danny0405 commented on a change in pull request #2430: URL: https://github.com/apache/hudi/pull/2430#discussion_r559431430
########## File path: hudi-flink/src/main/java/org/apache/hudi/operator/StreamWriteFunction.java ########## @@ -0,0 +1,342 @@ +/* + * 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.operator; + +import org.apache.hudi.avro.HoodieAvroUtils; +import org.apache.hudi.client.FlinkTaskContextSupplier; +import org.apache.hudi.client.HoodieFlinkWriteClient; +import org.apache.hudi.client.WriteStatus; +import org.apache.hudi.client.common.HoodieFlinkEngineContext; +import org.apache.hudi.common.config.SerializableConfiguration; +import org.apache.hudi.common.model.HoodieRecord; +import org.apache.hudi.common.model.HoodieRecordPayload; +import org.apache.hudi.common.model.WriteOperationType; +import org.apache.hudi.common.util.ObjectSizeCalculator; +import org.apache.hudi.keygen.KeyGenerator; +import org.apache.hudi.operator.event.BatchWriteSuccessEvent; +import org.apache.hudi.util.StreamerUtil; + +import org.apache.avro.Schema; +import org.apache.avro.generic.GenericRecord; +import org.apache.flink.annotation.VisibleForTesting; +import org.apache.flink.configuration.Configuration; +import org.apache.flink.formats.avro.RowDataToAvroConverters; +import org.apache.flink.runtime.operators.coordination.OperatorEventGateway; +import org.apache.flink.runtime.state.FunctionInitializationContext; +import org.apache.flink.runtime.state.FunctionSnapshotContext; +import org.apache.flink.streaming.api.checkpoint.CheckpointedFunction; +import org.apache.flink.streaming.api.functions.KeyedProcessFunction; +import org.apache.flink.table.types.logical.RowType; +import org.apache.flink.util.Collector; +import org.apache.flink.util.Preconditions; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +import java.io.IOException; +import java.util.ArrayList; +import java.util.Collections; +import java.util.List; +import java.util.Random; +import java.util.concurrent.locks.Condition; +import java.util.concurrent.locks.ReentrantLock; +import java.util.function.BiFunction; + +/** + * Sink function to write the data to the underneath filesystem. + * + * <p><h2>Work Flow</h2> + * + * <p>The function firstly buffers the data as a batch of {@link HoodieRecord}s, + * It flushes(write) the records batch when a Flink checkpoint starts. After a batch has been written successfully, + * the function notifies its operator coordinator {@link StreamWriteOperatorCoordinator} to mark a successful write. + * + * <p><h2>Exactly-once Semantics</h2> + * + * <p>The task implements exactly-once semantics by buffering the data between checkpoints. The operator coordinator + * starts a new instant on the time line when a checkpoint triggers, the coordinator checkpoints always + * start before its operator, so when this function starts a checkpoint, a REQUESTED instant already exists. + * The function process thread then block data buffering and the checkpoint thread starts flushing the existing data buffer. + * When the existing data buffer write successfully, the process thread unblock and start buffering again for the next round checkpoint. + * Because any checkpoint failures would trigger the write rollback, it implements the exactly-once semantics. + * + * <p><h2>Fault Tolerance</h2> + * + * <p>The operator coordinator checks the validity for the last instant when it starts a new one. The operator rolls back + * the written data and throws when any error occurs. This means any checkpoint or task failure would trigger a failover. + * The operator coordinator would try several times when committing the writestatus. + * + * <p>Note: The function task requires the input stream be partitioned by the partition fields to avoid different write tasks + * write to the same file group that conflict. The general case for partition path is a datetime field, + * so the sink task is very possible to have IO bottleneck, the more flexible solution is to shuffle the + * data by the file group IDs. + * + * @param <I> Type of the input record + * @see StreamWriteOperatorCoordinator + */ +public class StreamWriteFunction<K, I, O> + extends KeyedProcessFunction<K, I, O> + implements CheckpointedFunction { + + private static final long serialVersionUID = 1L; + + private static final Logger LOG = LoggerFactory.getLogger(StreamWriteFunction.class); + + /** + * Write buffer for a checkpoint. + */ + private transient List<HoodieRecord> buffer; + + /** + * The buffer lock to control data buffering/flushing. + */ + private transient ReentrantLock bufferLock; + + /** + * The condition to decide whether to add new records into the buffer. + */ + private transient Condition addToBufferCondition; + + /** + * Flag saying whether there is an on-going checkpoint. + */ + private volatile boolean onCheckpointing = false; + + /** + * Config options. + */ + private final Configuration config; + + /** + * Id of current subtask. + */ + private int taskID; + + /** + * Write Client. + */ + private transient HoodieFlinkWriteClient writeClient; + + private transient BiFunction<List<HoodieRecord>, String, List<WriteStatus>> writeFunction; + + /** + * HoodieKey generator. + */ + private transient KeyGenerator