fix(homepage): make entity counts execute in parallel and make cache …

…configurable (datahub-project#7249)
looppi · Feb 15, 2023 · 2b3c8fb · 2b3c8fb
1 parent 74c18a9
commit 2b3c8fb
Show file tree

Hide file tree

Showing 16 changed files with 125 additions and 27 deletions.
diff --git a/metadata-io/src/main/java/com/linkedin/metadata/config/CacheConfiguration.java b/metadata-io/src/main/java/com/linkedin/metadata/config/CacheConfiguration.java
@@ -0,0 +1,10 @@
+package com.linkedin.metadata.config;
+
+import lombok.Data;
+
+
+@Data
+public class CacheConfiguration {
+  PrimaryCacheConfiguration primary;
+  HomepageCacheConfiguration homepage;
+}
diff --git a/metadata-io/src/main/java/com/linkedin/metadata/config/EntityDocCountCacheConfiguration.java b/metadata-io/src/main/java/com/linkedin/metadata/config/EntityDocCountCacheConfiguration.java
@@ -0,0 +1,9 @@
+package com.linkedin.metadata.config;
+
+import lombok.Data;
+
+
+@Data
+public class EntityDocCountCacheConfiguration {
+  long ttlSeconds;
+}
diff --git a/metadata-io/src/main/java/com/linkedin/metadata/config/HomepageCacheConfiguration.java b/metadata-io/src/main/java/com/linkedin/metadata/config/HomepageCacheConfiguration.java
@@ -0,0 +1,9 @@
+package com.linkedin.metadata.config;
+
+import lombok.Data;
+
+
+@Data
+public class HomepageCacheConfiguration {
+  EntityDocCountCacheConfiguration entityCounts;
+}
diff --git a/metadata-io/src/main/java/com/linkedin/metadata/config/PrimaryCacheConfiguration.java b/metadata-io/src/main/java/com/linkedin/metadata/config/PrimaryCacheConfiguration.java
@@ -0,0 +1,10 @@
+package com.linkedin.metadata.config;
+
+import lombok.Data;
+
+
+@Data
+public class PrimaryCacheConfiguration {
+  long ttlSeconds;
+  long maxSize;
+}
diff --git a/...io/src/main/java/com/linkedin/metadata/search/aggregator/AllEntitiesSearchAggregator.java b/...io/src/main/java/com/linkedin/metadata/search/aggregator/AllEntitiesSearchAggregator.java
@@ -3,6 +3,7 @@
 import com.codahale.metrics.Timer;
 import com.linkedin.data.template.GetMode;
 import com.linkedin.data.template.LongMap;
+import com.linkedin.metadata.config.EntityDocCountCacheConfiguration;
 import com.linkedin.metadata.models.registry.EntityRegistry;
 import com.linkedin.metadata.query.SearchFlags;
 import com.linkedin.metadata.query.filter.Filter;
@@ -54,11 +55,12 @@ public AllEntitiesSearchAggregator(
       EntityRegistry entityRegistry,
       EntitySearchService entitySearchService,
       CachingEntitySearchService cachingEntitySearchService,
-      SearchRanker searchRanker) {
+      SearchRanker searchRanker,
+      EntityDocCountCacheConfiguration entityDocCountCacheConfiguration) {
     _entitySearchService = Objects.requireNonNull(entitySearchService);
     _searchRanker = Objects.requireNonNull(searchRanker);
     _cachingEntitySearchService = Objects.requireNonNull(cachingEntitySearchService);
-    _entityDocCountCache = new EntityDocCountCache(entityRegistry, entitySearchService);
+    _entityDocCountCache = new EntityDocCountCache(entityRegistry, entitySearchService, entityDocCountCacheConfiguration);
     _maxAggregationValueCount = DEFAULT_MAX_AGGREGATION_VALUES; // TODO: Make this externally configurable
   }
 

diff --git a/metadata-io/src/main/java/com/linkedin/metadata/search/cache/EntityDocCountCache.java b/metadata-io/src/main/java/com/linkedin/metadata/search/cache/EntityDocCountCache.java
@@ -1,8 +1,10 @@
 package com.linkedin.metadata.search.cache;
 
 import com.google.common.base.Suppliers;
+import com.linkedin.metadata.config.EntityDocCountCacheConfiguration;
 import com.linkedin.metadata.models.registry.EntityRegistry;
 import com.linkedin.metadata.search.EntitySearchService;
+import com.linkedin.metadata.utils.ConcurrencyUtils;
 import io.opentelemetry.extension.annotations.WithSpan;
 import java.util.List;
 import java.util.Map;
@@ -17,17 +19,18 @@ public class EntityDocCountCache {
   private final EntitySearchService _entitySearchService;
   private final Supplier<Map<String, Long>> entityDocCount;
 
