/**
* 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.hama.pipes;
import java.io.IOException;
import java.net.URI;
import java.net.URISyntaxException;
import java.net.URL;
import java.net.URLClassLoader;
import java.security.AccessController;
import java.security.PrivilegedAction;
import java.util.StringTokenizer;
import org.apache.commons.cli.BasicParser;
import org.apache.commons.cli.CommandLine;
import org.apache.commons.cli.Option;
import org.apache.commons.cli.OptionBuilder;
import org.apache.commons.cli.Options;
import org.apache.commons.cli.ParseException;
import org.apache.commons.cli.Parser;
import org.apache.commons.logging.Log;
import org.apache.commons.logging.LogFactory;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.filecache.DistributedCache;
import org.apache.hadoop.fs.FileStatus;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.util.GenericOptionsParser;
import org.apache.hadoop.util.Tool;
import org.apache.hama.HamaConfiguration;
import org.apache.hama.bsp.BSPJob;
import org.apache.hama.bsp.BSPJobClient;
import org.apache.hama.bsp.FileInputFormat;
import org.apache.hama.bsp.FileOutputFormat;
import org.apache.hama.bsp.HashPartitioner;
import org.apache.hama.bsp.InputFormat;
import org.apache.hama.bsp.OutputFormat;
import org.apache.hama.bsp.Partitioner;
import com.google.common.base.Joiner;
/**
* The main entry point and job submitter. It may either be used as a command
* line-based or API-based method to launch Pipes jobs.
*
* Adapted from Hadoop Pipes.
*
*/
public class Submitter implements Tool {
protected static final Log LOG = LogFactory.getLog(Submitter.class);
private HamaConfiguration conf;
public Submitter() {
this.conf = new HamaConfiguration();
}
public Submitter(HamaConfiguration conf) {
setConf(conf);
}
@Override
public HamaConfiguration getConf() {
return this.conf;
}
@Override
public void setConf(Configuration conf) {
this.conf = (HamaConfiguration) conf;
}
/**
* Get the URI of the CPU application's executable.
*
* @param conf
* @return the URI where the application's executable is located
*/
public static String getExecutable(Configuration conf) {
return conf.get("hama.pipes.executable");
}
/**
* Set the URI for the CPU application's executable. Normally this is a hdfs:
* location.
*
* @param conf
* @param executable The URI of the application's executable.
*/
public static void setExecutable(Configuration conf, String executable) {
conf.set("hama.pipes.executable", executable);
}
/**
* Set whether the job is using a Java RecordReader.
*
* @param conf the configuration to modify
* @param value the new value
*/
public static void setIsJavaRecordReader(Configuration conf, boolean value) {
conf.setBoolean("hama.pipes.java.recordreader", value);
}
/**
* Check whether the job is using a Java RecordReader
*
* @param conf the configuration to check
* @return is it a Java RecordReader?
*/
public static boolean getIsJavaRecordReader(Configuration conf) {
return conf.getBoolean("hama.pipes.java.recordreader", false);
}
/**
* Set whether the job will use a Java RecordWriter.
*
* @param conf the configuration to modify
* @param value the new value to set
*/
public static void setIsJavaRecordWriter(Configuration conf, boolean value) {
conf.setBoolean("hama.pipes.java.recordwriter", value);
}
/**
* Will the job use a Java RecordWriter?
*
* @param conf the configuration to check
* @return true, if the output of the job will be written by Java
*/
public static boolean getIsJavaRecordWriter(Configuration conf) {
return conf.getBoolean("hama.pipes.java.recordwriter", false);
}
/**
* Set the configuration, if it doesn't already have a value for the given
* key.
*
* @param conf the configuration to modify
* @param key the key to set
* @param value the new "default" value to set
*/
private static void setIfUnset(Configuration conf, String key, String value) {
if (conf.get(key) == null) {
conf.set(key, value);
}
}
/**
* Save away the user's original partitioner before we override it.
