Replication job to trigger setup and carbon flow for replica tables [WIP]

apps/spark/src/main/java/com/linkedin/openhouse/jobs/client/TablesClient.java

@@ -4,6 +4,7 @@
 import com.linkedin.openhouse.datalayout.strategy.DataLayoutStrategy;
 import com.linkedin.openhouse.jobs.util.DatabaseTableFilter;
 import com.linkedin.openhouse.jobs.util.DirectoryMetadata;
+import com.linkedin.openhouse.jobs.util.ReplicationConfig;
 import com.linkedin.openhouse.jobs.util.RetentionConfig;
 import com.linkedin.openhouse.jobs.util.RetryUtil;
 import com.linkedin.openhouse.jobs.util.TableDataLayoutMetadata;
@@ -15,6 +16,7 @@
 import com.linkedin.openhouse.tables.client.model.GetDatabaseResponseBody;
 import com.linkedin.openhouse.tables.client.model.GetTableResponseBody;
 import com.linkedin.openhouse.tables.client.model.Policies;
+import com.linkedin.openhouse.tables.client.model.Replication;
 import java.time.Duration;
 import java.util.AbstractMap;
 import java.util.ArrayList;
@@ -56,6 +58,11 @@ public Optional<RetentionConfig> getTableRetention(TableMetadata tableMetadata)
     return getTableRetention(response);
   }
 
+  public Optional<List<ReplicationConfig>> getTableReplication(TableMetadata tableMetadata) {
+    GetTableResponseBody response = getTable(tableMetadata);
+    return getTableReplication(response);
+  }
+
   private Optional<RetentionConfig> getTableRetention(GetTableResponseBody response) {
     // timePartitionSpec or retention.ColumnPattern should be present to run Retention job on a
     // table.
@@ -86,6 +93,31 @@ private Optional<RetentionConfig> getTableRetention(GetTableResponseBody respons
             .build());
   }
 
+  private Optional<List<ReplicationConfig>> getTableReplication(GetTableResponseBody response) {
+    // At least one replication config must be present
+    if (response == null
+        || response.getPolicies() == null
+        || response.getPolicies().getReplication() == null
+        || response.getPolicies().getReplication().getConfig().size() <= 0) {
+      return Optional.empty();
+    }
+    List<ReplicationConfig> replicationConfigList = new ArrayList<>();
+    Replication replication = response.getPolicies().getReplication();
+    List<com.linkedin.openhouse.tables.client.model.ReplicationConfig> replicationConfig =
+        replication.getConfig();
+
+    replicationConfig.forEach(
+        rc ->
+            replicationConfigList.add(
+                ReplicationConfig.builder()
+                    .cluster(rc.getDestination())
+                    .tableOwner(response.getTableCreator())
+                    .schedule(rc.getCronSchedule())
+                    .build()));
+    // since replicationConfigList is initialized, it cannot be null.
+    return Optional.of(replicationConfigList);
+  }
+
   protected GetTableResponseBody getTable(TableMetadata tableMetadata) {
     return getTable(tableMetadata.getDbName(), tableMetadata.getTableName());
   }
@@ -281,6 +313,7 @@ protected Optional<TableMetadata> mapTableResponseToTableMetadata(
             .isTimePartitioned(tableResponseBody.getTimePartitioning() != null)
             .isClustered(tableResponseBody.getClustering() != null)
             .retentionConfig(getTableRetention(tableResponseBody).orElse(null))
+            .replicationConfig(getTableReplication(tableResponseBody).orElse(null))
             .jobExecutionProperties(getJobExecutionProperties(tableResponseBody));
     builder.creationTimeMs(Objects.requireNonNull(tableResponseBody.getCreationTime()));
     return Optional.of(builder.build());

apps/spark/src/main/java/com/linkedin/openhouse/jobs/scheduler/tasks/OperationTasksBuilder.java

