flink-java/src/main/java/org/apache/flink/api/java/hadoop/mapreduce/HadoopOutputFormatBase.java

/*
 * 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.api.java.hadoop.mapreduce;

import org.apache.flink.annotation.Internal;
import org.apache.flink.api.common.io.FinalizeOnMaster;
import org.apache.flink.api.java.hadoop.common.HadoopOutputFormatCommonBase;
import org.apache.flink.api.java.hadoop.mapreduce.utils.HadoopUtils;
import org.apache.flink.configuration.Configuration;
import org.apache.hadoop.conf.Configurable;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.mapreduce.Job;
import org.apache.hadoop.mapreduce.JobContext;
import org.apache.hadoop.mapreduce.JobID;
import org.apache.hadoop.mapreduce.OutputCommitter;
import org.apache.hadoop.mapreduce.RecordWriter;
import org.apache.hadoop.mapreduce.TaskAttemptContext;
import org.apache.hadoop.mapreduce.TaskAttemptID;
import org.apache.hadoop.mapreduce.lib.output.FileOutputCommitter;
import org.apache.hadoop.security.Credentials;
import org.apache.hadoop.security.UserGroupInformation;

import java.io.IOException;
import java.io.ObjectInputStream;
import java.io.ObjectOutputStream;

import static org.apache.flink.api.java.hadoop.common.HadoopInputFormatCommonBase.getCredentialsFromUGI;

/**
 * Base class shared between the Java and Scala API of Flink
 */
@Internal
public abstract class HadoopOutputFormatBase<K, V, T> extends HadoopOutputFormatCommonBase<T> implements FinalizeOnMaster {

	private static final long serialVersionUID = 1L;

	// Mutexes to avoid concurrent operations on Hadoop OutputFormats.
	// Hadoop parallelizes tasks across JVMs which is why they might rely on this JVM isolation.
	// In contrast, Flink parallelizes using Threads, so multiple Hadoop OutputFormat instances
	// might be used in the same JVM.
	protected static final Object OPEN_MUTEX = new Object();
	protected static final Object CONFIGURE_MUTEX = new Object();
	protected static final Object CLOSE_MUTEX = new Object();

	protected org.apache.hadoop.conf.Configuration configuration;
	protected org.apache.hadoop.mapreduce.OutputFormat<K,V> mapreduceOutputFormat;
	protected transient RecordWriter<K,V> recordWriter;
	protected transient OutputCommitter outputCommitter;
	protected transient TaskAttemptContext context;
	protected transient int taskNumber;

	public HadoopOutputFormatBase(org.apache.hadoop.mapreduce.OutputFormat<K, V> mapreduceOutputFormat, Job job) {
		super(job.getCredentials());
		this.mapreduceOutputFormat = mapreduceOutputFormat;
		this.configuration = job.getConfiguration();
		HadoopUtils.mergeHadoopConf(configuration);
	}

	public org.apache.hadoop.conf.Configuration getConfiguration() {
		return this.configuration;
	}

	// --------------------------------------------------------------------------------------------
	//  OutputFormat
	// --------------------------------------------------------------------------------------------

	@Override
	public void configure(Configuration parameters) {

		// enforce sequential configure() calls
		synchronized (CONFIGURE_MUTEX) {
			if (this.mapreduceOutputFormat instanceof Configurable) {
				((Configurable) this.mapreduceOutputFormat).setConf(this.configuration);
			}
		}
	}

	/**
	 * create the temporary output file for hadoop RecordWriter.
	 * @param taskNumber The number of the parallel instance.
	 * @param numTasks The number of parallel tasks.
	 * @throws java.io.IOException
	 */
	@Override
	public void open(int taskNumber, int numTasks) throws IOException {

		// enforce sequential open() calls
		synchronized (OPEN_MUTEX) {
			if (Integer.toString(taskNumber + 1).length() > 6) {
				throw new IOException("Task id too large.");
			}

			this.taskNumber = taskNumber + 1;

			// for hadoop 2.2
			this.configuration.set("mapreduce.output.basename", "tmp");

			TaskAttemptID taskAttemptID = TaskAttemptID.forName("attempt__0000_r_"
					+ String.format("%" + (6 - Integer.toString(taskNumber + 1).length()) + "s", " ").replace(" ", "0")
					+ Integer.toString(taskNumber + 1)
					+ "_0");

			this.configuration.set("mapred.task.id", taskAttemptID.toString());
			this.configuration.setInt("mapred.task.partition", taskNumber + 1);
			// for hadoop 2.2
			this.configuration.set("mapreduce.task.attempt.id", taskAttemptID.toString());
			this.configuration.setInt("mapreduce.task.partition", taskNumber + 1);

			try {
				this.context = HadoopUtils.instantiateTaskAttemptContext(this.configuration, taskAttemptID);
				this.outputCommitter = this.mapreduceOutputFormat.getOutputCommitter(this.context);
				this.outputCommitter.setupJob(HadoopUtils.instantiateJobContext(this.configuration, new JobID()));
			} catch (Exception e) {
				throw new RuntimeException(e);
			}

			this.context.getCredentials().addAll(this.credentials);
			Credentials currentUserCreds = getCredentialsFromUGI(UserGroupInformation.getCurrentUser());
			if (currentUserCreds != null) {
				this.context.getCredentials().addAll(currentUserCreds);
			}

