/**
* Driver to copy srcPath to destPath depending on required protocol.
*
* @param args arguments
*/
static void copy(final Configuration conf, final Arguments args) throws IOException {
LOG.info("srcPaths=" + args.srcs);
LOG.info("destPath=" + args.dst);
checkSrcPath(conf, args.srcs);
JobConf job = createJobConf(conf);
if (args.preservedAttributes != null) {
job.set(PRESERVE_STATUS_LABEL, args.preservedAttributes);
}
if (args.mapredSslConf != null) {
job.set("dfs.https.client.keystore.resource", args.mapredSslConf);
}
// Initialize the mapper
try {
setup(conf, job, args);
JobClient.runJob(job);
finalize(conf, job, args.dst, args.preservedAttributes);
} finally {
// delete tmp
fullyDelete(job.get(TMP_DIR_LABEL), job);
// delete jobDirectory
fullyDelete(job.get(JOB_DIR_LABEL), job);
}
}
public void configure(JobConf conf) {
/*
* It reads all the configurations and distributed cache from outside.
*/
// Read number of nodes in input layer and output layer from configuration
inputNumdims = conf.get("numdims");
inputNumhid = conf.get("numhid");
// Read the weights from distributed cache
Path[] pathwaysFiles = new Path[0];
try {
pathwaysFiles = DistributedCache.getLocalCacheFiles(conf);
for (Path path : pathwaysFiles) {
/*
* this loop reads all the distributed cache files
* In fact, the driver program ensures that there is only one distributed cache file
*/
BufferedReader fis = new BufferedReader(new FileReader(path.toString()));
weightline = fis.readLine();
}
} catch (Exception e) {
e.printStackTrace();
}
}
public void configure(JobConf conf) {
numberOfCenters = Integer.valueOf(conf.get("numberOfCenters"));
centersDirectory = conf.get("centersReadDirectory");
try {
Configuration c = new Configuration();
FileSystem fs = FileSystem.get(c);
for (int index = 0; index < numberOfCenters; ++index) {
SequenceFile.Reader reader =
new SequenceFile.Reader(fs, new Path(centersDirectory + "/centers/" + index), c);
LongWritable key = new LongWritable();
Point value = new Point();
reader.next(key, value);
Point center = (Point) value;
centers.add(center);
reader.close();
}
} catch (IOException e) {
// do nothing
// I hope this doesn't happen
System.out.println("well, damn.");
e.printStackTrace();
}
}
// Mostly for setting up the symlinks. Note that when we setup the distributed
// cache, we didn't create the symlinks. This is done on a per task basis
// by the currently executing task.
public static void setupWorkDir(JobConf conf) throws IOException {
File workDir = new File(".").getAbsoluteFile();
FileUtil.fullyDelete(workDir);
if (DistributedCache.getSymlink(conf)) {
URI[] archives = DistributedCache.getCacheArchives(conf);
URI[] files = DistributedCache.getCacheFiles(conf);
Path[] localArchives = DistributedCache.getLocalCacheArchives(conf);
Path[] localFiles = DistributedCache.getLocalCacheFiles(conf);
if (archives != null) {
for (int i = 0; i < archives.length; i++) {
String link = archives[i].getFragment();
if (link != null) {
link = workDir.toString() + Path.SEPARATOR + link;
File flink = new File(link);
if (!flink.exists()) {
FileUtil.symLink(localArchives[i].toString(), link);
}
}
}
}
if (files != null) {
for (int i = 0; i < files.length; i++) {
String link = files[i].getFragment();
if (link != null) {
link = workDir.toString() + Path.SEPARATOR + link;
File flink = new File(link);
if (!flink.exists()) {
FileUtil.symLink(localFiles[i].toString(), link);
}
}
}
}
}
File jobCacheDir = null;
if (conf.getJar() != null) {
jobCacheDir = new File(new Path(conf.getJar()).getParent().toString());
}
// create symlinks for all the files in job cache dir in current
// workingdir for streaming
try {
DistributedCache.createAllSymlink(conf, jobCacheDir, workDir);
} catch (IOException ie) {
// Do not exit even if symlinks have not been created.
LOG.warn(StringUtils.stringifyException(ie));
}
// add java.io.tmpdir given by mapred.child.tmp
String tmp = conf.get("mapred.child.tmp", "./tmp");
Path tmpDir = new Path(tmp);
// if temp directory path is not absolute
// prepend it with workDir.
if (!tmpDir.isAbsolute()) {
tmpDir = new Path(workDir.toString(), tmp);
FileSystem localFs = FileSystem.getLocal(conf);
if (!localFs.mkdirs(tmpDir) && !localFs.getFileStatus(tmpDir).isDir()) {
throw new IOException("Mkdirs failed to create " + tmpDir.toString());
}
}
}
private static void finalize(
Configuration conf, JobConf jobconf, final Path destPath, String presevedAttributes)
throws IOException {
if (presevedAttributes == null) {
return;
}
EnumSet<FileAttribute> preseved = FileAttribute.parse(presevedAttributes);
if (!preseved.contains(FileAttribute.USER)
&& !preseved.contains(FileAttribute.GROUP)
&& !preseved.contains(FileAttribute.PERMISSION)) {
return;
}
FileSystem dstfs = destPath.getFileSystem(conf);
Path dstdirlist = new Path(jobconf.get(DST_DIR_LIST_LABEL));
SequenceFile.Reader in = null;
try {
in = new SequenceFile.Reader(dstdirlist.getFileSystem(jobconf), dstdirlist, jobconf);
Text dsttext = new Text();
FilePair pair = new FilePair();
for (; in.next(dsttext, pair); ) {
Path absdst = new Path(destPath, pair.output);
updatePermissions(pair.input, dstfs.getFileStatus(absdst), preseved, dstfs);
}
} finally {
checkAndClose(in);
}
}
private static IntWritable deduceInputFile(JobConf job) {
Path[] inputPaths = FileInputFormat.getInputPaths(job);
Path inputFile = new Path(job.get("map.input.file"));
// value == one for sort-input; value == two for sort-output
return (inputFile.getParent().equals(inputPaths[0])) ? sortInput : sortOutput;
}
/**
* Mapper configuration. Extracts source and destination file system, as well as top-level paths
* on source and destination directories. Gets the named file systems, to be used later in map.
