public JobBuilder compressType(CompressionType type) throws IOException {
if (type == CompressionType.NONE) {
_jobConf.setBoolean("mapred.output.compress", false);
} else {
_jobConf.setBoolean("mapred.output.compress", true);
}
_jobConf.set("mapred.output.compression.type", type.toString());
return this;
}
private JobConf createJobConf() {
JobConf job = new NutchJob(getConf());
job.setBoolean("segment.reader.co", this.co);
job.setBoolean("segment.reader.fe", this.fe);
job.setBoolean("segment.reader.ge", this.ge);
job.setBoolean("segment.reader.pa", this.pa);
job.setBoolean("segment.reader.pd", this.pd);
job.setBoolean("segment.reader.pt", this.pt);
return job;
}
public JobBuilder compressor(CompressionType type, Class<? extends CompressionCodec> codec)
throws IOException {
_jobConf.setBoolean("mapred.output.compress", true);
_jobConf.set("mapred.output.compression.type", type.toString());
_jobConf.setClass("mapred.output.compression.codec", codec, CompressionCodec.class);
return this;
}
/**
* {@inheritDoc}
*
* @see org.apache.hadoop.util.Tool#run(java.lang.String[])
*/
@Override
public int run(String[] args) throws Exception {
JobConf configuration = new JobConf(getConf(), WordCountExtended.class);
configuration.setJobName(JOB_NAME);
configuration.setOutputKeyClass(Text.class);
configuration.setOutputValueClass(IntWritable.class);
configuration.setMapperClass(Map.class);
configuration.setCombinerClass(Reduce.class);
configuration.setReducerClass(Reduce.class);
configuration.setInputFormat(TextInputFormat.class);
configuration.setOutputFormat(TextOutputFormat.class);
List<String> otherArgs = new ArrayList<String>();
for (int i = 0; i < args.length; ++i) {
if (JOB_SKIP_ARGUMENT.equals(args[i])) {
DistributedCache.addCacheFile(new Path(args[++i]).toUri(), configuration);
configuration.setBoolean(JOB_PARAMETER_SKIP_PATTERNS, true);
} else {
otherArgs.add(args[i]);
}
}
FileInputFormat.setInputPaths(configuration, new Path(otherArgs.get(0)));
FileOutputFormat.setOutputPath(configuration, new Path(otherArgs.get(1)));
JobClient.runJob(configuration);
return 0;
}
@Test
public void testEncryptedMerger() throws Throwable {
jobConf.setBoolean(MRJobConfig.MR_ENCRYPTED_INTERMEDIATE_DATA, true);
conf.setBoolean(MRJobConfig.MR_ENCRYPTED_INTERMEDIATE_DATA, true);
Credentials credentials = UserGroupInformation.getCurrentUser().getCredentials();
TokenCache.setShuffleSecretKey(new byte[16], credentials);
UserGroupInformation.getCurrentUser().addCredentials(credentials);
testInMemoryAndOnDiskMerger();
}
/**
* Creates and initializes a JobConf object that can be used to execute the DAG. The configuration
* object will contain configurations from mapred-site overlaid with key/value pairs from the
* hiveConf object. Finally it will also contain some hive specific configurations that do not
* change from DAG to DAG.
*
* @param hiveConf Current hiveConf for the execution
* @return JobConf base configuration for job execution
* @throws IOException
*/
public JobConf createConfiguration(HiveConf hiveConf) throws IOException {
hiveConf.setBoolean("mapred.mapper.new-api", false);
JobConf conf = (JobConf) MRHelpers.getBaseMRConfiguration(hiveConf);
conf.set("mapred.output.committer.class", NullOutputCommitter.class.getName());
conf.setBoolean("mapred.committer.job.setup.cleanup.needed", false);
conf.setBoolean("mapred.committer.job.task.cleanup.needed", false);
conf.setClass("mapred.output.format.class", HiveOutputFormatImpl.class, OutputFormat.class);
conf.set(MRJobConfig.OUTPUT_KEY_CLASS, HiveKey.class.getName());
conf.set(MRJobConfig.OUTPUT_VALUE_CLASS, BytesWritable.class.getName());
conf.set("mapred.partitioner.class", HiveConf.getVar(conf, HiveConf.ConfVars.HIVEPARTITIONER));
conf.set("tez.runtime.partitioner.class", MRPartitioner.class.getName());
return conf;
}
/**
* Test if {@link CompressionEmulationUtil#configureCompressionEmulation(
* org.apache.hadoop.mapred.JobConf, org.apache.hadoop.mapred.JobConf)} can extract compression
* related configuration parameters.