-  public EntityDocCountCache(EntityRegistry entityRegistry, EntitySearchService entitySearchService) {
+  public EntityDocCountCache(EntityRegistry entityRegistry, EntitySearchService entitySearchService,
+      EntityDocCountCacheConfiguration config) {
     _entityRegistry = entityRegistry;
     _entitySearchService = entitySearchService;
-    entityDocCount = Suppliers.memoizeWithExpiration(this::fetchEntityDocCount, 1, TimeUnit.MINUTES);
+    entityDocCount = Suppliers.memoizeWithExpiration(this::fetchEntityDocCount, config.getTtlSeconds(), TimeUnit.SECONDS);
   }
 
   private Map<String, Long> fetchEntityDocCount() {
-    return _entityRegistry.getEntitySpecs()
-        .keySet()
-        .stream()
-        .collect(Collectors.toMap(Function.identity(), _entitySearchService::docCount));
+    return ConcurrencyUtils
+        .transformAndCollectAsync(_entityRegistry.getEntitySpecs().keySet(),
+            Function.identity(),
+            Collectors.toMap(Function.identity(), _entitySearchService::docCount));
   }
 
   @WithSpan

diff --git a/metadata-io/src/test/java/com/linkedin/metadata/ESSampleDataFixture.java b/metadata-io/src/test/java/com/linkedin/metadata/ESSampleDataFixture.java
@@ -3,6 +3,7 @@
 import com.linkedin.entity.client.EntityClient;
 import com.linkedin.metadata.client.JavaEntityClient;
 import com.linkedin.metadata.config.ElasticSearchConfiguration;
+import com.linkedin.metadata.config.EntityDocCountCacheConfiguration;
 import com.linkedin.metadata.entity.EntityService;
 import com.linkedin.metadata.models.registry.EntityRegistry;
 import com.linkedin.metadata.search.SearchService;
@@ -102,9 +103,11 @@ protected SearchService searchService(
         int batchSize = 100;
         SearchRanker<Double> ranker = new SimpleRanker();
         CacheManager cacheManager = new ConcurrentMapCacheManager();
+        EntityDocCountCacheConfiguration entityDocCountCacheConfiguration = new EntityDocCountCacheConfiguration();
+        entityDocCountCacheConfiguration.setTtlSeconds(600L);
 
         SearchService service = new SearchService(
-                new EntityDocCountCache(entityRegistry, entitySearchService),
+                new EntityDocCountCache(entityRegistry, entitySearchService, entityDocCountCacheConfiguration),
                 new CachingEntitySearchService(
                         cacheManager,
                         entitySearchService,
@@ -122,7 +125,8 @@ protected SearchService searchService(
                                         batchSize,
                                         false
                                 ),
-                                ranker
+                                ranker,
+                                entityDocCountCacheConfiguration
                         ),
                         batchSize,
                         false

diff --git a/metadata-io/src/test/java/com/linkedin/metadata/ESSearchLineageFixture.java b/metadata-io/src/test/java/com/linkedin/metadata/ESSearchLineageFixture.java
@@ -3,6 +3,7 @@
 import com.linkedin.entity.client.EntityClient;
 import com.linkedin.metadata.client.JavaEntityClient;
 import com.linkedin.metadata.config.ElasticSearchConfiguration;
+import com.linkedin.metadata.config.EntityDocCountCacheConfiguration;
 import com.linkedin.metadata.entity.EntityService;
 import com.linkedin.metadata.graph.elastic.ESGraphQueryDAO;
 import com.linkedin.metadata.graph.elastic.ESGraphWriteDAO;
@@ -149,9 +150,11 @@ protected SearchService searchService(
         int batchSize = 100;
         SearchRanker<Double> ranker = new SimpleRanker();
         CacheManager cacheManager = new ConcurrentMapCacheManager();
+        EntityDocCountCacheConfiguration entityDocCountCacheConfiguration = new EntityDocCountCacheConfiguration();
+        entityDocCountCacheConfiguration.setTtlSeconds(600L);
 