*/
public void configure(JobConf job) {
destPath = new Path(job.get(DST_DIR_LABEL, "/"));
try {
destFileSys = destPath.getFileSystem(job);
} catch (IOException ex) {
throw new RuntimeException("Unable to get the named file system.", ex);
}
sizeBuf = job.getInt("copy.buf.size", 128 * 1024);
buffer = new byte[sizeBuf];
ignoreReadFailures = job.getBoolean(Options.IGNORE_READ_FAILURES.propertyname, false);
preserve_status = job.getBoolean(Options.PRESERVE_STATUS.propertyname, false);
if (preserve_status) {
preseved = FileAttribute.parse(job.get(PRESERVE_STATUS_LABEL));
}
update = job.getBoolean(Options.UPDATE.propertyname, false);
overwrite = !update && job.getBoolean(Options.OVERWRITE.propertyname, false);
this.job = job;
}
/**
* @param split
* @param jobConf
* @throws IOException
*/
public XMLRecordReader(FileSplit split, JobConf jobConf) throws IOException {
if (jobConf.get(START_TAG_KEY) == null || jobConf.get(END_TAG_KEY) == null) {
throw new RuntimeException("Error! XML start and end tags unspecified!");
}
startTag = jobConf.get(START_TAG_KEY).getBytes("utf-8");
endTag = jobConf.get(END_TAG_KEY).getBytes("utf-8");
start = split.getStart();
Path file = split.getPath();
CompressionCodecFactory compressionCodecs = new CompressionCodecFactory(jobConf);
CompressionCodec codec = compressionCodecs.getCodec(file);
FileSystem fs = file.getFileSystem(jobConf);
if (codec != null) {
LOG.info("Reading compressed file...");
fsin = new DataInputStream(codec.createInputStream(fs.open(file)));
end = Long.MAX_VALUE;
} else {
LOG.info("Reading uncompressed file...");
FSDataInputStream fileIn = fs.open(file);
fileIn.seek(start);
fsin = fileIn;
end = start + split.getLength();
}
recordStartPos = start;
// Because input streams of gzipped files are not seekable (specifically,
// do not support
// getPos), we need to keep track of bytes consumed ourselves.
pos = start;
}
/** @see org.apache.hadoop.mapred.JobSubmissionProtocol#getStagingAreaDir() */
public String getStagingAreaDir() throws IOException {
Path stagingRootDir =
new Path(conf.get("mapreduce.jobtracker.staging.root.dir", "/tmp/hadoop/mapred/staging"));
UserGroupInformation ugi = UserGroupInformation.getCurrentUser();
String user;
randid = rand.nextInt(Integer.MAX_VALUE);
if (ugi != null) {
user = ugi.getShortUserName() + randid;
} else {
user = "******" + randid;
}
return fs.makeQualified(new Path(stagingRootDir, user + "/.staging")).toString();
}
/**
* Produce splits such that each is no greater than the quotient of the total size and the
* number of splits requested.
*
* @param job The handle to the JobConf object
* @param numSplits Number of splits requested
*/
public InputSplit[] getSplits(JobConf job, int numSplits) throws IOException {
int cnfiles = job.getInt(SRC_COUNT_LABEL, -1);
long cbsize = job.getLong(TOTAL_SIZE_LABEL, -1);
String srcfilelist = job.get(SRC_LIST_LABEL, "");
if (cnfiles < 0 || cbsize < 0 || "".equals(srcfilelist)) {
throw new RuntimeException(
"Invalid metadata: #files("
+ cnfiles
+ ") total_size("
+ cbsize
+ ") listuri("
+ srcfilelist
+ ")");
}
Path src = new Path(srcfilelist);
FileSystem fs = src.getFileSystem(job);
FileStatus srcst = fs.getFileStatus(src);
ArrayList<FileSplit> splits = new ArrayList<FileSplit>(numSplits);
LongWritable key = new LongWritable();
FilePair value = new FilePair();
final long targetsize = cbsize / numSplits;
long pos = 0L;
long last = 0L;
long acc = 0L;
long cbrem = srcst.getLen();
SequenceFile.Reader sl = null;
try {
sl = new SequenceFile.Reader(fs, src, job);
for (; sl.next(key, value); last = sl.getPosition()) {
// if adding this split would put this split past the target size,
// cut the last split and put this next file in the next split.