*/
@Test
public void testExtractCompressionConfigs() {
JobConf source = new JobConf();
JobConf target = new JobConf();
// set the default values
source.setBoolean(FileOutputFormat.COMPRESS, false);
source.set(FileOutputFormat.COMPRESS_CODEC, "MyDefaultCodec");
source.set(FileOutputFormat.COMPRESS_TYPE, "MyDefaultType");
source.setBoolean(MRJobConfig.MAP_OUTPUT_COMPRESS, false);
source.set(MRJobConfig.MAP_OUTPUT_COMPRESS_CODEC, "MyDefaultCodec2");
CompressionEmulationUtil.configureCompressionEmulation(source, target);
// check default values
assertFalse(target.getBoolean(FileOutputFormat.COMPRESS, true));
assertEquals("MyDefaultCodec", target.get(FileOutputFormat.COMPRESS_CODEC));
assertEquals("MyDefaultType", target.get(FileOutputFormat.COMPRESS_TYPE));
assertFalse(target.getBoolean(MRJobConfig.MAP_OUTPUT_COMPRESS, true));
assertEquals("MyDefaultCodec2", target.get(MRJobConfig.MAP_OUTPUT_COMPRESS_CODEC));
assertFalse(CompressionEmulationUtil.isInputCompressionEmulationEnabled(target));
// set new values
source.setBoolean(FileOutputFormat.COMPRESS, true);
source.set(FileOutputFormat.COMPRESS_CODEC, "MyCodec");
source.set(FileOutputFormat.COMPRESS_TYPE, "MyType");
source.setBoolean(MRJobConfig.MAP_OUTPUT_COMPRESS, true);
source.set(MRJobConfig.MAP_OUTPUT_COMPRESS_CODEC, "MyCodec2");
org.apache.hadoop.mapred.FileInputFormat.setInputPaths(source, "file.gz");
target = new JobConf(); // reset
CompressionEmulationUtil.configureCompressionEmulation(source, target);
// check new values
assertTrue(target.getBoolean(FileOutputFormat.COMPRESS, false));
assertEquals("MyCodec", target.get(FileOutputFormat.COMPRESS_CODEC));
assertEquals("MyType", target.get(FileOutputFormat.COMPRESS_TYPE));
assertTrue(target.getBoolean(MRJobConfig.MAP_OUTPUT_COMPRESS, false));
assertEquals("MyCodec2", target.get(MRJobConfig.MAP_OUTPUT_COMPRESS_CODEC));
assertTrue(CompressionEmulationUtil.isInputCompressionEmulationEnabled(target));
}
@Override
public int run(String[] args) throws Exception {
final int ret = parseArgs(args);
if (ret < 0) {
return ret;
}
JobConf config = new JobConf(getConf(), TfIdfNovelty.class);
config.setJobName("Influence-TfIdfNovelty");
config.set(Fields.BASIS.get(), basisPath);
if (datesPath != null) {
config.set(Fields.DOC_DATES.get(), datesPath);
}
config.setBoolean(Fields.IGNORE.get(), ignoreDocs);
if (bands > 0) {
config.setInt(Fields.BANDS.get(), bands);
}
if (rows > 0) {
config.setInt(Fields.ROWS.get(), rows);
}
SetupHelper.getInstance()
.setSequenceInput(config, inputPath)
.setSequenceOutput(config, outputPath);
config.setMapOutputKeyClass(HashBandWritable.class);
config.setMapOutputValueClass(DocumentWithVectorWritable.class);
config.setMapperClass(TfIdfNoveltyLshMapper.class);
if (outputBuckets) {
config.setOutputKeyClass(HashBandWritable.class);
config.setOutputValueClass(IntArrayWritable.class);
config.setReducerClass(TfIdfNoveltyIdentityReducer.class);
} else {
config.setOutputKeyClass(Text.class);
config.setOutputValueClass(VectorWritable.class);
config.setReducerClass(TfIdfNoveltyReducer.class);
}
// Delete the output directory if it exists already.
