseInferenceActionRequest;
 import org.elasticsearch.xpack.core.inference.action.InferenceAction;
+import org.elasticsearch.xpack.inference.InferenceLicenceCheck;
 import org.elasticsearch.xpack.inference.InferencePlugin;
 import org.elasticsearch.xpack.inference.action.task.StreamingTaskManager;
 import org.elasticsearch.xpack.inference.registry.InferenceEndpointRegistry;
@@ -51,7 +50,6 @@
 import static org.elasticsearch.core.Strings.format;
 import static org.elasticsearch.inference.telemetry.InferenceStats.responseAttributes;
 import static org.elasticsearch.inference.telemetry.InferenceStats.serviceAndResponseAttributes;
-import static org.elasticsearch.xpack.inference.InferencePlugin.INFERENCE_API_FEATURE;
 
 /**
  * Base class for transport actions that handle inference requests.
@@ -112,16 +110,17 @@ protected abstract void doInference(
 
     @Override
     protected void doExecute(Task task, Request request, ActionListener<InferenceAction.Response> listener) {
-        if (INFERENCE_API_FEATURE.check(licenseState) == false) {
-            listener.onFailure(LicenseUtils.newComplianceException(XPackField.INFERENCE));
-            return;
-        }
 
         var timer = InferenceTimer.start();
 
         var getModelListener = ActionListener.wrap((Model model) -> {
             var serviceName = model.getConfigurations().getService();
 
+            if (InferenceLicenceCheck.isServiceLicenced(serviceName, licenseState) == false) {
+                listener.onFailure(InferenceLicenceCheck.complianceException(serviceName));
+                return;
+            }
+
             try {
                 validateRequest(request, model);
             } catch (Exception e) {

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

@@ -14,7 +14,6 @@
 import org.elasticsearch.action.support.ActionFilters;
 import org.elasticsearch.action.support.master.TransportMasterNodeAction;
 import org.elasticsearch.client.internal.Client;
-import org.elasticsearch.client.internal.OriginSettingClient;
 import org.elasticsearch.cluster.ClusterState;
 import org.elasticsearch.cluster.block.ClusterBlockException;
 import org.elasticsearch.cluster.block.ClusterBlockLevel;
@@ -34,19 +33,18 @@
 import org.elasticsearch.inference.ModelConfigurations;
 import org.elasticsearch.inference.TaskType;
 import org.elasticsearch.injection.guice.Inject;
-import org.elasticsearch.license.LicenseUtils;
 import org.elasticsearch.license.XPackLicenseState;
 import org.elasticsearch.rest.RestStatus;
 import org.elasticsearch.tasks.Task;
 import org.elasticsearch.threadpool.ThreadPool;
 import org.elasticsearch.transport.TransportService;
 import org.elasticsearch.xcontent.XContentParser;
 import org.elasticsearch.xcontent.XContentParserConfiguration;
-import org.elasticsearch.xpack.core.XPackField;
 import org.elasticsearch.xpack.core.inference.action.PutInferenceModelAction;
 import org.elasticsearch.xpack.core.ml.inference.assignment.TrainedModelAssignmentUtils;
 import org.elasticsearch.xpack.core.ml.job.messages.Messages;
 import org.elasticsearch.xpack.core.ml.utils.ExceptionsHelper;
+import org.elasticsearch.xpack.inference.InferenceLicenceCheck;
 import org.elasticsearch.xpack.inference.InferencePlugin;
 import org.elasticsearch.xpack.inference.registry.ModelRegistry;
 import org.elasticsearch.xpack.inference.services.ServiceUtils;
@@ -60,8 +58,6 @@
 import java.util.Set;
 
 import static org.elasticsearch.core.Strings.format;
-import static org.elasticsearch.xpack.core.ClientHelper.INFERENCE_ORIGIN;
-import static org.elasticsearch.xpack.inference.InferencePlugin.INFERENCE_API_FEATURE;
 import static org.elasticsearch.xpack.inference.InferencePlugin.UTILITY_THREAD_POOL_NAME;
 import static org.elasticsearch.xpack.inference.common.SemanticTextInfoExtractor.getModelSettingsForIndicesReferencingInferenceEndpoints;
 import static org.elasticsearch.xpack.inference.mapper.SemanticTextFieldMapper.canMergeModelSettings;
@@ -76,7 +72,6 @@ public class TransportPutInferenceModelAction extends TransportMasterNodeAction<
     private final XPackLicenseState licenseState;
     private final ModelRegistry modelRegistry;
     private final InferenceServiceRegistry serviceRegistry;
-    private final OriginSettingClient client;
     private volatile boolean skipValidationAndStart;
     private final ProjectResolver projectResolver;
 