			// compatible for hadoop 2.2.0, the temporary output directory is different from hadoop 1.2.1
			if (outputCommitter instanceof FileOutputCommitter) {
				this.configuration.set("mapreduce.task.output.dir", ((FileOutputCommitter) this.outputCommitter).getWorkPath().toString());
			}

			try {
				this.recordWriter = this.mapreduceOutputFormat.getRecordWriter(this.context);
			} catch (InterruptedException e) {
				throw new IOException("Could not create RecordWriter.", e);
			}
		}
	}

	/**
	 * commit the task by moving the output file out from the temporary directory.
	 * @throws java.io.IOException
	 */
	@Override
	public void close() throws IOException {

		// enforce sequential close() calls
		synchronized (CLOSE_MUTEX) {
			try {
				this.recordWriter.close(this.context);
			} catch (InterruptedException e) {
				throw new IOException("Could not close RecordReader.", e);
			}

			if (this.outputCommitter.needsTaskCommit(this.context)) {
				this.outputCommitter.commitTask(this.context);
			}

			Path outputPath = new Path(this.configuration.get("mapred.output.dir"));

			// rename tmp-file to final name
			FileSystem fs = FileSystem.get(outputPath.toUri(), this.configuration);

			String taskNumberStr = Integer.toString(this.taskNumber);
			String tmpFileTemplate = "tmp-r-00000";
			String tmpFile = tmpFileTemplate.substring(0, 11 - taskNumberStr.length()) + taskNumberStr;

			if (fs.exists(new Path(outputPath.toString() + "/" + tmpFile))) {
				fs.rename(new Path(outputPath.toString() + "/" + tmpFile), new Path(outputPath.toString() + "/" + taskNumberStr));
			}
		}
	}

	@Override
	public void finalizeGlobal(int parallelism) throws IOException {

		JobContext jobContext;
		TaskAttemptContext taskContext;
		try {
			TaskAttemptID taskAttemptID = TaskAttemptID.forName("attempt__0000_r_"
					+ String.format("%" + (6 - Integer.toString(1).length()) + "s"," ").replace(" ", "0")
					+ Integer.toString(1)
					+ "_0");

			jobContext = HadoopUtils.instantiateJobContext(this.configuration, new JobID());
			taskContext = HadoopUtils.instantiateTaskAttemptContext(this.configuration, taskAttemptID);
			this.outputCommitter = this.mapreduceOutputFormat.getOutputCommitter(taskContext);
		} catch (Exception e) {
			throw new RuntimeException(e);
		}

		jobContext.getCredentials().addAll(this.credentials);
		Credentials currentUserCreds = getCredentialsFromUGI(UserGroupInformation.getCurrentUser());
		if(currentUserCreds != null) {
			jobContext.getCredentials().addAll(currentUserCreds);
		}

		// finalize HDFS output format
		if(this.outputCommitter != null) {
			this.outputCommitter.commitJob(jobContext);
		}
	}
	
	// --------------------------------------------------------------------------------------------
	//  Custom serialization methods
	// --------------------------------------------------------------------------------------------
	
	private void writeObject(ObjectOutputStream out) throws IOException {
		super.write(out);
		out.writeUTF(this.mapreduceOutputFormat.getClass().getName());
		this.configuration.write(out);
	}
	
	@SuppressWarnings("unchecked")
	private void readObject(ObjectInputStream in) throws IOException, ClassNotFoundException {
		super.read(in);
		String hadoopOutputFormatClassName = in.readUTF();
		
		org.apache.hadoop.conf.Configuration configuration = new org.apache.hadoop.conf.Configuration();
		configuration.readFields(in);
		
		if(this.configuration == null) {
			this.configuration = configuration;
		}
		
		try {
			this.mapreduceOutputFormat = (org.apache.hadoop.mapreduce.OutputFormat<K,V>) Class.forName(hadoopOutputFormatClassName, true, Thread.currentThread().getContextClassLoader()).newInstance();
		} catch (Exception e) {
			throw new RuntimeException("Unable to instantiate the hadoop output format", e);
		}
	}
}