FileSystem.get(getConf()).delete(new Path(outputPath), true);
JobClient.runJob(config);
return 0;
}
public static void seekTest(FileSystem fs, boolean fastCheck) throws Exception {
fs.delete(READ_DIR, true);
JobConf job = new JobConf(conf, TestFileSystem.class);
job.setBoolean("fs.test.fastCheck", fastCheck);
FileInputFormat.setInputPaths(job, CONTROL_DIR);
job.setInputFormat(SequenceFileInputFormat.class);
job.setMapperClass(SeekMapper.class);
job.setReducerClass(LongSumReducer.class);
FileOutputFormat.setOutputPath(job, READ_DIR);
job.setOutputKeyClass(Text.class);
job.setOutputValueClass(LongWritable.class);
job.setNumReduceTasks(1);
JobClient.runJob(job);
}
public void testTotalOrderBinarySearch() throws Exception {
TotalOrderPartitioner<Text, NullWritable> partitioner =
new TotalOrderPartitioner<Text, NullWritable>();
JobConf job = new JobConf();
Path p =
TestTotalOrderPartitioner.<Text>writePartitionFile(
"totalorderbinarysearch", job, splitStrings);
job.setBoolean("total.order.partitioner.natural.order", false);
job.setMapOutputKeyClass(Text.class);
try {
partitioner.configure(job);
NullWritable nw = NullWritable.get();
for (Check<Text> chk : testStrings) {
assertEquals(
chk.data.toString(),
chk.part,
partitioner.getPartition(chk.data, nw, splitStrings.length + 1));
}
} finally {
p.getFileSystem(job).delete(p);
}
}
public void testTotalOrderCustomComparator() throws Exception {
TotalOrderPartitioner<Text, NullWritable> partitioner =
new TotalOrderPartitioner<Text, NullWritable>();
JobConf job = new JobConf();
Text[] revSplitStrings = Arrays.copyOf(splitStrings, splitStrings.length);
Arrays.sort(revSplitStrings, new ReverseStringComparator());
Path p =
TestTotalOrderPartitioner.<Text>writePartitionFile(
"totalordercustomcomparator", job, revSplitStrings);
job.setBoolean("total.order.partitioner.natural.order", false);
job.setMapOutputKeyClass(Text.class);
job.setOutputKeyComparatorClass(ReverseStringComparator.class);
ArrayList<Check<Text>> revCheck = new ArrayList<Check<Text>>();
revCheck.add(new Check<Text>(new Text("aaaaa"), 9));
revCheck.add(new Check<Text>(new Text("aaabb"), 9));
revCheck.add(new Check<Text>(new Text("aabbb"), 9));
revCheck.add(new Check<Text>(new Text("aaaaa"), 9));
revCheck.add(new Check<Text>(new Text("babbb"), 8));
revCheck.add(new Check<Text>(new Text("baabb"), 8));
revCheck.add(new Check<Text>(new Text("yai"), 1));
revCheck.add(new Check<Text>(new Text("yak"), 1));
revCheck.add(new Check<Text>(new Text("z"), 0));
revCheck.add(new Check<Text>(new Text("ddngo"), 4));
revCheck.add(new Check<Text>(new Text("hi"), 3));
try {
partitioner.configure(job);
NullWritable nw = NullWritable.get();
for (Check<Text> chk : revCheck) {
assertEquals(
chk.data.toString(),
chk.part,
partitioner.getPartition(chk.data, nw, splitStrings.length + 1));
}
} finally {
p.getFileSystem(job).delete(p);
}
}
/**
* Test {@link CompressionEmulationUtil#getPossiblyDecompressedInputStream(Path, Configuration,
* long)} and {@link CompressionEmulationUtil#getPossiblyCompressedOutputStream(Path,
* Configuration)}.
