danny0405 commented on code in PR #18348: URL: https://github.com/apache/hudi/pull/18348#discussion_r3534764010
########## hudi-client/hudi-client-common/src/main/java/org/apache/hudi/metadata/index/record/PartitionedRecordIndexer.java: ########## @@ -0,0 +1,93 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.hudi.metadata.index.record; + +import org.apache.hudi.common.engine.HoodieEngineContext; +import org.apache.hudi.common.table.HoodieTableMetaClient; +import org.apache.hudi.common.util.Lazy; +import org.apache.hudi.common.util.ValidationUtils; +import org.apache.hudi.config.HoodieWriteConfig; +import org.apache.hudi.core.index.record.HoodieRecordIndex; +import org.apache.hudi.metadata.BucketizedMetadataTableFileGroupIndexParser; +import org.apache.hudi.metadata.MetadataPartitionType; +import org.apache.hudi.metadata.index.model.DataPartitionAndRecords; +import org.apache.hudi.metadata.index.model.IndexPartitionInitialization; +import org.apache.hudi.metadata.model.FileInfo; +import org.apache.hudi.metadata.model.FileSliceAndPartition; + +import lombok.extern.slf4j.Slf4j; + +import java.io.IOException; +import java.util.ArrayList; +import java.util.Collections; +import java.util.HashMap; +import java.util.List; +import java.util.Map; +import java.util.TreeMap; +import java.util.stream.Collectors; + +import static org.apache.hudi.metadata.HoodieTableMetadataUtil.createRecordIndexDefinition; +import static org.apache.hudi.metadata.MetadataPartitionType.RECORD_INDEX; + +/** + * Implementation of the partitioned {@link MetadataPartitionType#RECORD_INDEX} index + */ +@Slf4j +public class PartitionedRecordIndexer extends BaseRecordIndexer { + public PartitionedRecordIndexer(HoodieEngineContext engineContext, HoodieWriteConfig dataTableWriteConfig, + HoodieTableMetaClient dataTableMetaClient) { + super(engineContext, dataTableWriteConfig, dataTableMetaClient); + } + + @Override + public List<IndexPartitionInitialization> buildInitialization(String dataTableInstantTime, String instantTimeForPartition, Map<String, List<FileInfo>> partitionToAllFilesMap, + Lazy<List<FileSliceAndPartition>> lazyPartitionFileSlices) throws IOException { + createRecordIndexDefinition(dataTableMetaClient, Collections.singletonMap(HoodieRecordIndex.IS_PARTITIONED_OPTION, "true")); + Map<String, List<FileSliceAndPartition>> partitionFileSlicePairsMap = lazyPartitionFileSlices.get().stream() + .collect(Collectors.groupingBy(FileSliceAndPartition::partitionPath)); + Map<String, DataPartitionAndRecords> fileGroupCountAndRecordsPairMap = new HashMap<>(partitionFileSlicePairsMap.size()); + int maxParallelismPerHudiPartition = partitionFileSlicePairsMap.isEmpty() + ? 1 : Math.max(1, dataTableWriteConfig.getMetadataConfig().getRecordIndexMaxParallelism() / partitionFileSlicePairsMap.size()); + int totalFileGroupCount = 0; + for (Map.Entry<String, List<FileSliceAndPartition>> entry: partitionFileSlicePairsMap.entrySet()) { + String partition = entry.getKey(); + List<FileSliceAndPartition> fileSliceAndPartitions = entry.getValue(); + log.info("Initializing partitioned record index from data partition {}", partition); + DataPartitionAndRecords dataPartitionAndRecords = initializeRecordIndexPartition(partition, fileSliceAndPartitions, maxParallelismPerHudiPartition); + fileGroupCountAndRecordsPairMap.put(partition, dataPartitionAndRecords); + totalFileGroupCount += dataPartitionAndRecords.numFileGroups(); + } + log.info("Initializing partitioned record index with {} mappings", totalFileGroupCount); + + List<DataPartitionAndRecords> initializationList = new ArrayList<>(); + // Generate the file groups + TreeMap<String, Integer> partitionSizes = new TreeMap<>(); + for (Map.Entry<String, DataPartitionAndRecords> entry: fileGroupCountAndRecordsPairMap.entrySet()) { + String dataPartition = entry.getKey(); + DataPartitionAndRecords dataPartitionAndRecords = entry.getValue(); + ValidationUtils.checkArgument(dataPartitionAndRecords.numFileGroups() > 0, "FileGroup count for partitioned RLI data partition " + dataPartition + " should be > 0"); + partitionSizes.put(dataPartition, dataPartitionAndRecords.numFileGroups()); + initializationList.add(dataPartitionAndRecords); + } + + return Collections.singletonList(IndexPartitionInitialization.of(totalFileGroupCount, Review Comment: This returns an initialized record-index partition with `totalFileGroupCount == 0` when `lazyPartitionFileSlices` is empty. `initializeMetadataPartition` then skips `initializeFileGroups`, still bulk-commits empty data