if (acc + key.get() > targetsize && acc != 0) {
long splitsize = last - pos;
splits.add(new FileSplit(src, pos, splitsize, (String[]) null));
cbrem -= splitsize;
pos = last;
acc = 0L;
}
acc += key.get();
}
} finally {
checkAndClose(sl);
}
if (cbrem != 0) {
splits.add(new FileSplit(src, pos, cbrem, (String[]) null));
}
return splits.toArray(new FileSplit[splits.size()]);
}
@Override
public void configure(JobConf job) {
super.configure(job);
this.localDir = job.get("mapred.child.tmp");
try {
FileAndWriter faw = openFileWriter(localDir, 1);
s1File = faw.file;
s1FileWriter = faw.writer;
faw = openFileWriter(localDir, 2);
s2File = faw.file;
s2FileWriter = faw.writer;
} catch (IOException e) {
e.printStackTrace();
}
}
TaskTrackerRunner(int trackerId, int numDir, String hostname, JobConf cfg) throws IOException {
this.trackerId = trackerId;
this.numDir = numDir;
localDirs = new String[numDir];
final JobConf conf;
if (cfg == null) {
conf = createJobConf();
} else {
conf = createJobConf(cfg);
}
if (hostname != null) {
conf.set("slave.host.name", hostname);
}
conf.set("mapred.task.tracker.http.address", "0.0.0.0:0");
conf.set("mapred.task.tracker.report.address", "127.0.0.1:" + taskTrackerPort);
File localDirBase = new File(conf.get("mapred.local.dir")).getAbsoluteFile();
localDirBase.mkdirs();
StringBuffer localPath = new StringBuffer();
for (int i = 0; i < numDir; ++i) {
File ttDir = new File(localDirBase, Integer.toString(trackerId) + "_" + i);
if (!ttDir.mkdirs()) {
if (!ttDir.isDirectory()) {
throw new IOException("Mkdirs failed to create " + ttDir);
}
}
localDirs[i] = ttDir.toString();
if (i != 0) {
localPath.append(",");
}
localPath.append(localDirs[i]);
}
conf.set("mapred.local.dir", localPath.toString());
LOG.info("mapred.local.dir is " + localPath);
try {
tt =
ugi.doAs(
new PrivilegedExceptionAction<TaskTracker>() {
public TaskTracker run() throws InterruptedException, IOException {
return createTaskTracker(conf);
}
});
isInitialized = true;
} catch (Throwable e) {
isDead = true;
tt = null;
LOG.error("task tracker " + trackerId + " crashed", e);
}
}
public InputSplit[] getSplits(JobConf job, int numSplits) throws IOException {
Path src =
new Path(
job.get(
org.apache.hadoop.mapreduce.GenericMRLoadGenerator.INDIRECT_INPUT_FILE, null));
FileSystem fs = src.getFileSystem(job);
ArrayList<IndirectSplit> splits = new ArrayList<IndirectSplit>(numSplits);
LongWritable key = new LongWritable();
Text value = new Text();
for (SequenceFile.Reader sl = new SequenceFile.Reader(fs, src, job); sl.next(key, value); ) {
splits.add(new IndirectSplit(new Path(value.toString()), key.get()));
}
return splits.toArray(new IndirectSplit[splits.size()]);
}
/**
* @param taskid
* @param workDir
* @return
* @throws IOException
*/
static Path createChildTmpDir(File workDir, JobConf conf) throws IOException {
// add java.io.tmpdir given by mapred.child.tmp
String tmp = conf.get("mapred.child.tmp", "./tmp");
Path tmpDir = new Path(tmp);
// if temp directory path is not absolute, prepend it with workDir.
if (!tmpDir.isAbsolute()) {
tmpDir = new Path(workDir.toString(), tmp);
FileSystem localFs = FileSystem.getLocal(conf);
if (!localFs.mkdirs(tmpDir) && !localFs.getFileStatus(tmpDir).isDir()) {
throw new IOException("Mkdirs failed to create " + tmpDir.toString());
}
}
return tmpDir;
}
/**
* Creates the configuration for mumak simulation. This is kept modular mostly for testing
* purposes. so that the standard configuration can be modified before passing it to the init()
* function.
*
* @return JobConf: the configuration for the SimulatorJobTracker
*/
JobConf createMumakConf() {
JobConf jobConf = new JobConf(getConf());
jobConf.setClass(
"topology.node.switch.mapping.impl", StaticMapping.class, DNSToSwitchMapping.class);
jobConf.set("fs.default.name", "file:///");
jobConf.set("mapred.job.tracker", "localhost:8012");
jobConf.setInt("mapred.jobtracker.job.history.block.size", 512);
jobConf.setInt("mapred.jobtracker.job.history.buffer.size", 512);
jobConf.setLong("mapred.tasktracker.expiry.interval", 5000);
jobConf.setInt("mapred.reduce.copy.backoff", 4);
jobConf.setLong("mapred.job.reuse.jvm.num.tasks", -1);
jobConf.setUser("mumak");
jobConf.set(
"mapred.system.dir",
jobConf.get("hadoop.log.dir", "/tmp/hadoop-" + jobConf.getUser()) + "/mapred/system");
return jobConf;
}
public void configure(JobConf job) {
// 'key' == sortInput for sort-input; key == sortOutput for sort-output
key = deduceInputFile(job);
if (key == sortOutput) {
partitioner = new HashPartitioner<WritableComparable, Writable>();
// Figure the 'current' partition and no. of reduces of the 'sort'
try {
URI inputURI = new URI(job.get("map.input.file"));
String inputFile = inputURI.getPath();
partition =
Integer.valueOf(inputFile.substring(inputFile.lastIndexOf("part") + 5)).intValue();
noSortReducers = job.getInt("sortvalidate.sort.reduce.tasks", -1);
} catch (Exception e) {
System.err.println("Caught: " + e);
System.exit(-1);
}
}
}
/**
* Validates map phase progress after each record is processed by map task using custom task
* reporter.
*/
public void testMapProgress() throws Exception {
JobConf job = new JobConf();
fs = FileSystem.getLocal(job);
Path rootDir = new Path(TEST_ROOT_DIR);
createInputFile(rootDir);
job.setNumReduceTasks(0);
TaskAttemptID taskId = TaskAttemptID.forName("attempt_200907082313_0424_m_000000_0");
job.setClass("mapreduce.job.outputformat.class", NullOutputFormat.class, OutputFormat.class);
job.set(org.apache.hadoop.mapreduce.lib.input.FileInputFormat.INPUT_DIR, TEST_ROOT_DIR);
jobId = taskId.getJobID();
JobContext jContext = new JobContextImpl(job, jobId);
InputFormat<?, ?> input = ReflectionUtils.newInstance(jContext.getInputFormatClass(), job);
List<InputSplit> splits = input.getSplits(jContext);
JobSplitWriter.createSplitFiles(
new Path(TEST_ROOT_DIR), job, new Path(TEST_ROOT_DIR).getFileSystem(job), splits);
TaskSplitMetaInfo[] splitMetaInfo =
SplitMetaInfoReader.readSplitMetaInfo(jobId, fs, job, new Path(TEST_ROOT_DIR));
job.setUseNewMapper(true); // use new api
for (int i = 0; i < splitMetaInfo.length; i++) { // rawSplits.length is 1
map =
new TestMapTask(
job.get(JTConfig.JT_SYSTEM_DIR, "/tmp/hadoop/mapred/system") + jobId + "job.xml",
taskId,
i,
splitMetaInfo[i].getSplitIndex(),
1);
JobConf localConf = new JobConf(job);
map.localizeConfiguration(localConf);
map.setConf(localConf);
map.run(localConf, fakeUmbilical);
}
// clean up
fs.delete(rootDir, true);
}
/**
* Map method. Copies one file from source file system to destination.