*/
@Test
public void testPossiblyCompressedDecompressedStreams() throws IOException {
JobConf conf = new JobConf();
FileSystem lfs = FileSystem.getLocal(conf);
String inputLine = "Hi Hello!";
CompressionEmulationUtil.setCompressionEmulationEnabled(conf, true);
CompressionEmulationUtil.setInputCompressionEmulationEnabled(conf, true);
conf.setBoolean(FileOutputFormat.COMPRESS, true);
conf.setClass(FileOutputFormat.COMPRESS_CODEC, GzipCodec.class, CompressionCodec.class);
// define the test's root temp directory
Path rootTempDir =
new Path(System.getProperty("test.build.data", "/tmp"))
.makeQualified(lfs.getUri(), lfs.getWorkingDirectory());
Path tempDir = new Path(rootTempDir, "TestPossiblyCompressedDecompressedStreams");
lfs.delete(tempDir, true);
// create a compressed file
Path compressedFile = new Path(tempDir, "test");
OutputStream out =
CompressionEmulationUtil.getPossiblyCompressedOutputStream(compressedFile, conf);
BufferedWriter writer = new BufferedWriter(new OutputStreamWriter(out));
writer.write(inputLine);
writer.close();
// now read back the data from the compressed stream
compressedFile = compressedFile.suffix(".gz");
InputStream in =
CompressionEmulationUtil.getPossiblyDecompressedInputStream(compressedFile, conf, 0);
BufferedReader reader = new BufferedReader(new InputStreamReader(in));
String readLine = reader.readLine();
assertEquals("Compression/Decompression error", inputLine, readLine);
reader.close();
}
/**
* Initialize DFSCopyFileMapper specific job-configuration.
*
* @param conf : The dfs/mapred configuration.
* @param jobConf : The handle to the jobConf object to be initialized.
* @param args Arguments
*/
private static void setup(Configuration conf, JobConf jobConf, final Arguments args)
throws IOException {
jobConf.set(DST_DIR_LABEL, args.dst.toUri().toString());
// set boolean values
final boolean update = args.flags.contains(Options.UPDATE);
final boolean overwrite = !update && args.flags.contains(Options.OVERWRITE);
jobConf.setBoolean(Options.UPDATE.propertyname, update);
jobConf.setBoolean(Options.OVERWRITE.propertyname, overwrite);
jobConf.setBoolean(
Options.IGNORE_READ_FAILURES.propertyname,
args.flags.contains(Options.IGNORE_READ_FAILURES));
jobConf.setBoolean(
Options.PRESERVE_STATUS.propertyname, args.flags.contains(Options.PRESERVE_STATUS));
final String randomId = getRandomId();
JobClient jClient = new JobClient(jobConf);
Path jobDirectory = new Path(jClient.getSystemDir(), NAME + "_" + randomId);
jobConf.set(JOB_DIR_LABEL, jobDirectory.toString());
FileSystem dstfs = args.dst.getFileSystem(conf);
boolean dstExists = dstfs.exists(args.dst);
boolean dstIsDir = false;
if (dstExists) {
dstIsDir = dstfs.getFileStatus(args.dst).isDir();
}
// default logPath
Path logPath = args.log;
if (logPath == null) {
String filename = "_distcp_logs_" + randomId;
if (!dstExists || !dstIsDir) {
Path parent = args.dst.getParent();
if (!dstfs.exists(parent)) {
dstfs.mkdirs(parent);
}
logPath = new Path(parent, filename);
} else {
logPath = new Path(args.dst, filename);
}
}
FileOutputFormat.setOutputPath(jobConf, logPath);
// create src list, dst list
FileSystem jobfs = jobDirectory.getFileSystem(jobConf);
Path srcfilelist = new Path(jobDirectory, "_distcp_src_files");
jobConf.set(SRC_LIST_LABEL, srcfilelist.toString());
SequenceFile.Writer src_writer =
SequenceFile.createWriter(
jobfs,
jobConf,
srcfilelist,
LongWritable.class,
FilePair.class,
SequenceFile.CompressionType.NONE);
Path dstfilelist = new Path(jobDirectory, "_distcp_dst_files");
SequenceFile.Writer dst_writer =
SequenceFile.createWriter(
jobfs, jobConf, dstfilelist, Text.class, Text.class, SequenceFile.CompressionType.NONE);
Path dstdirlist = new Path(jobDirectory, "_distcp_dst_dirs");
jobConf.set(DST_DIR_LIST_LABEL, dstdirlist.toString());
SequenceFile.Writer dir_writer =
SequenceFile.createWriter(
jobfs,
jobConf,
dstdirlist,
Text.class,
FilePair.class,
SequenceFile.CompressionType.NONE);
// handle the case where the destination directory doesn't exist
// and we've only a single src directory OR we're updating/overwriting
// the contents of the destination directory.
