lirui-apache commented on a change in pull request #13963: URL: https://github.com/apache/flink/pull/13963#discussion_r519256075
########## File path: flink-connectors/flink-connector-hive/src/main/java/org/apache/flink/connectors/hive/ContinuousHiveSplitEnumerator.java ########## @@ -0,0 +1,230 @@ +/* + * 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.flink.connectors.hive; + +import org.apache.flink.api.connector.source.SourceEvent; +import org.apache.flink.api.connector.source.SplitEnumerator; +import org.apache.flink.api.connector.source.SplitEnumeratorContext; +import org.apache.flink.api.connector.source.event.RequestSplitEvent; +import org.apache.flink.api.java.tuple.Tuple2; +import org.apache.flink.connector.file.src.FileSourceSplit; +import org.apache.flink.connector.file.src.PendingSplitsCheckpoint; +import org.apache.flink.connector.file.src.assigners.FileSplitAssigner; +import org.apache.flink.connectors.hive.read.HiveSourceSplit; +import org.apache.flink.table.catalog.ObjectPath; +import org.apache.flink.table.filesystem.ContinuousPartitionFetcher; + +import org.apache.hadoop.hive.metastore.api.Partition; +import org.apache.hadoop.mapred.JobConf; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +import java.io.IOException; +import java.util.ArrayList; +import java.util.Collection; +import java.util.Collections; +import java.util.Comparator; +import java.util.HashSet; +import java.util.Iterator; +import java.util.LinkedHashMap; +import java.util.List; +import java.util.Map; +import java.util.Optional; +import java.util.Set; +import java.util.concurrent.locks.ReadWriteLock; +import java.util.concurrent.locks.ReentrantReadWriteLock; + +/** + * A continuously monitoring {@link SplitEnumerator} for hive source. + */ +public class ContinuousHiveSplitEnumerator<T extends Comparable<T>> implements SplitEnumerator<HiveSourceSplit, PendingSplitsCheckpoint<HiveSourceSplit>> { + + private static final Logger LOG = LoggerFactory.getLogger(ContinuousHiveSplitEnumerator.class); + + private final SplitEnumeratorContext<HiveSourceSplit> enumeratorContext; + private final LinkedHashMap<Integer, String> readersAwaitingSplit; + private final FileSplitAssigner splitAssigner; + private final long discoveryInterval; + + private final JobConf jobConf; + private final ObjectPath tablePath; + + private final ContinuousPartitionFetcher<Partition, T> fetcher; + private final HiveTableSource.HiveContinuousPartitionFetcherContext<T> fetcherContext; + + private final ReadWriteLock stateLock = new ReentrantReadWriteLock(); + // the maximum partition read offset seen so far + private volatile T currentReadOffset; + // the partitions that have been processed for a given read offset + private final Set<List<String>> seenPartitions; + + public ContinuousHiveSplitEnumerator( + SplitEnumeratorContext<HiveSourceSplit> enumeratorContext, + T currentReadOffset, + Collection<List<String>> seenPartitions, + FileSplitAssigner splitAssigner, + long discoveryInterval, + JobConf jobConf, + ObjectPath tablePath, + ContinuousPartitionFetcher<Partition, T> fetcher, + HiveTableSource.HiveContinuousPartitionFetcherContext<T> fetcherContext) { + this.enumeratorContext = enumeratorContext; + this.currentReadOffset = currentReadOffset; + this.seenPartitions = new HashSet<>(seenPartitions); + this.splitAssigner = splitAssigner; + this.discoveryInterval = discoveryInterval; + this.jobConf = jobConf; + this.tablePath = tablePath; + this.fetcher = fetcher; + this.fetcherContext = fetcherContext; + readersAwaitingSplit = new LinkedHashMap<>(); + } + + @Override + public void start() { + try { + fetcherContext.open(); + enumeratorContext.callAsync( + this::monitorAndGetSplits, + this::handleNewSplits, + discoveryInterval, + discoveryInterval); + } catch (Exception e) { + throw new FlinkHiveException("Failed to start continuous split enumerator", e); + } + } + + @Override + public void handleSourceEvent(int subtaskId, SourceEvent sourceEvent) { + if (sourceEvent instanceof RequestSplitEvent) { + readersAwaitingSplit.put(subtaskId, ((RequestSplitEvent) sourceEvent).hostName()); + assignSplits(); + } else { + LOG.error("Received unrecognized event: {}", sourceEvent); + } + } + + @Override + public void addSplitsBack(List<HiveSourceSplit> splits, int subtaskId) { + LOG.debug("Continuous Hive Source Enumerator adds splits back: {}", splits); + stateLock.writeLock().lock(); + try { + splitAssigner.addSplits(new ArrayList<>(splits)); + } finally { + stateLock.writeLock().unlock(); + } + } + + @Override + public void addReader(int subtaskId) { + // this source is purely lazy-pull-based, nothing to do upon registration + } + + @Override + public PendingSplitsCheckpoint<HiveSourceSplit> snapshotState() throws Exception { + stateLock.readLock().lock(); + try { + Collection<HiveSourceSplit> remainingSplits = (Collection<HiveSourceSplit>) (Collection<?>) splitAssigner.remainingSplits(); + return new ContinuousHivePendingSplitsCheckpoint(remainingSplits, currentReadOffset, seenPartitions); + } finally { + stateLock.readLock().unlock(); + } + } + + @Override + public void close() throws IOException { + try { + fetcherContext.close(); + } catch (Exception e) { + throw new IOException(e); + } + } + + private Void monitorAndGetSplits() throws Exception { + stateLock.writeLock().lock(); + try { + List<Tuple2<Partition, T>> partitions = fetcher.fetchPartitions(fetcherContext, currentReadOffset); + if (partitions.isEmpty()) { + return null; + } + + partitions.sort(Comparator.comparing(o -> o.f1)); + List<HiveSourceSplit> newSplits = new ArrayList<>(); + // the max offset of new partitions + T maxOffset = currentReadOffset; + Set<List<String>> nextSeen = new HashSet<>(); + for (Tuple2<Partition, T> tuple2 : partitions) { + Partition partition = tuple2.f0; + List<String> partSpec = partition.getValues(); + if (seenPartitions.add(partSpec)) { + T offset = tuple2.f1; + if (offset.compareTo(currentReadOffset) > 0) { + nextSeen.add(partSpec); + } + if (offset.compareTo(maxOffset) > 0) { + maxOffset = offset; + } + LOG.info("Found new partition {} of table {}, generating splits for it", + partSpec, tablePath.getFullName()); + newSplits.addAll(HiveSourceFileEnumerator.createInputSplits( + 0, Collections.singletonList(fetcherContext.toHiveTablePartition(partition)), jobConf)); + } + } + currentReadOffset = maxOffset; + splitAssigner.addSplits(new ArrayList<>(newSplits)); + if (!nextSeen.isEmpty()) { + seenPartitions.clear(); Review comment: Renamed to `seenPartitionsSinceOffset`. Hopefully that makes more sense. ---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: [email protected]