*
* @param key src len
* @param value FilePair (FileStatus src, Path dst)
* @param out Log of failed copies
* @param reporter
*/
public void map(
LongWritable key,
FilePair value,
OutputCollector<WritableComparable<?>, Text> out,
Reporter reporter)
throws IOException {
final FileStatus srcstat = value.input;
final Path relativedst = new Path(value.output);
try {
copy(srcstat, relativedst, out, reporter);
} catch (IOException e) {
++failcount;
reporter.incrCounter(Counter.FAIL, 1);
updateStatus(reporter);
final String sfailure = "FAIL " + relativedst + " : " + StringUtils.stringifyException(e);
out.collect(null, new Text(sfailure));
LOG.info(sfailure);
try {
for (int i = 0; i < 3; ++i) {
try {
final Path tmp = new Path(job.get(TMP_DIR_LABEL), relativedst);
if (destFileSys.delete(tmp, true)) break;
} catch (Throwable ex) {
// ignore, we are just cleaning up
LOG.debug("Ignoring cleanup exception", ex);
}
// update status, so we don't get timed out
updateStatus(reporter);
Thread.sleep(3 * 1000);
}
} catch (InterruptedException inte) {
throw (IOException) new IOException().initCause(inte);
}
} finally {
updateStatus(reporter);
}
}
@Override
public final void run() {
try {
// before preparing the job localize
// all the archives
TaskAttemptID taskid = t.getTaskID();
LocalDirAllocator lDirAlloc = new LocalDirAllocator("mapred.local.dir");
File jobCacheDir = null;
if (conf.getJar() != null) {
jobCacheDir = new File(new Path(conf.getJar()).getParent().toString());
}
File workDir =
new File(
lDirAlloc
.getLocalPathToRead(
TaskTracker.getJobCacheSubdir()
+ Path.SEPARATOR
+ t.getJobID()
+ Path.SEPARATOR
+ t.getTaskID()
+ Path.SEPARATOR
+ MRConstants.WORKDIR,
conf)
.toString());
URI[] archives = DistributedCache.getCacheArchives(conf);
URI[] files = DistributedCache.getCacheFiles(conf);
FileStatus fileStatus;
FileSystem fileSystem;
Path localPath;
String baseDir;
if ((archives != null) || (files != null)) {
if (archives != null) {
String[] archivesTimestamps = DistributedCache.getArchiveTimestamps(conf);
Path[] p = new Path[archives.length];
for (int i = 0; i < archives.length; i++) {
fileSystem = FileSystem.get(archives[i], conf);
fileStatus = fileSystem.getFileStatus(new Path(archives[i].getPath()));
String cacheId = DistributedCache.makeRelative(archives[i], conf);
String cachePath = TaskTracker.getCacheSubdir() + Path.SEPARATOR + cacheId;
if (lDirAlloc.ifExists(cachePath, conf)) {
localPath = lDirAlloc.getLocalPathToRead(cachePath, conf);
} else {
localPath = lDirAlloc.getLocalPathForWrite(cachePath, fileStatus.getLen(), conf);
}
baseDir = localPath.toString().replace(cacheId, "");
p[i] =
DistributedCache.getLocalCache(
archives[i],
conf,
new Path(baseDir),
fileStatus,
true,
Long.parseLong(archivesTimestamps[i]),
new Path(workDir.getAbsolutePath()),
false);
}
DistributedCache.setLocalArchives(conf, stringifyPathArray(p));
}
if ((files != null)) {
String[] fileTimestamps = DistributedCache.getFileTimestamps(conf);
Path[] p = new Path[files.length];
for (int i = 0; i < files.length; i++) {
fileSystem = FileSystem.get(files[i], conf);
fileStatus = fileSystem.getFileStatus(new Path(files[i].getPath()));
String cacheId = DistributedCache.makeRelative(files[i], conf);
String cachePath = TaskTracker.getCacheSubdir() + Path.SEPARATOR + cacheId;
if (lDirAlloc.ifExists(cachePath, conf)) {
localPath = lDirAlloc.getLocalPathToRead(cachePath, conf);
} else {
localPath = lDirAlloc.getLocalPathForWrite(cachePath, fileStatus.getLen(), conf);
}
baseDir = localPath.toString().replace(cacheId, "");
p[i] =
DistributedCache.getLocalCache(
files[i],
conf,
new Path(baseDir),
fileStatus,
false,
Long.parseLong(fileTimestamps[i]),
new Path(workDir.getAbsolutePath()),
false);
}
DistributedCache.setLocalFiles(conf, stringifyPathArray(p));
}
Path localTaskFile = new Path(t.getJobFile());
FileSystem localFs = FileSystem.getLocal(conf);
localFs.delete(localTaskFile, true);
OutputStream out = localFs.create(localTaskFile);
try {
conf.writeXml(out);
} finally {
out.close();
}
}
if (!prepare()) {
return;
}
String sep = System.getProperty("path.separator");
StringBuffer classPath = new StringBuffer();
// start with same classpath as parent process
classPath.append(System.getProperty("java.class.path"));
classPath.append(sep);
if (!workDir.mkdirs()) {
if (!workDir.isDirectory()) {
LOG.fatal("Mkdirs failed to create " + workDir.toString());
}
}
String jar = conf.getJar();
if (jar != null) {
// if jar exists, it into workDir
File[] libs = new File(jobCacheDir, "lib").listFiles();
if (libs != null) {
for (int i = 0; i < libs.length; i++) {
classPath.append(sep); // add libs from jar to classpath
classPath.append(libs[i]);
}
}
classPath.append(sep);
classPath.append(new File(jobCacheDir, "classes"));
classPath.append(sep);
classPath.append(jobCacheDir);
}
// include the user specified classpath
// archive paths
Path[] archiveClasspaths = DistributedCache.getArchiveClassPaths(conf);
if (archiveClasspaths != null && archives != null) {
Path[] localArchives = DistributedCache.getLocalCacheArchives(conf);
if (localArchives != null) {
for (int i = 0; i < archives.length; i++) {
for (int j = 0; j < archiveClasspaths.length; j++) {
if (archives[i].getPath().equals(archiveClasspaths[j].toString())) {
classPath.append(sep);
classPath.append(localArchives[i].toString());
}
}
}
}
}
// file paths
Path[] fileClasspaths = DistributedCache.getFileClassPaths(conf);
if (fileClasspaths != null && files != null) {
Path[] localFiles = DistributedCache.getLocalCacheFiles(conf);
if (localFiles != null) {
for (int i = 0; i < files.length; i++) {
for (int j = 0; j < fileClasspaths.length; j++) {
if (files[i].getPath().equals(fileClasspaths[j].toString())) {
classPath.append(sep);
classPath.append(localFiles[i].toString());
}
}
}
}
}
classPath.append(sep);
classPath.append(workDir);
// Build exec child jmv args.