final boolean special = (args.srcs.size() == 1 && !dstExists) || update || overwrite;
int srcCount = 0, cnsyncf = 0, dirsyn = 0;
long fileCount = 0L, byteCount = 0L, cbsyncs = 0L;
try {
for (Iterator<Path> srcItr = args.srcs.iterator(); srcItr.hasNext(); ) {
final Path src = srcItr.next();
FileSystem srcfs = src.getFileSystem(conf);
FileStatus srcfilestat = srcfs.getFileStatus(src);
Path root = special && srcfilestat.isDir() ? src : src.getParent();
if (srcfilestat.isDir()) {
++srcCount;
}
Stack<FileStatus> pathstack = new Stack<FileStatus>();
for (pathstack.push(srcfilestat); !pathstack.empty(); ) {
FileStatus cur = pathstack.pop();
FileStatus[] children = srcfs.listStatus(cur.getPath());
for (int i = 0; i < children.length; i++) {
boolean skipfile = false;
final FileStatus child = children[i];
final String dst = makeRelative(root, child.getPath());
++srcCount;
if (child.isDir()) {
pathstack.push(child);
} else {
// skip file if the src and the dst files are the same.
skipfile = update && sameFile(srcfs, child, dstfs, new Path(args.dst, dst));
// skip file if it exceed file limit or size limit
skipfile |=
fileCount == args.filelimit || byteCount + child.getLen() > args.sizelimit;
if (!skipfile) {
++fileCount;
byteCount += child.getLen();
if (LOG.isTraceEnabled()) {
LOG.trace("adding file " + child.getPath());
}
++cnsyncf;
cbsyncs += child.getLen();
if (cnsyncf > SYNC_FILE_MAX || cbsyncs > BYTES_PER_MAP) {
src_writer.sync();
dst_writer.sync();
cnsyncf = 0;
cbsyncs = 0L;
}
}
}
if (!skipfile) {
src_writer.append(
new LongWritable(child.isDir() ? 0 : child.getLen()), new FilePair(child, dst));
}
dst_writer.append(new Text(dst), new Text(child.getPath().toString()));
}
if (cur.isDir()) {
String dst = makeRelative(root, cur.getPath());
dir_writer.append(new Text(dst), new FilePair(cur, dst));
if (++dirsyn > SYNC_FILE_MAX) {
dirsyn = 0;
dir_writer.sync();
}
}
}
}
} finally {
checkAndClose(src_writer);
checkAndClose(dst_writer);
checkAndClose(dir_writer);
}
FileStatus dststatus = null;
try {
dststatus = dstfs.getFileStatus(args.dst);
} catch (FileNotFoundException fnfe) {
LOG.info(args.dst + " does not exist.");
}
// create dest path dir if copying > 1 file
if (dststatus == null) {
if (srcCount > 1 && !dstfs.mkdirs(args.dst)) {
throw new IOException("Failed to create" + args.dst);
}
}
final Path sorted = new Path(jobDirectory, "_distcp_sorted");
checkDuplication(jobfs, dstfilelist, sorted, conf);
if (dststatus != null && args.flags.contains(Options.DELETE)) {
deleteNonexisting(dstfs, dststatus, sorted, jobfs, jobDirectory, jobConf, conf);
}
Path tmpDir =
new Path(
(dstExists && !dstIsDir) || (!dstExists && srcCount == 1)
? args.dst.getParent()
: args.dst,
"_distcp_tmp_" + randomId);
jobConf.set(TMP_DIR_LABEL, tmpDir.toUri().toString());
LOG.info("srcCount=" + srcCount);
jobConf.setInt(SRC_COUNT_LABEL, srcCount);
jobConf.setLong(TOTAL_SIZE_LABEL, byteCount);
setMapCount(byteCount, jobConf);
}
/**
* Set if native hadoop libraries, if present, can be used for this job.