keyGenerator; + + /** + * Row type of the input. + */ + private final RowType rowType; + + /** + * Avro schema of the input. + */ + private final Schema avroSchema; + + private transient RowDataToAvroConverters.RowDataToAvroConverter converter; + + /** + * The REQUESTED instant we write the data. + */ + private volatile String currentInstant; + + /** + * Gateway to send operator events to the operator coordinator. + */ + private transient OperatorEventGateway eventGateway; + + /** + * Constructs a StreamingSinkFunction. + * + * @param rowType The input row type + * @param config The config options + */ + public StreamWriteFunction(RowType rowType, Configuration config) { + this.rowType = rowType; + this.avroSchema = org.apache.flink.formats.avro.typeutils + .AvroSchemaConverter.convertToSchema(rowType); + this.config = config; + } + + @Override + public void open(Configuration parameters) throws IOException { + this.taskID = getRuntimeContext().getIndexOfThisSubtask(); + this.keyGenerator = StreamerUtil.createKeyGenerator(HoodieOptions.flatOptions(this.config)); + this.converter = RowDataToAvroConverters.createConverter(this.rowType); + initBuffer(); + initWriteClient(); + initWriteFunction(); + } + + @Override + public void initializeState(FunctionInitializationContext context) { + // no operation + } + + @Override + public void snapshotState(FunctionSnapshotContext functionSnapshotContext) { + bufferLock.lock(); + try { + // Based on the fact that the coordinator starts the checkpoint first, + // it would check the validity. + this.onCheckpointing = true; + this.currentInstant = this.writeClient.getInflightAndRequestedInstant(this.config.get(HoodieOptions.TABLE_TYPE)); + Preconditions.checkNotNull(this.currentInstant, + "No inflight instant when flushing data"); + // wait for the buffer data flush out and request a new instant + flushBuffer(); + // signal the task thread to start buffering + addToBufferCondition.signal(); + } finally { + this.onCheckpointing = false; + bufferLock.unlock(); + } + } + + @Override + public void processElement(I value, KeyedProcessFunction<K, I, O>.Context ctx, Collector<O> out) throws Exception { + bufferLock.lock(); + try { + if (onCheckpointing) { + addToBufferCondition.await(); + } + this.buffer.add(toHoodie(value)); + } finally { + bufferLock.unlock(); + } + } + + @Override + public void close() { + if (this.writeClient != null) { + this.writeClient.close(); + } + } + + // ------------------------------------------------------------------------- + // Getter/Setter + // ------------------------------------------------------------------------- + + @VisibleForTesting + @SuppressWarnings("rawtypes") + public List<HoodieRecord> getBuffer() { + return buffer; + } + + @VisibleForTesting + @SuppressWarnings("rawtypes") + public HoodieFlinkWriteClient getWriteClient() { + return writeClient; + } + + public void setOperatorEventGateway(OperatorEventGateway operatorEventGateway) { + this.eventGateway = operatorEventGateway; + } + + // ------------------------------------------------------------------------- + // Utilities + // ------------------------------------------------------------------------- + + private void initBuffer() { + this.buffer = new ArrayList<>(); + this.bufferLock = new ReentrantLock(); + this.addToBufferCondition = this.bufferLock.newCondition(); + } + + private void initWriteClient() { + HoodieFlinkEngineContext context = + new HoodieFlinkEngineContext( + new SerializableConfiguration(StreamerUtil.getHadoopConf()), + new FlinkTaskContextSupplier(getRuntimeContext())); + + writeClient = new HoodieFlinkWriteClient<>(context, StreamerUtil.getHoodieClientConfig(this.config)); + } + + private void initWriteFunction() { + final String writeOperation = this.config.get(HoodieOptions.OPERATION); + switch (WriteOperationType.fromValue(writeOperation)) { + case INSERT: + this.writeFunction = (records, instantTime) -> this.writeClient.insert(records, instantTime); + break; + case UPSERT: + this.writeFunction = (records, instantTime) -> this.writeClient.upsert(records, instantTime); + break; + default: + throw new RuntimeException("Unsupported write operation : " + writeOperation); + } + } + + // Keep for mini-batch write. + private static class BufferSizeEstimator { + private final Random random = new Random(47); + + private long lastRecordSize = -1L; + private long totalSize = 0L; + + long estimate(Object record) { + if (lastRecordSize == -1 || sampling()) { + lastRecordSize = ObjectSizeCalculator.getObjectSize(record); + } + totalSize += lastRecordSize; + return totalSize; + } + + boolean sampling() { + // 0.01 sampling percentage + return random.nextInt(100) == 1; + } + + void reset() { + this.lastRecordSize = -1L; + this.totalSize = 0L; + } + } + + /** + * Converts the give record to a {@link HoodieRecord}. + * + * @param record The input record + * @return HoodieRecord based on the configuration + * @throws IOException if error occurs + */ + @SuppressWarnings("rawtypes") + private HoodieRecord toHoodie(I record) throws IOException { Review comment: Agree, thanks ~ ---------------------------------------------------------------- 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: us...@infra.apache.org