Vector<String> vargs = new Vector<String>(8);
File jvm = // use same jvm as parent
new File(new File(System.getProperty("java.home"), "bin"), "java");
vargs.add(jvm.toString());
// Add child (task) java-vm options.
//
// The following symbols if present in mapred.child.java.opts value are
// replaced:
// + @taskid@ is interpolated with value of TaskID.
// Other occurrences of @ will not be altered.
//
// Example with multiple arguments and substitutions, showing
// jvm GC logging, and start of a passwordless JVM JMX agent so can
// connect with jconsole and the likes to watch child memory, threads
// and get thread dumps.
//
// <property>
// <name>mapred.child.java.opts</name>
// <value>-verbose:gc -Xloggc:/tmp/@[email protected] \
// -Dcom.sun.management.jmxremote.authenticate=false \
// -Dcom.sun.management.jmxremote.ssl=false \
// </value>
// </property>
//
String javaOpts = conf.get("mapred.child.java.opts", "-Xmx200m");
javaOpts = javaOpts.replace("@taskid@", taskid.toString());
String[] javaOptsSplit = javaOpts.split(" ");
// Add java.library.path; necessary for loading native libraries.
//
// 1. To support native-hadoop library i.e. libhadoop.so, we add the
// parent processes' java.library.path to the child.
// 2. We also add the 'cwd' of the task to it's java.library.path to help
// users distribute native libraries via the DistributedCache.
// 3. The user can also specify extra paths to be added to the
// java.library.path via mapred.child.java.opts.
//
String libraryPath = System.getProperty("java.library.path");
if (libraryPath == null) {
libraryPath = workDir.getAbsolutePath();
} else {
libraryPath += sep + workDir;
}
boolean hasUserLDPath = false;
for (int i = 0; i < javaOptsSplit.length; i++) {
if (javaOptsSplit[i].startsWith("-Djava.library.path=")) {
javaOptsSplit[i] += sep + libraryPath;
hasUserLDPath = true;
break;
}
}
if (!hasUserLDPath) {
vargs.add("-Djava.library.path=" + libraryPath);
}
for (int i = 0; i < javaOptsSplit.length; i++) {
vargs.add(javaOptsSplit[i]);
}
// add java.io.tmpdir given by mapred.child.tmp
String tmp = conf.get("mapred.child.tmp", "./tmp");
Path tmpDir = new Path(tmp);
// if temp directory path is not absolute
// prepend it with workDir.
if (!tmpDir.isAbsolute()) {
tmpDir = new Path(workDir.toString(), tmp);
}
FileSystem localFs = FileSystem.getLocal(conf);
if (!localFs.mkdirs(tmpDir) && !localFs.getFileStatus(tmpDir).isDir()) {
throw new IOException("Mkdirs failed to create " + tmpDir.toString());
}
vargs.add("-Djava.io.tmpdir=" + tmpDir.toString());
// Add classpath.
vargs.add("-classpath");
vargs.add(classPath.toString());
// Setup the log4j prop
long logSize = TaskLog.getTaskLogLength(conf);
vargs.add(
"-Dhadoop.log.dir=" + new File(System.getProperty("hadoop.log.dir")).getAbsolutePath());
vargs.add("-Dhadoop.root.logger=INFO,TLA");
vargs.add("-Dhadoop.tasklog.taskid=" + taskid);
vargs.add("-Dhadoop.tasklog.totalLogFileSize=" + logSize);
if (conf.getProfileEnabled()) {
if (conf.getProfileTaskRange(t.isMapTask()).isIncluded(t.getPartition())) {
File prof = TaskLog.getTaskLogFile(taskid, TaskLog.LogName.PROFILE);
vargs.add(String.format(conf.getProfileParams(), prof.toString()));
}
}
// Add main class and its arguments
vargs.add(Child.class.getName()); // main of Child
// pass umbilical address
InetSocketAddress address = tracker.getTaskTrackerReportAddress();
vargs.add(address.getAddress().getHostAddress());
vargs.add(Integer.toString(address.getPort()));
vargs.add(taskid.toString()); // pass task identifier
String pidFile = null;
if (tracker.isTaskMemoryManagerEnabled()) {
pidFile =
lDirAlloc
.getLocalPathForWrite(
(TaskTracker.getPidFilesSubdir() + Path.SEPARATOR + taskid), this.conf)
.toString();
}
// set memory limit using ulimit if feasible and necessary ...