*
* @param jobConf job configuration
* @param loadNativeLibraries can native hadoop libraries be loaded
*/
public void setLoadNativeLibraries(JobConf jobConf, boolean loadNativeLibraries) {
jobConf.setBoolean("hadoop.native.lib", loadNativeLibraries);
}
boolean createJobConfAndParseArgs(String... args) throws ParseException, IOException {
job = new JobConf(getConf(), Crush.class);
/*
* Turn off speculative execution because that's just wasting network io.
*/
job.setMapSpeculativeExecution(false);
job.setReduceSpeculativeExecution(false);
/*
* Turn off pre-emption because we don't want to kill a task after two hours of network io.
*/
job.set("mapred.fairscheduler.preemption", "false");
tmpDir = new Path("tmp/crush-" + UUID.randomUUID());
outDir = new Path(tmpDir, "out");
double threshold = 0.75;
List<String> regexes = asList(".+");
List<String> replacements =
asList("crushed_file-${crush.timestamp}-${crush.task.num}-${crush.file.num}");
List<String> inFormats = asList(SequenceFileInputFormat.class.getName());
List<String> outFormats = asList(SequenceFileOutputFormat.class.getName());
String crushTimestamp;
Options options = buildOptions();
CommandLine cli = new GnuParser().parse(options, args);
if (cli.hasOption("?")) {
BufferedReader reader =
new BufferedReader(
new InputStreamReader(getClass().getClassLoader().getResourceAsStream("help.txt")));
try {
String line;
while (null != (line = reader.readLine())) {
System.out.println(line);
}
} finally {
reader.close();
}
return false;
}
if (cli.hasOption("verbose")) {
console = Verbosity.VERBOSE;
} else if (cli.hasOption("info")) {
console = Verbosity.INFO;
} else {
console = Verbosity.NONE;
}
if (cli.hasOption("ignore-regex")) {
ignoredFiles = Pattern.compile(cli.getOptionValue("ignore-regex")).matcher("");
}
excludeSingleFileDirs = !cli.hasOption("include-single-file-dirs");
String[] nonOptions = cli.getArgs();
if (2 == nonOptions.length) {
/*
* Stand alone mode accepts two arguments.
*/
mode = Mode.STAND_ALONE;
srcDir = new Path(nonOptions[0]);
dest = new Path(nonOptions[1]);
if (cli.hasOption("input-format")) {
inFormats = asList(cli.getOptionValue("input-format"));
}
if (cli.hasOption("output-format")) {
outFormats = asList(cli.getOptionValue("output-format"));
}
replacements = asList(dest.getName());
crushTimestamp = Long.toString(currentTimeMillis());
} else {
/*
* The previous version expected three or four arguments. The third one specified the number of tasks to use, which is an
* integral number, just like the third argument in the new version, which is a timestamp. We tell the two apart by looking
* at the value of the argument. A timestamp is going to be a huge, 14-digit number while the number of tasks should be much
* smaller.
*/
if ((args.length == 4 || args.length == 3)
&& args.length == nonOptions.length
&& args[2].length() != 14) {
int maxTasks = Integer.parseInt(args[2]);
if (maxTasks <= 0 || maxTasks > 4000) {
throw new IllegalArgumentException("Tasks must be in the range [1, 4000]: " + maxTasks);
}
job.setInt("mapred.reduce.tasks", maxTasks);
maxFileBlocks = Integer.MAX_VALUE;
crushTimestamp = Long.toString(currentTimeMillis());
srcDir = new Path(args[0]);
dest = new Path(args[1]);
mode = Mode.CLONE;
if (args.length == 4) {
if (args[3].equals("TEXT")) {
/*
* These are the defaults except with text input and output formats.
*/
inFormats = asList(TextInputFormat.class.getName());
outFormats = asList(TextOutputFormat.class.getName());
} else if (!args[3].equals("SEQUENCE")) {
throw new IllegalArgumentException("Type must be either TEXT or SEQUENCE: " + args[3]);
}
}
} else {
/*
* V2 style arguments.
