stevenzwu commented on code in PR #10179: URL: https://github.com/apache/iceberg/pull/10179#discussion_r1619419797
########## flink/v1.19/flink/src/main/java/org/apache/iceberg/flink/sink/IcebergSink.java: ########## @@ -0,0 +1,780 @@ +/* + * 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.iceberg.flink.sink; + +import static org.apache.iceberg.TableProperties.AVRO_COMPRESSION; +import static org.apache.iceberg.TableProperties.AVRO_COMPRESSION_LEVEL; +import static org.apache.iceberg.TableProperties.ORC_COMPRESSION; +import static org.apache.iceberg.TableProperties.ORC_COMPRESSION_STRATEGY; +import static org.apache.iceberg.TableProperties.PARQUET_COMPRESSION; +import static org.apache.iceberg.TableProperties.PARQUET_COMPRESSION_LEVEL; +import static org.apache.iceberg.TableProperties.WRITE_DISTRIBUTION_MODE; + +import java.io.IOException; +import java.io.UncheckedIOException; +import java.time.Duration; +import java.util.List; +import java.util.Map; +import java.util.Optional; +import java.util.Set; +import java.util.UUID; +import java.util.function.Function; +import org.apache.flink.api.common.functions.MapFunction; +import org.apache.flink.api.common.typeinfo.TypeInformation; +import org.apache.flink.api.connector.sink2.Committer; +import org.apache.flink.api.connector.sink2.CommitterInitContext; +import org.apache.flink.api.connector.sink2.Sink; +import org.apache.flink.api.connector.sink2.SinkWriter; +import org.apache.flink.api.connector.sink2.SupportsCommitter; +import org.apache.flink.configuration.Configuration; +import org.apache.flink.configuration.ReadableConfig; +import org.apache.flink.core.io.SimpleVersionedSerializer; +import org.apache.flink.streaming.api.connector.sink2.CommittableMessage; +import org.apache.flink.streaming.api.connector.sink2.CommittableMessageTypeInfo; +import org.apache.flink.streaming.api.connector.sink2.SupportsPostCommitTopology; +import org.apache.flink.streaming.api.connector.sink2.SupportsPreCommitTopology; +import org.apache.flink.streaming.api.connector.sink2.SupportsPreWriteTopology; +import org.apache.flink.streaming.api.datastream.DataStream; +import org.apache.flink.streaming.api.datastream.DataStreamSink; +import org.apache.flink.streaming.api.datastream.SingleOutputStreamOperator; +import org.apache.flink.table.api.TableSchema; +import org.apache.flink.table.data.RowData; +import org.apache.flink.table.data.util.DataFormatConverters; +import org.apache.flink.table.types.DataType; +import org.apache.flink.table.types.logical.RowType; +import org.apache.flink.types.Row; +import org.apache.flink.util.Preconditions; +import org.apache.iceberg.DataFile; +import org.apache.iceberg.DistributionMode; +import org.apache.iceberg.FileFormat; +import org.apache.iceberg.PartitionField; +import org.apache.iceberg.PartitionSpec; +import org.apache.iceberg.Schema; +import org.apache.iceberg.SerializableTable; +import org.apache.iceberg.Table; +import org.apache.iceberg.flink.FlinkSchemaUtil; +import org.apache.iceberg.flink.FlinkWriteConf; +import org.apache.iceberg.flink.FlinkWriteOptions; +import org.apache.iceberg.flink.TableLoader; +import org.apache.iceberg.flink.sink.committer.SinkAggregator; +import org.apache.iceberg.flink.sink.committer.SinkCommittable; +import org.apache.iceberg.flink.sink.committer.SinkCommittableSerializer; +import org.apache.iceberg.flink.sink.committer.SinkCommitter; +import org.apache.iceberg.flink.sink.committer.WriteResultSerializer; +import org.apache.iceberg.flink.sink.writer.IcebergSinkWriter; +import org.apache.iceberg.flink.util.FlinkCompatibilityUtil; +import org.apache.iceberg.io.WriteResult; +import org.apache.iceberg.relocated.com.google.common.annotations.VisibleForTesting; +import org.apache.iceberg.relocated.com.google.common.collect.Lists; +import org.apache.iceberg.relocated.com.google.common.collect.Maps; +import org.apache.iceberg.relocated.com.google.common.collect.Sets; +import org.apache.iceberg.types.TypeUtil; +import org.apache.iceberg.util.SerializableSupplier; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +/** + * Flink v2 sink offer different hooks to insert custom topologies into the sink. We will use the + * following: + * + * <ul> + * <li>{@link SupportsPreWriteTopology} which redistributes the data to the writers based on the + * {@link DistributionMode} + * <li>{@link org.apache.flink.api.connector.sink2.SinkWriter} which writes data files, and + * generates a {@link SinkCommittable} to store the {@link org.apache.iceberg.io.WriteResult} + * <li>{@link SupportsPreCommitTopology} which we use to to place the {@link SinkAggregator} which + * merges the individual {@link org.apache.flink.api.connector.sink2.SinkWriter}'s {@link + * org.apache.iceberg.io.WriteResult}s to a single {@link + * org.apache.iceberg.flink.sink.DeltaManifests} + * <li>{@link SinkCommitter} which stores the incoming {@link + * org.apache.iceberg.flink.sink.DeltaManifests}s in state for recovery, and commits them to + * the Iceberg table + * <li>{@link SupportsPostCommitTopology} we could use for incremental compaction later. This is + * not implemented yet. + * </ul> + * + * The job graph looks like below: + * + * <pre>{@code + * Flink sink + * +-----------------------------------------------------------------------------------+ + * | | + * +-------+ | +----------+ +-------------+ +---------------+ | + * | Map 1 | ==> | | writer 1 | | committer 1 | ---> | post commit 1 | | + * +-------+ | +----------+ +-------------+ +---------------+ | + * | \ / \ | + * | \ / \ | + * | \ / \ | + * +-------+ | +----------+ \ +-------------------+ / +-------------+ \ +---------------+ | + * | Map 2 | ==> | | writer 2 | --->| commit aggregator | | committer 2 | | post commit 2 | | + * +-------+ | +----------+ +-------------------+ +-------------+ +---------------+ | + * | Commit only on | + * | committer 1 | + * +-----------------------------------------------------------------------------------+ + * }</pre> + */ +public class IcebergSink + implements Sink<RowData>, + SupportsPreWriteTopology<RowData>, + SupportsCommitter<SinkCommittable>, + SupportsPreCommitTopology<WriteResult, SinkCommittable>, + SupportsPostCommitTopology<SinkCommittable> { + private Table initTable; + private final TableLoader tableLoader; + private final Map<String, String> snapshotProperties; + private String uidPrefix; + private Committer<SinkCommittable> sinkCommitter; + private final String sinkId; + private transient Builder builder; + private Map<String, String> writeProperties; + private RowType flinkRowType; + private SerializableSupplier<Table> tableSupplier; + private transient FlinkWriteConf flinkWriteConf; + private List<Integer> equalityFieldIds; + private boolean upsertMode; + private FileFormat dataFileFormat; + private long targetDataFileSize; + private String branch; + private boolean overwriteMode; + private int workerPoolSize; + private static final Logger LOG = LoggerFactory.getLogger(IcebergSink.class); + + private IcebergSink(Builder builder) { + this.builder = builder; + this.tableLoader = builder.tableLoader(); + this.snapshotProperties = builder.snapshotSummary; + + // We generate a random UUID every time when a sink is created. + // This is used to separate files generated by different sinks writing the same table. + // Also used to generate the aggregator operator name + this.sinkId = UUID.randomUUID().toString(); + } + + @Override + public SinkWriter<RowData> createWriter(InitContext context) throws IOException { + RowDataTaskWriterFactory taskWriterFactory = + new RowDataTaskWriterFactory( + tableSupplier, + flinkRowType, + targetDataFileSize, + dataFileFormat, + writeProperties, + equalityFieldIds, + upsertMode); + IcebergStreamWriterMetrics metrics = + new IcebergStreamWriterMetrics(context.metricGroup(), initTable.name()); + return new IcebergSinkWriter( + tableSupplier, + taskWriterFactory, + metrics, + context.getSubtaskId(), + context.getAttemptNumber()); + } + + @Override + public Committer<SinkCommittable> createCommitter(CommitterInitContext context) { + IcebergFilesCommitterMetrics metrics = + new IcebergFilesCommitterMetrics(context.metricGroup(), initTable.name()); + return sinkCommitter != null + ? sinkCommitter + : new SinkCommitter( + tableLoader, + branch, + snapshotProperties, + overwriteMode, + workerPoolSize, + sinkId, + metrics); + } + + @Override + public SimpleVersionedSerializer<SinkCommittable> getCommittableSerializer() { + return new SinkCommittableSerializer(); + } + + @Override + public void addPostCommitTopology(DataStream<CommittableMessage<SinkCommittable>> committables) { + // TODO Support small file compaction + } + + @Override + public DataStream<RowData> addPreWriteTopology(DataStream<RowData> inputDataStream) { + return distributeDataStream(inputDataStream); + } + + @Override + public DataStream<CommittableMessage<SinkCommittable>> addPreCommitTopology( + DataStream<CommittableMessage<WriteResult>> writeResults) { + TypeInformation<CommittableMessage<SinkCommittable>> typeInformation = + CommittableMessageTypeInfo.of(this::getCommittableSerializer); + return writeResults + .global() + .transform( + prefixIfNotNull(uidPrefix, initTable.name() + "-" + sinkId + "-pre-commit-topology"), + typeInformation, + new SinkAggregator(tableLoader, sinkId)) + .uid(prefixIfNotNull(uidPrefix, "pre-commit-topology")) + .setParallelism(1) + .setMaxParallelism(1) + .global(); Review Comment: ``` global forces all output records send to subtask 0 of the downstream committer operator. This is to ensure commit only happen in one committer subtask. Once upstream Flink provides the capability of setting committer operator parallelism to 1, this can be removed. link to Flink discussion thread. ``` -- 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: issues-unsubscr...@iceberg.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@iceberg.apache.org For additional commands, e-mail: issues-h...@iceberg.apache.org