String[] ulimitCmd = Shell.getUlimitMemoryCommand(conf);
List<String> setup = null;
if (ulimitCmd != null) {
setup = new ArrayList<String>();
for (String arg : ulimitCmd) {
setup.add(arg);
}
}
// Set up the redirection of the task's stdout and stderr streams
File stdout = TaskLog.getTaskLogFile(taskid, TaskLog.LogName.STDOUT);
File stderr = TaskLog.getTaskLogFile(taskid, TaskLog.LogName.STDERR);
stdout.getParentFile().mkdirs();
tracker.getTaskTrackerInstrumentation().reportTaskLaunch(taskid, stdout, stderr);
Map<String, String> env = new HashMap<String, String>();
StringBuffer ldLibraryPath = new StringBuffer();
ldLibraryPath.append(workDir.toString());
String oldLdLibraryPath = null;
oldLdLibraryPath = System.getenv("LD_LIBRARY_PATH");
if (oldLdLibraryPath != null) {
ldLibraryPath.append(sep);
ldLibraryPath.append(oldLdLibraryPath);
}
env.put("LD_LIBRARY_PATH", ldLibraryPath.toString());
jvmManager.launchJvm(
this,
jvmManager.constructJvmEnv(
setup, vargs, stdout, stderr, logSize, workDir, env, pidFile, conf));
synchronized (lock) {
while (!done) {
lock.wait();
}
}
tracker.getTaskTrackerInstrumentation().reportTaskEnd(t.getTaskID());
if (exitCodeSet) {
if (!killed && exitCode != 0) {
if (exitCode == 65) {
tracker.getTaskTrackerInstrumentation().taskFailedPing(t.getTaskID());
}
throw new IOException("Task process exit with nonzero status of " + exitCode + ".");
}
}
} catch (FSError e) {
LOG.fatal("FSError", e);
try {
tracker.fsError(t.getTaskID(), e.getMessage());
} catch (IOException ie) {
LOG.fatal(t.getTaskID() + " reporting FSError", ie);
}
} catch (Throwable throwable) {
LOG.warn(t.getTaskID() + " Child Error", throwable);
ByteArrayOutputStream baos = new ByteArrayOutputStream();
throwable.printStackTrace(new PrintStream(baos));
try {
tracker.reportDiagnosticInfo(t.getTaskID(), baos.toString());
} catch (IOException e) {
LOG.warn(t.getTaskID() + " Reporting Diagnostics", e);
}
} finally {
try {
URI[] archives = DistributedCache.getCacheArchives(conf);
URI[] files = DistributedCache.getCacheFiles(conf);
if (archives != null) {
for (int i = 0; i < archives.length; i++) {
DistributedCache.releaseCache(archives[i], conf);
}
}
if (files != null) {
for (int i = 0; i < files.length; i++) {
DistributedCache.releaseCache(files[i], conf);
}
}
} catch (IOException ie) {
LOG.warn("Error releasing caches : Cache files might not have been cleaned up");
}
tracker.reportTaskFinished(t.getTaskID(), false);
if (t.isMapTask()) {
tracker.addFreeMapSlot();
} else {
tracker.addFreeReduceSlot();
}
}
}
private String getCJTJavaOpts(JobConf conf) {
return conf.get("mapred.corona.standalonecjt.java.opts", JobConf.DEFAULT_MAPRED_TASK_JAVA_OPTS);
}
@Override
public void configure(JobConf conf) {
TaskAttemptID id = TaskAttemptID.forName(conf.get("mapred.task.id"));
shouldFail = id.getId() == 0 && id.getTaskID().getId() == 0;
}
/**
* Get the environment variables for the child map/reduce tasks.
*
* @param jobConf job configuration
* @return the environment variables for the child map/reduce tasks or <code>null</code> if
* unspecified
* @deprecated Use environment variables specific to the map or reduce tasks set via {@link
* JobConf#MAPRED_MAP_TASK_ENV} or {@link JobConf#MAPRED_REDUCE_TASK_ENV}
*/
public String getChildEnv(JobConf jobConf) {
return jobConf.get(JobConf.MAPRED_TASK_ENV);
}
private String getVMEnvironment(
String errorInfo,
File workDir,
JobConf conf,
Map<String, String> env,
TaskAttemptID taskid,
long logSize)
throws Throwable {
StringBuffer ldLibraryPath = new StringBuffer();
ldLibraryPath.append(workDir.toString());
String oldLdLibraryPath = null;
oldLdLibraryPath = System.getenv("LD_LIBRARY_PATH");
if (oldLdLibraryPath != null) {
ldLibraryPath.append(SYSTEM_PATH_SEPARATOR);
ldLibraryPath.append(oldLdLibraryPath);
}
env.put("LD_LIBRARY_PATH", ldLibraryPath.toString());
env.put(HADOOP_WORK_DIR, workDir.toString());
// put jobTokenFile name into env
String jobTokenFile = conf.get(TokenCache.JOB_TOKENS_FILENAME);
LOG.debug("putting jobToken file name into environment " + jobTokenFile);
env.put(UserGroupInformation.HADOOP_TOKEN_FILE_LOCATION, jobTokenFile);
// for the child of task jvm, set hadoop.root.logger
env.put("HADOOP_ROOT_LOGGER", "INFO,TLA");
String hadoopClientOpts = System.getenv("HADOOP_CLIENT_OPTS");
if (hadoopClientOpts == null) {
hadoopClientOpts = "";
} else {
hadoopClientOpts = hadoopClientOpts + " ";
}
hadoopClientOpts =
hadoopClientOpts
+ "-Dhadoop.tasklog.taskid="
+ taskid
+ " -Dhadoop.tasklog.iscleanup="
+ t.isTaskCleanupTask()
+ " -Dhadoop.tasklog.totalLogFileSize="
+ logSize;
// following line is a backport from jira MAPREDUCE-1286
env.put("HADOOP_CLIENT_OPTS", hadoopClientOpts);
// add the env variables passed by the user
String mapredChildEnv = getChildEnv(conf);
if (mapredChildEnv != null && mapredChildEnv.length() > 0) {
String childEnvs[] = mapredChildEnv.split(",");
for (String cEnv : childEnvs) {
try {
String[] parts = cEnv.split("="); // split on '='
String value = env.get(parts[0]);
if (value != null) {
// replace $env with the child's env constructed by tt's
// example LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/tmp
value = parts[1].replace("$" + parts[0], value);
} else {
// this key is not configured by the tt for the child .. get it
// from the tt's env
// example PATH=$PATH:/tmp
value = System.getenv(parts[0]);
if (value != null) {
// the env key is present in the tt's env
value = parts[1].replace("$" + parts[0], value);
} else {
// the env key is note present anywhere .. simply set it
// example X=$X:/tmp or X=/tmp
value = parts[1].replace("$" + parts[0], "");
}
}
env.put(parts[0], value);
} catch (Throwable t) {
// set the error msg
errorInfo =
"Invalid User environment settings : "
+ mapredChildEnv
+ ". Failed to parse user-passed environment param."
