YuweiXiao commented on code in PR #4958: URL: https://github.com/apache/hudi/pull/4958#discussion_r917398587
########## hudi-client/hudi-spark-client/src/main/java/org/apache/hudi/client/clustering/update/strategy/SparkConsistentHashingDuplicateUpdateStrategy.java: ########## @@ -0,0 +1,146 @@ +/* + * 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.client.clustering.update.strategy; + +import org.apache.hudi.avro.model.HoodieClusteringGroup; +import org.apache.hudi.avro.model.HoodieClusteringPlan; +import org.apache.hudi.client.clustering.plan.strategy.SparkConsistentBucketClusteringPlanStrategy; +import org.apache.hudi.common.data.HoodieData; +import org.apache.hudi.common.engine.HoodieEngineContext; +import org.apache.hudi.common.fs.FSUtils; +import org.apache.hudi.common.model.ConsistentHashingNode; +import org.apache.hudi.common.model.HoodieAvroRecord; +import org.apache.hudi.common.model.HoodieConsistentHashingMetadata; +import org.apache.hudi.common.model.HoodieFileGroupId; +import org.apache.hudi.common.model.HoodieRecord; +import org.apache.hudi.common.model.HoodieRecordLocation; +import org.apache.hudi.common.model.HoodieRecordPayload; +import org.apache.hudi.common.table.timeline.HoodieInstant; +import org.apache.hudi.common.table.timeline.HoodieTimeline; +import org.apache.hudi.common.util.ClusteringUtils; +import org.apache.hudi.common.util.Option; +import org.apache.hudi.common.util.ValidationUtils; +import org.apache.hudi.common.util.collection.Pair; +import org.apache.hudi.exception.HoodieException; +import org.apache.hudi.index.HoodieIndexUtils; +import org.apache.hudi.index.bucket.ConsistentBucketIdentifier; +import org.apache.hudi.index.bucket.HoodieSparkConsistentBucketIndex; +import org.apache.hudi.table.HoodieTable; +import org.apache.hudi.table.action.cluster.strategy.UpdateStrategy; + +import org.apache.log4j.LogManager; +import org.apache.log4j.Logger; + +import java.util.Arrays; +import java.util.Collections; +import java.util.HashMap; +import java.util.HashSet; +import java.util.List; +import java.util.Map; +import java.util.Set; +import java.util.stream.Collectors; +import java.util.stream.Stream; + +/** + * Update strategy for (consistent hashing) bucket index + * If updates to file groups that are under clustering are identified, then generate + * two same records for each update, routing to both old and new file groups + */ +public class SparkConsistentHashingDuplicateUpdateStrategy<T extends HoodieRecordPayload<T>> extends UpdateStrategy<T, HoodieData<HoodieRecord<T>>> { + + private static final Logger LOG = LogManager.getLogger(SparkConsistentHashingDuplicateUpdateStrategy.class); + + public SparkConsistentHashingDuplicateUpdateStrategy(HoodieEngineContext engineContext, HoodieTable table, Set<HoodieFileGroupId> fileGroupsInPendingClustering) { + super(engineContext, table, fileGroupsInPendingClustering); + } + + @Override + public Pair<HoodieData<HoodieRecord<T>>, Set<HoodieFileGroupId>> handleUpdate(HoodieData<HoodieRecord<T>> taggedRecordsRDD) { + if (fileGroupsInPendingClustering.isEmpty()) { + return Pair.of(taggedRecordsRDD, Collections.emptySet()); + } + + HoodieData<HoodieRecord<T>> filteredRecordsRDD = taggedRecordsRDD.filter(r -> { + ValidationUtils.checkState(r.getCurrentLocation() != null); + return fileGroupsInPendingClustering.contains(new HoodieFileGroupId(r.getPartitionPath(), r.getCurrentLocation().getFileId())); + }); + + if (filteredRecordsRDD.count() == 0) { + return Pair.of(taggedRecordsRDD, Collections.emptySet()); + } + + // Read all pending/ongoing clustering plans + List<Pair<HoodieInstant, HoodieClusteringPlan>> instantPlanPairs = + table.getMetaClient().getActiveTimeline().filterInflightsAndRequested().filter(instant -> instant.getAction().equals(HoodieTimeline.REPLACE_COMMIT_ACTION)).getInstants() + .map(instant -> ClusteringUtils.getClusteringPlan(table.getMetaClient(), instant)) + .flatMap(o -> o.isPresent() ? Stream.of(o.get()) : Stream.empty()) + .collect(Collectors.toList()); + + // Construct child node for each partition & build the bucket identifier + final Set<String> partitions = new HashSet<>(filteredRecordsRDD.map(HoodieRecord::getPartitionPath).distinct().collectAsList()); + Map<String, HoodieConsistentHashingMetadata> partitionToHashingMeta = new HashMap<>(); + Map<String, String> partitionToInstant = new HashMap<>(); + for (Pair<HoodieInstant, HoodieClusteringPlan> pair : instantPlanPairs) { + String instant = pair.getLeft().getTimestamp(); + HoodieClusteringPlan plan = pair.getRight(); + extractHashingMetadataFromClusteringPlan(instant, plan, partitions, partitionToHashingMeta, partitionToInstant); + } + Map<String, ConsistentBucketIdentifier> partitionToIdentifier = partitionToHashingMeta.entrySet().stream() + .collect(Collectors.toMap(Map.Entry::getKey, e -> new ConsistentBucketIdentifier(e.getValue()))); + + // Produce records tagged with new record location + List<String> indexKeyFields = Arrays.asList(table.getConfig().getBucketIndexHashField().split(",")); + HoodieData<HoodieRecord<T>> redirectedRecordsRDD = filteredRecordsRDD.map(r -> { + ConsistentHashingNode node = partitionToIdentifier.get(r.getPartitionPath()).getBucket(r.getKey(), indexKeyFields); + return HoodieIndexUtils.getTaggedRecord(new HoodieAvroRecord(r.getKey(), r.getData(), r.getOperation()), Review Comment: `filteredRecordsRDD` is filtered based on the partitions of clustering plans. And `partitionToIdentifier` is initialized based on that. So I guess we may not need the check here, and it is also a hot code path and the checking will hurt the performance. And in the extreme case where null value happens, the execution will be stopped because of exception, which is also an expected behavior. -- 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]