*
* @param conf the configuration to modify
* @param cls the user's partitioner class
*/
static void setJavaPartitioner(Configuration conf, Class<?> cls) {
conf.set("hama.pipes.partitioner", cls.getName());
}
/**
* Get the user's original partitioner.
*
* @param conf the configuration to look in
* @return the class that the user submitted
*/
@SuppressWarnings("rawtypes")
static Class<? extends Partitioner> getJavaPartitioner(Configuration conf) {
return conf.getClass("hama.pipes.partitioner", HashPartitioner.class,
Partitioner.class);
}
/**
* Does the user want to keep the command file for debugging? If this is true,
* pipes will write a copy of the command data to a file in the task directory
* named "downlink.data", which may be used to run the C++ program under the
* debugger. You probably also want to set
* JobConf.setKeepFailedTaskFiles(true) to keep the entire directory from
* being deleted. To run using the data file, set the environment variable
* "hadoop.pipes.command.file" to point to the file.
*
* @param conf the configuration to check
* @return will the framework save the command file?
*/
public static boolean getKeepCommandFile(Configuration conf) {
return conf.getBoolean("hama.pipes.command-file.keep", false);
}
/**
* Set whether to keep the command file for debugging
*
* @param conf the configuration to modify
* @param keep the new value
*/
public static void setKeepCommandFile(Configuration conf, boolean keep) {
conf.setBoolean("hama.pipes.command-file.keep", keep);
}
/**
* Submit a job to the cluster. All of the necessary modifications to the job
* to run under pipes are made to the configuration.
*
* @param conf the job to submit to the cluster (MODIFIED)
* @throws IOException
*/
public static void runJob(BSPJob job) throws IOException {
setupPipesJob(job);
BSPJobClient.runJob(job);
}
private static void setupPipesJob(BSPJob job) throws IOException {
job.setBspClass(PipesBSP.class);
job.setJarByClass(PipesBSP.class);
String textClassname = Text.class.getName();
setIfUnset(job.getConfiguration(), "bsp.input.key.class", textClassname);
setIfUnset(job.getConfiguration(), "bsp.input.value.class", textClassname);
setIfUnset(job.getConfiguration(), "bsp.output.key.class", textClassname);
setIfUnset(job.getConfiguration(), "bsp.output.value.class", textClassname);
setIfUnset(job.getConfiguration(), "bsp.job.name", "Hama Pipes Job");
LOG.debug("isJavaRecordReader: "
+ getIsJavaRecordReader(job.getConfiguration()));
LOG.debug("BspClass: " + job.getBspClass().getName());
// conf.setInputFormat(NLineInputFormat.class);
LOG.debug("InputFormat: " + job.getInputFormat());
LOG.debug("InputKeyClass: " + job.getInputKeyClass().getName());
LOG.debug("InputValueClass: " + job.getInputValueClass().getName());
LOG.debug("OutputKeyClass: " + job.getOutputKeyClass().getName());
LOG.debug("OutputValueClass: " + job.getOutputValueClass().getName());
if ((!job.getOutputKeyClass().getName().equals(textClassname))
|| (!job.getOutputValueClass().getName().equals(textClassname)))
throw new IllegalArgumentException(
"Hama Pipes does only support Text as Key/Value output!");
LOG.debug("bsp.master.address: "
+ job.getConfiguration().get("bsp.master.address"));
LOG.debug("bsp.local.tasks.maximum: "
+ job.getConfiguration().get("bsp.local.tasks.maximum"));
LOG.debug("NumBspTask: " + job.getNumBspTask());
LOG.debug("fs.default.name: "
+ job.getConfiguration().get("fs.default.name"));
String exec = getExecutable(job.getConfiguration());
if (exec == null) {
throw new IllegalArgumentException("No application defined.");
}
URI[] fileCache = DistributedCache.getCacheFiles(job.getConfiguration());
if (fileCache == null) {
fileCache = new URI[1];
} else {
URI[] tmp = new URI[fileCache.length + 1];
System.arraycopy(fileCache, 0, tmp, 1, fileCache.length);
fileCache = tmp;
}
try {
fileCache[0] = new URI(exec);
} catch (URISyntaxException e) {
IOException ie = new IOException("Problem parsing execable URI " + exec);
ie.initCause(e);
throw ie;
}
DistributedCache.setCacheFiles(fileCache, job.getConfiguration());
}
/**
* A command line parser for the CLI-based Pipes job submitter.