@@ -48,6 +48,19 @@ private List<OperationTask<?>> prepareTableOperationTaskList(JobConf.JobTypeEnum
     return processMetadataList(tableMetadataList, jobType);
   }
 
+  private List<OperationTask<?>> prepareReplicationOperationTaskList(JobConf.JobTypeEnum jobType) {
+    List<TableMetadata> replicationSetupTableMetadataList = tablesClient.getTableMetadataList();
+    // filters tables which are primary and hava replication config defined
+    replicationSetupTableMetadataList =
+        replicationSetupTableMetadataList.stream()
+            .filter(m -> m.isPrimary() && (m.getReplicationConfig() != null))
+            .collect(Collectors.toList());
+    log.info(
+        "Fetched metadata for {} tables for replication setup task",
+        replicationSetupTableMetadataList.size());
+    return processMetadataList(replicationSetupTableMetadataList, jobType);
+  }
+
   private List<OperationTask<?>> prepareTableDirectoryOperationTaskList(
       JobConf.JobTypeEnum jobType) {
     List<DirectoryMetadata> directoryMetadataList = tablesClient.getOrphanTableDirectories();
@@ -152,6 +165,8 @@ public List<OperationTask<?>> buildOperationTaskList(
       case STAGED_FILES_DELETION:
       case DATA_LAYOUT_STRATEGY_GENERATION:
         return prepareTableOperationTaskList(jobType);
+      case REPLICATION:
+        return prepareReplicationOperationTaskList(jobType);
       case DATA_LAYOUT_STRATEGY_EXECUTION:
         return prepareDataLayoutOperationTaskList(jobType, properties, meter);
       case ORPHAN_DIRECTORY_DELETION:

apps/spark/src/main/java/com/linkedin/openhouse/jobs/util/ReplicationConfig.java

@@ -0,0 +1,17 @@
+package com.linkedin.openhouse.jobs.util;
+
+import lombok.Builder;
+import lombok.EqualsAndHashCode;
+import lombok.Getter;
+import lombok.ToString;
+
+/** Table retention config class. This is app side representation of /tables policies->retention */
+@Builder
+@Getter
+@EqualsAndHashCode
+@ToString
+public class ReplicationConfig {
+  private final String schedule;
+  private final String tableOwner;
+  private final String cluster;
+}

apps/spark/src/main/java/com/linkedin/openhouse/jobs/util/TableMetadata.java

@@ -1,6 +1,7 @@
 package com.linkedin.openhouse.jobs.util;
 
 import java.util.HashMap;
+import java.util.List;
 import java.util.Map;
 import javax.annotation.Nullable;
 import lombok.Builder;
@@ -25,6 +26,7 @@ public class TableMetadata extends Metadata {
   @Builder.Default protected @NonNull Map<String, String> jobExecutionProperties = new HashMap<>();
   protected @Nullable RetentionConfig retentionConfig;
   protected @Nullable HistoryConfig historyConfig;
+  protected @Nullable List<ReplicationConfig> replicationConfig;
 
   public String fqtn() {
     return String.format("%s.%s", dbName, tableName);

apps/spark/src/test/java/com/linkedin/openhouse/jobs/clients/TablesClientTest.java

@@ -18,6 +18,8 @@
 import com.linkedin.openhouse.tables.client.model.GetDatabaseResponseBody;
 import com.linkedin.openhouse.tables.client.model.GetTableResponseBody;
 import com.linkedin.openhouse.tables.client.model.Policies;
+import com.linkedin.openhouse.tables.client.model.Replication;
+import com.linkedin.openhouse.tables.client.model.ReplicationConfig;
 import com.linkedin.openhouse.tables.client.model.Retention;
 import com.linkedin.openhouse.tables.client.model.RetentionColumnPattern;
 import com.linkedin.openhouse.tables.client.model.TimePartitionSpec;
@@ -415,6 +417,33 @@ void testNonPartitionedTableWithPatternGetRetentionConfig() {
     Mockito.verify(apiMock, Mockito.times(1)).getTableV1(testDbName, testTableNamePartitioned);
   }
 