and marks `record_index` available. The next insert into a new data partition reaches `getRecordTaggerPartitionedRLI`, where no file slices exist for that partition, so tagging fails instead of creating the first RLI file group. Can we either skip initialization until there is at least one data partition, or create the configured min file groups for an empty/fresh table? ########## hudi-client/hudi-client-common/src/test/java/org/apache/hudi/metadata/index/record/TestPartitionedRecordIndexer.java: ########## @@ -0,0 +1,108 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. + */ + +package org.apache.hudi.metadata.index.record; + +import org.apache.hudi.common.config.HoodieMetadataConfig; +import org.apache.hudi.common.data.HoodieData; +import org.apache.hudi.common.engine.HoodieEngineContext; +import org.apache.hudi.common.engine.HoodieLocalEngineContext; +import org.apache.hudi.common.model.FileSlice; +import org.apache.hudi.common.model.HoodieRecord; +import org.apache.hudi.common.table.HoodieTableConfig; +import org.apache.hudi.common.table.HoodieTableMetaClient; +import org.apache.hudi.common.util.Lazy; +import org.apache.hudi.common.util.Option; +import org.apache.hudi.config.HoodieWriteConfig; +import org.apache.hudi.metadata.HoodieMetadataPayload; +import org.apache.hudi.metadata.HoodieTableMetadataUtil; +import org.apache.hudi.metadata.index.model.DataPartitionAndRecords; +import org.apache.hudi.metadata.index.model.IndexPartitionInitialization; +import org.apache.hudi.metadata.model.FileSliceAndPartition; + +import org.junit.jupiter.api.Test; +import org.mockito.MockedStatic; + +import java.io.IOException; +import java.util.Arrays; +import java.util.Collections; +import java.util.List; +import java.util.Map; + +import static org.apache.hudi.common.testutils.HoodieTestUtils.getDefaultStorageConf; +import static org.junit.jupiter.api.Assertions.assertEquals; +import static org.mockito.ArgumentMatchers.any; +import static org.mockito.Mockito.mock; +import static org.mockito.Mockito.mockStatic; +import static org.mockito.Mockito.when; + +class TestPartitionedRecordIndexer { + + @SuppressWarnings("unchecked") + @Test + void testInitializeWithRealEngineContextAndIndexDataContent() throws IOException { + HoodieEngineContext engineContext = new HoodieLocalEngineContext(getDefaultStorageConf()); + HoodieWriteConfig writeConfig = mock(HoodieWriteConfig.class); + HoodieMetadataConfig metadataConfig = mock(HoodieMetadataConfig.class); + HoodieTableMetaClient dataMetaClient = mock(HoodieTableMetaClient.class); + HoodieTableConfig tableConfig = mock(HoodieTableConfig.class); + + when(writeConfig.getMetadataConfig()).thenReturn(metadataConfig); + when(metadataConfig.getRecordIndexMaxParallelism()).thenReturn(8); + when(dataMetaClient.getTableConfig()).thenReturn(tableConfig); + + HoodieData<HoodieRecord> p1Data = (HoodieData<HoodieRecord>) (HoodieData<?>) engineContext.parallelize( + Collections.singletonList(HoodieMetadataPayload.createPartitionFilesRecord("p1_data", + Collections.singletonMap("f1.parquet", 1L), Collections.emptyList())), + 1); + HoodieData<HoodieRecord> p2Data = (HoodieData<HoodieRecord>) (HoodieData<?>) engineContext.parallelize( + Collections.singletonList(HoodieMetadataPayload.createPartitionFilesRecord("p2_data", + Collections.singletonMap("f2.parquet", 2L), Collections.emptyList())), + 1); + + DataPartitionAndRecords p1Init = new DataPartitionAndRecords(1, Option.of("p1"), p1Data); + DataPartitionAndRecords p2Init = new DataPartitionAndRecords(2, Option.of("p2"), p2Data); + + FileSliceAndPartition fs1 = FileSliceAndPartition.of("p1", new FileSlice("p1", "001", "f1")); + FileSliceAndPartition fs2 = FileSliceAndPartition.of("p2", new FileSlice("p2", "001", "f2")); + List<FileSliceAndPartition> input = Arrays.asList(fs1, fs2); + + ExposedPartitionedRecordIndexer indexer = new ExposedPartitionedRecordIndexer(engineContext, writeConfig, dataMetaClient, p1Init, p2Init); + + try (MockedStatic<HoodieTableMetadataUtil> mockedUtil = mockStatic(HoodieTableMetadataUtil.class)) { + List<IndexPartitionInitialization> initializationList = indexer.buildInitialization("001", "002", Map.of(), Lazy.lazily(() -> input)); Review Comment: Could we add an empty `lazyPartitionFileSlices` case here? The current test only covers non-empty input, so it misses the zero-file-group initialization path in `PartitionedRecordIndexer` where `totalFileGroupCount` and `dataPartitionAndRecords` both end up empty. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