*/
static class CommandLineParser {
private Options options = new Options();
void addOption(String longName, boolean required, String description,
String paramName) {
OptionBuilder.withArgName(paramName);
OptionBuilder.hasArgs(1);
OptionBuilder.withDescription(description);
OptionBuilder.isRequired(required);
Option option = OptionBuilder.create(longName);
options.addOption(option);
}
void addArgument(String name, boolean required, String description) {
OptionBuilder.withArgName(name);
OptionBuilder.hasArgs(1);
OptionBuilder.withDescription(description);
OptionBuilder.isRequired(required);
Option option = OptionBuilder.create();
options.addOption(option);
}
Parser createParser() {
Parser result = new BasicParser();
return result;
}
void printUsage() {
// The CLI package should do this for us, but I can't figure out how
// to make it print something reasonable.
System.out.println("bin/hama pipes");
System.out.println(" [-input <path>] // Input directory");
System.out.println(" [-output <path>] // Output directory");
System.out.println(" [-jar <jar file> // jar filename");
System.out.println(" [-inputformat <class>] // InputFormat class");
System.out
.println(" [-bspTasks <number>] // Number of bsp tasks to launch");
System.out.println(" [-partitioner <class>] // Java Partitioner");
System.out.println(" [-combiner <class>] // Java Combiner class");
System.out.println(" [-outputformat <class>] // Java RecordWriter");
System.out
.println(" [-cachefiles <space separated paths>] // Additional cache files like libs, can be globbed with wildcards");
System.out.println(" [-program <executable>] // executable URI");
System.out
.println(" [-programArgs <argument>] // arguments for the program");
System.out
.println(" [-interpreter <executable>] // interpreter, like python or bash");
System.out
.println(" [-streaming <true|false>] // if supplied, streaming is used instead of pipes");
System.out.println(" [-jobname <name>] // sets the name of this job");
System.out.println();
GenericOptionsParser.printGenericCommandUsage(System.out);
}
}
private static <InterfaceType> Class<? extends InterfaceType> getClass(
CommandLine cl, String key, HamaConfiguration conf,
Class<InterfaceType> cls) throws ClassNotFoundException {
return conf.getClassByName(cl.getOptionValue(key)).asSubclass(cls);
}
@Override
public int run(String[] args) throws Exception {
CommandLineParser cli = new CommandLineParser();
if (args.length == 0) {
cli.printUsage();
return 1;
}
cli.addOption("input", false, "input path for bsp", "path");
cli.addOption("output", false, "output path from bsp", "path");
cli.addOption("jar", false, "job jar file", "path");
cli.addOption("inputformat", false, "java classname of InputFormat",
"class");
// cli.addArgument("javareader", false, "is the RecordReader in Java");
cli.addOption("partitioner", false, "java classname of Partitioner",
"class");
cli.addOption("outputformat", false, "java classname of OutputFormat",
"class");
cli.addOption("cachefiles", false, "additional cache files to add",
"space delimited paths");
cli.addOption("interpreter", false, "interpreter, like python or bash",
"executable");
cli.addOption("jobname", false, "the jobname", "name");
cli.addOption("programArgs", false, "program arguments", "arguments");
cli.addOption("bspTasks", false, "how many bsp tasks to launch", "number");
cli.addOption("streaming", false,
"if supplied, streaming is used instead of pipes", "");
cli.addOption(
"jobconf",
false,
"\"n1=v1,n2=v2,..\" (Deprecated) Optional. Add or override a JobConf property.",
"key=val");
cli.addOption("program", false, "URI to application executable", "class");
Parser parser = cli.createParser();
try {
// check generic arguments -conf