         SearchService service = new SearchService(
-                new EntityDocCountCache(entityRegistry, entitySearchService),
+                new EntityDocCountCache(entityRegistry, entitySearchService, entityDocCountCacheConfiguration),
                 new CachingEntitySearchService(
                         cacheManager,
                         entitySearchService,
@@ -169,7 +172,8 @@ protected SearchService searchService(
                                         batchSize,
                                         false
                                 ),
-                                ranker
+                                ranker,
+                                entityDocCountCacheConfiguration
                         ),
                         batchSize,
                         false

diff --git a/metadata-io/src/test/java/com/linkedin/metadata/search/LineageSearchServiceTest.java b/metadata-io/src/test/java/com/linkedin/metadata/search/LineageSearchServiceTest.java
@@ -9,6 +9,7 @@
 import com.linkedin.data.schema.annotation.PathSpecBasedSchemaAnnotationVisitor;
 import com.linkedin.metadata.ESTestConfiguration;
 import com.linkedin.metadata.TestEntityUtil;
+import com.linkedin.metadata.config.EntityDocCountCacheConfiguration;
 import com.linkedin.metadata.graph.EntityLineageResult;
 import com.linkedin.metadata.graph.GraphService;
 import com.linkedin.metadata.graph.LineageDirection;
@@ -91,13 +92,16 @@ public void setup() {
 
   private void resetService(boolean withCache) {
     CachingEntitySearchService cachingEntitySearchService = new CachingEntitySearchService(_cacheManager, _elasticSearchService, 100, true);
+    EntityDocCountCacheConfiguration entityDocCountCacheConfiguration = new EntityDocCountCacheConfiguration();
+    entityDocCountCacheConfiguration.setTtlSeconds(600L);
     _lineageSearchService = new LineageSearchService(
         new SearchService(
-            new EntityDocCountCache(_entityRegistry, _elasticSearchService),
+            new EntityDocCountCache(_entityRegistry, _elasticSearchService, entityDocCountCacheConfiguration),
             cachingEntitySearchService,
             new CachingAllEntitiesSearchAggregator(
                 _cacheManager,
-                new AllEntitiesSearchAggregator(_entityRegistry, _elasticSearchService, cachingEntitySearchService,  new SimpleRanker()),
+                new AllEntitiesSearchAggregator(_entityRegistry, _elasticSearchService, cachingEntitySearchService,
+                    new SimpleRanker(), entityDocCountCacheConfiguration),
                 100,
                 true),
             new SimpleRanker()),

diff --git a/metadata-io/src/test/java/com/linkedin/metadata/search/SearchServiceTest.java b/metadata-io/src/test/java/com/linkedin/metadata/search/SearchServiceTest.java
@@ -8,6 +8,7 @@
 import com.linkedin.common.urn.Urn;
 import com.linkedin.data.template.StringArray;
 import com.linkedin.metadata.ESTestConfiguration;
+import com.linkedin.metadata.config.EntityDocCountCacheConfiguration;
 import com.linkedin.metadata.models.registry.EntityRegistry;
 import com.linkedin.metadata.models.registry.SnapshotEntityRegistry;
 import com.linkedin.metadata.query.SearchFlags;
@@ -82,16 +83,19 @@ private void resetSearchService() {
         _elasticSearchService,
         100,
         true);
+
+    EntityDocCountCacheConfiguration entityDocCountCacheConfiguration = new EntityDocCountCacheConfiguration();
+    entityDocCountCacheConfiguration.setTtlSeconds(600L);
     _searchService = new SearchService(
-      new EntityDocCountCache(_entityRegistry, _elasticSearchService),
+      new EntityDocCountCache(_entityRegistry, _elasticSearchService, entityDocCountCacheConfiguration),
       cachingEntitySearchService,
       new CachingAllEntitiesSearchAggregator(
           _cacheManager,
           new AllEntitiesSearchAggregator(
               _entityRegistry,
               _elasticSearchService,
               cachingEntitySearchService,
-              new SimpleRanker()),
+              new SimpleRanker(), entityDocCountCacheConfiguration),
           100,
           true),
       new SimpleRanker());
@@ -161,6 +165,9 @@ public void testSearchService() throws Exception {
     assertEquals(searchResult.getEntities().get(0).getEntity(), urn2);
     clearCache();
 