+  @Test
+  void testPrimaryTableWithReplicationConfig() {
+    GetTableResponseBody primaryTableWithReplicationConfigResponseBodyMock =
+        createPrimaryTableWithReplicationPolicyResponseBodyMock(
+            testDbName, testTableName, "schedule", "interval", "cluster");
+    Mono<GetTableResponseBody> responseMock = (Mono<GetTableResponseBody>) Mockito.mock(Mono.class);
+    Mockito.when(responseMock.block(any(Duration.class)))
+        .thenReturn(primaryTableWithReplicationConfigResponseBodyMock);
+    Mockito.when(apiMock.getTableV1(testDbName, testTableName)).thenReturn(responseMock);
+    Optional<List<com.linkedin.openhouse.jobs.util.ReplicationConfig>> result =
+        client.getTableReplication(
+            TableMetadata.builder().dbName(testDbName).tableName(testTableName).build());
+    Assertions.assertTrue(
+        result.isPresent(), "Retention config must be present for a test partitioned table");
+    List<com.linkedin.openhouse.jobs.util.ReplicationConfig> replicationConfigs = new ArrayList<>();
+    com.linkedin.openhouse.jobs.util.ReplicationConfig replicationConfig =
+        com.linkedin.openhouse.jobs.util.ReplicationConfig.builder()
+            .schedule("schedule")
+            .cluster("cluster")
+            .tableOwner("")
+            .build();
+    replicationConfigs.add(replicationConfig);
+    Assertions.assertEquals(replicationConfigs, result.orElse(null));
+    Mockito.verify(responseMock, Mockito.times(1)).block(any(Duration.class));
+    Mockito.verify(apiMock, Mockito.times(1)).getTableV1(testDbName, testTableName);
+  }
+
   @Test
   void getDatabases() {
     GetAllDatabasesResponseBody allDatabasesResponseBodyMock =
@@ -535,6 +564,24 @@ private GetTableResponseBody createNonPartitionedTableWithPatternResponseBodyMoc
     return setUpResponseBodyMock(dbName, tableName, null, policies);
   }
 
+  private GetTableResponseBody createPrimaryTableWithReplicationPolicyResponseBodyMock(
+      String dbName, String tableName, String schedule, String interval, String cluster) {
+    Policies policies = Mockito.mock(Policies.class);
+    Replication replication = Mockito.mock(Replication.class);
+    List<ReplicationConfig> replicationConfigs = new ArrayList<>();
+    ReplicationConfig replicationConfig = Mockito.mock(ReplicationConfig.class);
+    replicationConfigs.add(replicationConfig);
+    replication.setConfig(replicationConfigs);
+
+    policies.setReplication(replication);
+    Mockito.when(replication.getConfig()).thenReturn(replicationConfigs);
+    Mockito.when(policies.getReplication()).thenReturn(replication);
+    Mockito.when(replicationConfig.getCronSchedule()).thenReturn(schedule);
+    Mockito.when(replicationConfig.getDestination()).thenReturn(cluster);
+    Mockito.when(replicationConfig.getInterval()).thenReturn(interval);
+    return setUpResponseBodyMock(dbName, tableName, null, policies);
+  }
+
   private GetTableResponseBody createPartitionedTableNullPoliciesResponseBodyMock(
       String dbName, String tableName, String partitionColummName) {
     TimePartitionSpec partitionSpec = Mockito.mock(TimePartitionSpec.class);

services/jobs/src/main/java/com/linkedin/openhouse/jobs/model/JobConf.java

@@ -33,6 +33,7 @@ public enum JobType {
     ORPHAN_DIRECTORY_DELETION,
     TABLE_STATS_COLLECTION,
     DATA_LAYOUT_STRATEGY_GENERATION,
-    DATA_LAYOUT_STRATEGY_EXECUTION
+    DATA_LAYOUT_STRATEGY_EXECUTION,
+    REPLICATION
   }
 }