GenericOptionsParser genericParser = new GenericOptionsParser(getConf(),
args);
// get other arguments
CommandLine results = parser.parse(cli.options,
genericParser.getRemainingArgs());
BSPJob job = new BSPJob(getConf());
if (results.hasOption("input")) {
FileInputFormat.setInputPaths(job, results.getOptionValue("input"));
}
if (results.hasOption("output")) {
FileOutputFormat.setOutputPath(job,
new Path(results.getOptionValue("output")));
}
if (results.hasOption("jar")) {
job.setJar(results.getOptionValue("jar"));
}
if (results.hasOption("jobname")) {
job.setJobName(results.getOptionValue("jobname"));
}
if (results.hasOption("inputformat")) {
setIsJavaRecordReader(job.getConfiguration(), true);
job.setInputFormat(getClass(results, "inputformat", conf,
InputFormat.class));
}
if (results.hasOption("partitioner")) {
job.setPartitioner(getClass(results, "partitioner", conf,
Partitioner.class));
}
if (results.hasOption("outputformat")) {
setIsJavaRecordWriter(job.getConfiguration(), true);
job.setOutputFormat(getClass(results, "outputformat", conf,
OutputFormat.class));
}
if (results.hasOption("streaming")) {
LOG.info("Streaming enabled!");
job.set("hama.streaming.enabled", "true");
}
if (results.hasOption("jobconf")) {
LOG.warn("-jobconf option is deprecated, please use -D instead.");
String options = results.getOptionValue("jobconf");
StringTokenizer tokenizer = new StringTokenizer(options, ",");
while (tokenizer.hasMoreTokens()) {
String keyVal = tokenizer.nextToken().trim();
String[] keyValSplit = keyVal.split("=", 2);
job.set(keyValSplit[0], keyValSplit[1]);
}
}
if (results.hasOption("bspTasks")) {
int optionValue = Integer.parseInt(results.getOptionValue("bspTasks"));
conf.setInt("bsp.local.tasks.maximum", optionValue);
conf.setInt("bsp.peers.num", optionValue);
}
if (results.hasOption("program")) {
String executablePath = results.getOptionValue("program");
setExecutable(job.getConfiguration(), executablePath);
DistributedCache.addCacheFile(new Path(executablePath).toUri(), conf);
}
if (results.hasOption("interpreter")) {
job.getConfiguration().set("hama.pipes.executable.interpretor",
results.getOptionValue("interpreter"));
}
if (results.hasOption("programArgs")) {
job.getConfiguration().set("hama.pipes.executable.args",
Joiner.on(" ").join(results.getOptionValues("programArgs")));
// job.getConfiguration().set("hama.pipes.resolve.executable.args",
// "true");
}
if (results.hasOption("cachefiles")) {
FileSystem fs = FileSystem.get(getConf());
String[] optionValues = results.getOptionValues("cachefiles");
for (String s : optionValues) {
Path path = new Path(s);
FileStatus[] globStatus = fs.globStatus(path);
for (FileStatus f : globStatus) {
if (!f.isDir()) {
DistributedCache.addCacheFile(f.getPath().toUri(),
job.getConfiguration());
} else {
LOG.info("Ignoring directory " + f.getPath() + " while globbing.");
}
}
}
}
// if they gave us a jar file, include it into the class path
String jarFile = job.getJar();
if (jarFile != null) {
@SuppressWarnings("deprecation")
final URL[] urls = new URL[] { FileSystem.getLocal(conf)
.pathToFile(new Path(jarFile)).toURL() };
// FindBugs complains that creating a URLClassLoader should be
// in a doPrivileged() block.
ClassLoader loader = AccessController
.doPrivileged(new PrivilegedAction<ClassLoader>() {
@Override
public ClassLoader run() {
return new URLClassLoader(urls);
}
});
conf.setClassLoader(loader);
}
runJob(job);
return 0;
} catch (ParseException pe) {
LOG.info("Error : " + pe);
cli.printUsage();
return 1;
}
}
/**
* Submit a pipes job based on the command line arguments.
*/
public static void main(String[] args) throws Exception {
int exitCode = new Submitter().run(args);
System.exit(exitCode);
}
}