+    long docCount = _elasticSearchService.docCount(ENTITY_NAME);
+    assertEquals(docCount, 2L);
+
     _elasticSearchService.deleteDocument(ENTITY_NAME, urn.toString());
     _elasticSearchService.deleteDocument(ENTITY_NAME, urn2.toString());
     syncAfterWrite(_bulkProcessor);

diff --git a/metadata-service/factories/src/main/java/com/linkedin/gms/factory/common/CacheConfig.java b/metadata-service/factories/src/main/java/com/linkedin/gms/factory/common/CacheConfig.java
@@ -21,10 +21,10 @@
 @Configuration
 public class CacheConfig {
 
-  @Value("${CACHE_TTL_SECONDS:600}")
+  @Value("${cache.primary.ttlSeconds:600}")
   private int cacheTtlSeconds;
 
-  @Value("${CACHE_MAX_SIZE:10000}")
+  @Value("${cache.primary.maxSize:10000}")
   private int cacheMaxSize;
 
   @Value("${searchService.cache.hazelcast.serviceName:hazelcast-service}")

diff --git a/...ervice/factories/src/main/java/com/linkedin/gms/factory/config/ConfigurationProvider.java b/...ervice/factories/src/main/java/com/linkedin/gms/factory/config/ConfigurationProvider.java
@@ -3,6 +3,7 @@
 import com.datahub.authentication.AuthenticationConfiguration;
 import com.datahub.authorization.AuthorizationConfiguration;
 import com.linkedin.datahub.graphql.featureflags.FeatureFlags;
+import com.linkedin.metadata.config.CacheConfiguration;
 import com.linkedin.metadata.config.DataHubConfiguration;
 import com.linkedin.metadata.config.ElasticSearchConfiguration;
 import com.linkedin.metadata.config.IngestionConfiguration;
@@ -69,4 +70,9 @@ public class ConfigurationProvider {
    * System Update configurations
    */
   private SystemUpdateConfiguration systemUpdate;
+
+  /**
+   * Configuration for caching
+   */
+  private CacheConfiguration cache;
 }
diff --git a/...ies/src/main/java/com/linkedin/gms/factory/search/AllEntitiesSearchAggregatorFactory.java b/...ies/src/main/java/com/linkedin/gms/factory/search/AllEntitiesSearchAggregatorFactory.java
@@ -1,5 +1,6 @@
 package com.linkedin.gms.factory.search;
 
+import com.linkedin.gms.factory.config.ConfigurationProvider;
 import com.linkedin.gms.factory.spring.YamlPropertySourceFactory;
 import com.linkedin.metadata.models.registry.EntityRegistry;
 import com.linkedin.metadata.search.EntitySearchService;
@@ -38,11 +39,12 @@ public class AllEntitiesSearchAggregatorFactory {
   @Bean(name = "allEntitiesSearchAggregator")
   @Primary
   @Nonnull
-  protected AllEntitiesSearchAggregator getInstance() {
+  protected AllEntitiesSearchAggregator getInstance(ConfigurationProvider configurationProvider) {
     return new AllEntitiesSearchAggregator(
         entityRegistry,
         entitySearchService,
         cachingEntitySearchService,
-        searchRanker);
+        searchRanker,
+        configurationProvider.getCache().getHomepage().getEntityCounts());
   }
 }
diff --git a/...service/factories/src/main/java/com/linkedin/gms/factory/search/SearchServiceFactory.java b/...service/factories/src/main/java/com/linkedin/gms/factory/search/SearchServiceFactory.java
@@ -1,5 +1,6 @@
 package com.linkedin.gms.factory.search;
 