+ " Expecting : env1=value1,env2=value2...";
LOG.warn(errorInfo);
throw t;
}
}
}
return errorInfo;
}
/**
* Copy a file to a destination.
*
* @param srcstat src path and metadata
* @param dstpath dst path
* @param reporter
*/
private void copy(
FileStatus srcstat,
Path relativedst,
OutputCollector<WritableComparable<?>, Text> outc,
Reporter reporter)
throws IOException {
Path absdst = new Path(destPath, relativedst);
int totfiles = job.getInt(SRC_COUNT_LABEL, -1);
assert totfiles >= 0 : "Invalid file count " + totfiles;
// if a directory, ensure created even if empty
if (srcstat.isDir()) {
if (destFileSys.exists(absdst)) {
if (!destFileSys.getFileStatus(absdst).isDir()) {
throw new IOException("Failed to mkdirs: " + absdst + " is a file.");
}
} else if (!destFileSys.mkdirs(absdst)) {
throw new IOException("Failed to mkdirs " + absdst);
}
// TODO: when modification times can be set, directories should be
// emitted to reducers so they might be preserved. Also, mkdirs does
// not currently return an error when the directory already exists;
// if this changes, all directory work might as well be done in reduce
return;
}
if (destFileSys.exists(absdst) && !overwrite && !needsUpdate(srcstat, destFileSys, absdst)) {
outc.collect(null, new Text("SKIP: " + srcstat.getPath()));
++skipcount;
reporter.incrCounter(Counter.SKIP, 1);
updateStatus(reporter);
return;
}
Path tmpfile = new Path(job.get(TMP_DIR_LABEL), relativedst);
long cbcopied = 0L;
FSDataInputStream in = null;
FSDataOutputStream out = null;
try {
// open src file
try {
in = srcstat.getPath().getFileSystem(job).open(srcstat.getPath());
} catch (IOException e) {
LOG.error("Failed to open src file " + srcstat.getPath() + ", ignore and return");
in = null;
return;
}
reporter.incrCounter(Counter.BYTESEXPECTED, srcstat.getLen());
// open tmp file
out = create(tmpfile, reporter, srcstat);
// copy file
for (int cbread; (cbread = in.read(buffer)) >= 0; ) {
out.write(buffer, 0, cbread);
cbcopied += cbread;
reporter.setStatus(
String.format("%.2f ", cbcopied * 100.0 / srcstat.getLen())
+ absdst
+ " [ "
+ StringUtils.humanReadableInt(cbcopied)
+ " / "
+ StringUtils.humanReadableInt(srcstat.getLen())
+ " ]");
}
} finally {
checkAndClose(in);
checkAndClose(out);
}
if (cbcopied != srcstat.getLen()) {
if (srcstat.getLen() == 0 && cbcopied > 0) {
LOG.info("most likely see a WAL file corruption: " + srcstat.getPath());
} else {
throw new IOException(
"File size not matched: copied "
+ bytesString(cbcopied)
+ " to tmpfile (="
+ tmpfile
+ ") but expected "
+ bytesString(srcstat.getLen())
+ " from "
+ srcstat.getPath());
}
} else {
if (totfiles == 1) {
// Copying a single file; use dst path provided by user as destination
// rather than destination directory, if a file
Path dstparent = absdst.getParent();
if (!(destFileSys.exists(dstparent) && destFileSys.getFileStatus(dstparent).isDir())) {
absdst = dstparent;
}
}
if (destFileSys.exists(absdst) && destFileSys.getFileStatus(absdst).isDir()) {
throw new IOException(absdst + " is a directory");
}
if (!destFileSys.mkdirs(absdst.getParent())) {
throw new IOException("Failed to craete parent dir: " + absdst.getParent());
}
rename(tmpfile, absdst);
FileStatus dststat = destFileSys.getFileStatus(absdst);
if (dststat.getLen() != srcstat.getLen()) {
destFileSys.delete(absdst, false);
throw new IOException(
"File size not matched: copied "
+ bytesString(dststat.getLen())
+ " to dst (="
+ absdst
+ ") but expected "
+ bytesString(srcstat.getLen())
+ " from "
+ srcstat.getPath());
}
updatePermissions(srcstat, dststat);
}
// report at least once for each file
++copycount;
reporter.incrCounter(Counter.BYTESCOPIED, cbcopied);
reporter.incrCounter(Counter.COPY, 1);
updateStatus(reporter);
}
/**
* Initiate components in the simulation. The JobConf is create separately and passed to the
* init().
*
* @param JobConf: The configuration for the jobtracker.
* @throws InterruptedException
* @throws IOException if trace or topology files cannot be opened.