@@ -110,7 +105,6 @@ public TransportPutInferenceModelAction(
         clusterService.getClusterSettings()
             .addSettingsUpdateConsumer(InferencePlugin.SKIP_VALIDATE_AND_START, this::setSkipValidationAndStart);
         this.projectResolver = projectResolver;
-        this.client = new OriginSettingClient(client, INFERENCE_ORIGIN);
     }
 
     @Override
@@ -120,11 +114,6 @@ protected void masterOperation(
         ClusterState state,
         ActionListener<PutInferenceModelAction.Response> listener
     ) throws Exception {
-        if (INFERENCE_API_FEATURE.check(licenseState) == false) {
-            listener.onFailure(LicenseUtils.newComplianceException(XPackField.INFERENCE));
-            return;
-        }
-
         if (modelRegistry.containsDefaultConfigId(request.getInferenceEntityId())) {
             listener.onFailure(
                 new ElasticsearchStatusException(
@@ -150,6 +139,11 @@ protected void masterOperation(
             return;
         }
 
+        if (InferenceLicenceCheck.isServiceLicenced(serviceName, licenseState) == false) {
+            listener.onFailure(InferenceLicenceCheck.complianceException(serviceName));
+            return;
+        }
+
         if (List.of(OLD_ELSER_SERVICE_NAME, ElasticsearchInternalService.NAME).contains(serviceName)) {
             // required for BWC of elser service in elasticsearch service TODO remove when elser service deprecated
             requestAsMap.put(ModelConfigurations.SERVICE, serviceName);

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

@@ -35,21 +35,20 @@
 import org.elasticsearch.inference.TaskType;
 import org.elasticsearch.inference.UnparsedModel;
 import org.elasticsearch.injection.guice.Inject;
-import org.elasticsearch.license.LicenseUtils;
 import org.elasticsearch.license.XPackLicenseState;
 import org.elasticsearch.rest.RestStatus;
 import org.elasticsearch.tasks.Task;
 import org.elasticsearch.threadpool.ThreadPool;
 import org.elasticsearch.transport.TransportService;
 import org.elasticsearch.xcontent.XContentParser;
 import org.elasticsearch.xcontent.XContentParserConfiguration;
-import org.elasticsearch.xpack.core.XPackField;
 import org.elasticsearch.xpack.core.inference.action.UpdateInferenceModelAction;
 import org.elasticsearch.xpack.core.ml.action.CreateTrainedModelAssignmentAction;
 import org.elasticsearch.xpack.core.ml.action.UpdateTrainedModelDeploymentAction;
 import org.elasticsearch.xpack.core.ml.inference.assignment.TrainedModelAssignmentUtils;
 import org.elasticsearch.xpack.core.ml.job.messages.Messages;
 import org.elasticsearch.xpack.core.ml.utils.ExceptionsHelper;
+import org.elasticsearch.xpack.inference.InferenceLicenceCheck;
 import org.elasticsearch.xpack.inference.registry.ModelRegistry;
 import org.elasticsearch.xpack.inference.services.elasticsearch.ElasticsearchInternalModel;
 import org.elasticsearch.xpack.inference.services.elasticsearch.ElasticsearchInternalService;
@@ -61,7 +60,6 @@
 import java.util.Optional;
 import java.util.concurrent.atomic.AtomicReference;
 
-import static org.elasticsearch.xpack.inference.InferencePlugin.INFERENCE_API_FEATURE;
 import static org.elasticsearch.xpack.inference.services.ServiceUtils.resolveTaskType;
 import static org.elasticsearch.xpack.inference.services.elasticsearch.ElasticsearchInternalServiceSettings.NUM_ALLOCATIONS;
 
@@ -113,11 +111,6 @@ protected void masterOperation(
         ClusterState state,
         ActionListener<UpdateInferenceModelAction.Response> masterListener
     ) {
-        if (INFERENCE_API_FEATURE.check(licenseState) == false) {
-            masterListener.onFailure(LicenseUtils.newComplianceException(XPackField.INFERENCE));
-            return;
-        }
-
         var bodyTaskType = request.getContentAsSettings().taskType();
         var resolvedTaskType = resolveTaskType(request.getTaskType(), bodyTaskType != null ? bodyTaskType.toString() : null);
 
@@ -137,10 +130,16 @@ protected void masterOperation(
                             unparsedModel.service()
                         )
                     );
-                } else {
-                    service.set(optionalService.get());
-                    listener.onResponse(unparsedModel);
+                    return;
                 }
+
+                if (InferenceLicenceCheck.isServiceLicenced(optionalService.get().name(), licenseState) == false) {
+                    listener.onFailure(InferenceLicenceCheck.complianceException(optionalService.get().name()));
+                    return;
+                }
+
+                service.set(optionalService.get());
+                listener.onResponse(unparsedModel);
             })
             .<Boolean>andThen((listener, existingUnparsedModel) -> {