+import com.linkedin.gms.factory.config.ConfigurationProvider;
 import com.linkedin.gms.factory.spring.YamlPropertySourceFactory;
 import com.linkedin.metadata.models.registry.EntityRegistry;
 import com.linkedin.metadata.search.EntitySearchService;
@@ -44,9 +45,10 @@ public class SearchServiceFactory {
   @Bean(name = "searchService")
   @Primary
   @Nonnull
-  protected SearchService getInstance() {
+  protected SearchService getInstance(ConfigurationProvider configurationProvider) {
     return new SearchService(
-        new EntityDocCountCache(entityRegistry, entitySearchService),
+        new EntityDocCountCache(entityRegistry, entitySearchService, configurationProvider.getCache()
+            .getHomepage().getEntityCounts()),
         cachingEntitySearchService,
         cachingAllEntitiesSearchAggregator,
         searchRanker);

diff --git a/metadata-service/factories/src/main/resources/application.yml b/metadata-service/factories/src/main/resources/application.yml
@@ -255,3 +255,11 @@ entityClient:
 usageClient:
   retryInterval: ${USAGE_CLIENT_RETRY_INTERVAL:2}
   numRetries: ${USAGE_CLIENT_NUM_RETRIES:3}
+
+cache:
+  primary:
+    ttlSeconds: ${CACHE_TTL_SECONDS:600}
+    maxSize: ${CACHE_MAX_SIZE:10000}
+  homepage:
+    entityCounts:
+      ttlSeconds: ${CACHE_ENTITY_COUNTS_TTL_SECONDS:600}
diff --git a/metadata-utils/src/main/java/com/linkedin/metadata/utils/ConcurrencyUtils.java b/metadata-utils/src/main/java/com/linkedin/metadata/utils/ConcurrencyUtils.java
@@ -1,11 +1,13 @@
 package com.linkedin.metadata.utils;
 
+import java.util.Collection;
 import java.util.List;
 import java.util.Objects;
 import java.util.concurrent.CompletableFuture;
 import java.util.concurrent.TimeUnit;
 import java.util.function.BiFunction;
 import java.util.function.Function;
+import java.util.stream.Collector;
 import java.util.stream.Collectors;
 import lombok.extern.slf4j.Slf4j;
 
@@ -15,32 +17,48 @@ public class ConcurrencyUtils {
   private ConcurrencyUtils() {
   }
 
+  public static <O, T> List<T> transformAndCollectAsync(List<O> originalList, Function<O, T> transformer) {
+    return transformAndCollectAsync(originalList, transformer, Collectors.toList());
+  }
+
   /**
    * Transforms original list into the final list using the function transformer in an asynchronous fashion
    * i.e. each element transform is run as a separate CompleteableFuture and then joined at the end
    */
-  public static <O, T> List<T> transformAndCollectAsync(List<O> originalList, Function<O, T> transformer) {
-    return originalList.stream()
+  public static <O, T, OUTPUT> OUTPUT transformAndCollectAsync(Collection<O> originalCollection,
+      Function<O, T> transformer, Collector<T, ?, OUTPUT> collector) {
+    return originalCollection.stream()
         .map(element -> CompletableFuture.supplyAsync(() -> transformer.apply(element)))
         .collect(Collectors.collectingAndThen(Collectors.toList(),
             completableFutureList -> completableFutureList.stream().map(CompletableFuture::join)))
-        .collect(Collectors.toList());
+        .collect(collector);
   }
 
+
   /**
    * Transforms original list into the final list using the function transformer in an asynchronous fashion
    * with exceptions handled by the input exceptionHandler
    * i.e. each element transform is run as a separate CompleteableFuture and then joined at the end
    */
   public static <O, T> List<T> transformAndCollectAsync(List<O> originalList, Function<O, T> transformer,
       BiFunction<O, Throwable, ? extends T> exceptionHandler) {
-    return originalList.stream()
+    return transformAndCollectAsync(originalList, transformer, exceptionHandler, Collectors.toList());
+  }
+
+  /**
+   * Transforms original list into the final list using the function transformer in an asynchronous fashion
+   * with exceptions handled by the input exceptionHandler
+   * i.e. each element transform is run as a separate CompleteableFuture and then joined at the end
+   */
+  public static <O, T, OUTPUT> OUTPUT transformAndCollectAsync(Collection<O> originalCollection,
+      Function<O, T> transformer, BiFunction<O, Throwable, ? extends T> exceptionHandler, Collector<T, ?, OUTPUT> collector) {
+    return originalCollection.stream()
         .map(element -> CompletableFuture.supplyAsync(() -> transformer.apply(element))
             .exceptionally(e -> exceptionHandler.apply(element, e)))
         .filter(Objects::nonNull)
         .collect(Collectors.collectingAndThen(Collectors.toList(),
             completableFutureList -> completableFutureList.stream().map(CompletableFuture::join)))
-        .collect(Collectors.toList());
+        .collect(collector);
   }
 
   /**