*/
@SuppressWarnings("deprecation")
void init(JobConf jobConf) throws InterruptedException, IOException {
FileSystem lfs = FileSystem.getLocal(getConf());
Path logPath = new Path(System.getProperty("hadoop.log.dir")).makeQualified(lfs);
jobConf.set("mapred.system.dir", logPath.toString());
jobConf.set("hadoop.job.history.location", (new Path(logPath, "history").toString()));
// start time for virtual clock
// possible improvement: set default value to sth more meaningful based on
// the 1st job
long now = getTimeProperty(jobConf, "mumak.start.time", System.currentTimeMillis());
jt = SimulatorJobTracker.startTracker(jobConf, now, this);
jt.offerService();
masterRandomSeed = jobConf.getLong("mumak.random.seed", System.nanoTime());
// max Map/Reduce tasks per node
int maxMaps =
getConf()
.getInt("mapred.tasktracker.map.tasks.maximum", SimulatorTaskTracker.DEFAULT_MAP_SLOTS);
int maxReduces =
getConf()
.getInt(
"mapred.tasktracker.reduce.tasks.maximum",
SimulatorTaskTracker.DEFAULT_REDUCE_SLOTS);
MachineNode defaultNode =
new MachineNode.Builder("default", 2)
.setMapSlots(maxMaps)
.setReduceSlots(maxReduces)
.build();
LoggedNetworkTopology topology =
new ClusterTopologyReader(new Path(topologyFile), jobConf).get();
// Setting the static mapping before removing numeric IP hosts.
setStaticMapping(topology);
if (getConf().getBoolean("mumak.topology.filter-numeric-ips", true)) {
removeIpHosts(topology);
}
ZombieCluster cluster = new ZombieCluster(topology, defaultNode);
// create TTs based on topology.json
long firstJobStartTime = startTaskTrackers(cluster, jobConf, now);
long subRandomSeed =
RandomSeedGenerator.getSeed("forSimulatorJobStoryProducer", masterRandomSeed);
JobStoryProducer jobStoryProducer =
new SimulatorJobStoryProducer(
new Path(traceFile), cluster, firstJobStartTime, jobConf, subRandomSeed);
final SimulatorJobSubmissionPolicy submissionPolicy =
SimulatorJobSubmissionPolicy.getPolicy(jobConf);
jc = new SimulatorJobClient(jt, jobStoryProducer, submissionPolicy);
queue.addAll(jc.init(firstJobStartTime));
// if the taskScheduler is CapacityTaskScheduler start off the JobInitialization
// threads too
if (jobConf
.get("mapred.jobtracker.taskScheduler")
.equals(CapacityTaskScheduler.class.getName())) {
LOG.info("CapacityScheduler used: starting simulatorThreads");
startSimulatorThreadsCapSched(now);
}
terminateTime = getTimeProperty(jobConf, "mumak.terminate.time", Long.MAX_VALUE);
}
/**
* Get the java command line options for the child map/reduce tasks.
*
* @param jobConf job configuration
* @param defaultValue default value
* @return the java command line options for child map/reduce tasks
* @deprecated Use command line options specific to map or reduce tasks set via {@link
* JobConf#MAPRED_MAP_TASK_JAVA_OPTS} or {@link JobConf#MAPRED_REDUCE_TASK_JAVA_OPTS}
*/
@Deprecated
public String getChildJavaOpts(JobConf jobConf, String defaultValue) {
return jobConf.get(JobConf.MAPRED_TASK_JAVA_OPTS, defaultValue);
}
static void checkRecords(
Configuration defaults, int noMaps, int noReduces, Path sortInput, Path sortOutput)
throws IOException {
JobConf jobConf = new JobConf(defaults, RecordChecker.class);
jobConf.setJobName("sortvalidate-record-checker");
jobConf.setInputFormat(SequenceFileInputFormat.class);
jobConf.setOutputFormat(SequenceFileOutputFormat.class);
jobConf.setOutputKeyClass(BytesWritable.class);
jobConf.setOutputValueClass(IntWritable.class);
jobConf.setMapperClass(Map.class);
jobConf.setReducerClass(Reduce.class);
JobClient client = new JobClient(jobConf);
ClusterStatus cluster = client.getClusterStatus();
if (noMaps == -1) {
noMaps = cluster.getTaskTrackers() * jobConf.getInt("test.sortvalidate.maps_per_host", 10);
}
if (noReduces == -1) {
noReduces = (int) (cluster.getMaxReduceTasks() * 0.9);
String sortReduces = jobConf.get("test.sortvalidate.reduces_per_host");
if (sortReduces != null) {
noReduces = cluster.getTaskTrackers() * Integer.parseInt(sortReduces);
}
}
jobConf.setNumMapTasks(noMaps);
jobConf.setNumReduceTasks(noReduces);
FileInputFormat.setInputPaths(jobConf, sortInput);
FileInputFormat.addInputPath(jobConf, sortOutput);
Path outputPath = new Path("/tmp/sortvalidate/recordchecker");
FileSystem fs = FileSystem.get(defaults);
if (fs.exists(outputPath)) {
fs.delete(outputPath, true);
}
FileOutputFormat.setOutputPath(jobConf, outputPath);
// Uncomment to run locally in a single process
// job_conf.set("mapred.job.tracker", "local");
Path[] inputPaths = FileInputFormat.getInputPaths(jobConf);
System.out.println(
"\nSortValidator.RecordChecker: Running on "
+ cluster.getTaskTrackers()
+ " nodes to validate sort from "
+ inputPaths[0]
+ ", "
+ inputPaths[1]
+ " into "
+ FileOutputFormat.getOutputPath(jobConf)
+ " with "
+ noReduces
+ " reduces.");
Date startTime = new Date();
System.out.println("Job started: " + startTime);
JobClient.runJob(jobConf);
Date end_time = new Date();
System.out.println("Job ended: " + end_time);
System.out.println(
"The job took " + (end_time.getTime() - startTime.getTime()) / 1000 + " seconds.");
}
public void configure(JobConf job) {
this.inputArgs = Utils.asArray(job.get("user.args"));
}
/** @see org.apache.hadoop.mapred.JobSubmissionProtocol#getSystemDir() */
public String getSystemDir() {
Path sysDir = new Path(conf.get("mapred.system.dir", "/tmp/hadoop/mapred/system"));
return fs.makeQualified(sysDir).toString();
}
@Override
public void configure(JobConf config) {
splitChar = (char) NumberUtils.toInt(config.get("user.split"));
}