*/
if (cli.hasOption("threshold")) {
threshold = Double.parseDouble(cli.getOptionValue("threshold"));
if (0 >= threshold
|| 1 < threshold
|| Double.isInfinite(threshold)
|| Double.isNaN(threshold)) {
throw new IllegalArgumentException(
"Block size threshold must be in (0, 1]: " + threshold);
}
}
if (cli.hasOption("max-file-blocks")) {
int maxFileBlocksOption = Integer.parseInt(cli.getOptionValue("max-file-blocks"));
if (0 > maxFileBlocksOption) {
throw new IllegalArgumentException(
"Maximum file size in blocks must be positive: " + maxFileBlocksOption);
}
maxFileBlocks = maxFileBlocksOption;
} else {
maxFileBlocks = 8;
}
if (cli.hasOption("regex")) {
regexes = asList(cli.getOptionValues("regex"));
}
if (cli.hasOption("replacement")) {
replacements = asList(cli.getOptionValues("replacement"));
}
if (cli.hasOption("input-format")) {
inFormats = asList(cli.getOptionValues("input-format"));
}
if (cli.hasOption("output-format")) {
outFormats = asList(cli.getOptionValues("output-format"));
}
if (3 != nonOptions.length) {
throw new IllegalArgumentException(
"Could not find source directory, out directory, and job timestamp");
}
srcDir = new Path(nonOptions[0]);
dest = new Path(nonOptions[1]);
crushTimestamp = nonOptions[2];
if (cli.hasOption("clone")) {
mode = Mode.CLONE;
} else {
mode = Mode.MAP_REDUCE;
}
if (!crushTimestamp.matches("\\d{14}")) {
throw new IllegalArgumentException(
"Crush timestamp must be 14 digits yyyymmddhhMMss: " + crushTimestamp);
}
}
dfsBlockSize = Long.parseLong(job.get("dfs.blocksize"));
maxEligibleSize = (long) (dfsBlockSize * threshold);
}
/*
* Add the crush specs and compression options to the configuration.
*/
job.set("crush.timestamp", crushTimestamp);
if (ignoredFiles != null) {
job.set("crush.ignore-regex", ignoredFiles.pattern().pattern());
}
if (regexes.size() != replacements.size()
|| replacements.size() != inFormats.size()
|| inFormats.size() != outFormats.size()) {
throw new IllegalArgumentException(
"Must be an equal number of regex, replacement, in-format, and out-format options");
}
job.setInt("crush.num.specs", regexes.size());
matchers = new ArrayList<Matcher>(regexes.size());
for (int i = 0; i < regexes.size(); i++) {
job.set(format("crush.%d.regex", i), regexes.get(i));
matchers.add(Pattern.compile(regexes.get(i)).matcher("dummy"));
job.set(format("crush.%d.regex.replacement", i), replacements.get(i));
String inFmt = inFormats.get(i);
if ("sequence".equals(inFmt)) {
inFmt = SequenceFileInputFormat.class.getName();
} else if ("text".equals(inFmt)) {
inFmt = TextInputFormat.class.getName();
} else {
try {
if (!FileInputFormat.class.isAssignableFrom(Class.forName(inFmt))) {
throw new IllegalArgumentException("Not a FileInputFormat:" + inFmt);
}
} catch (ClassNotFoundException e) {
throw new IllegalArgumentException("Not a FileInputFormat:" + inFmt);
}
}
job.set(format("crush.%d.input.format", i), inFmt);
String outFmt = outFormats.get(i);
if ("sequence".equals(outFmt)) {
outFmt = SequenceFileOutputFormat.class.getName();
} else if ("text".equals(outFmt)) {
outFmt = TextOutputFormat.class.getName();
} else {
try {
if (!FileOutputFormat.class.isAssignableFrom(Class.forName(outFmt))) {
throw new IllegalArgumentException("Not a FileOutputFormat:" + outFmt);
}
} catch (ClassNotFoundException e) {
throw new IllegalArgumentException("Not a FileOutputFormat:" + outFmt);
}
}
job.set(format("crush.%d.output.format", i), outFmt);
}
String codec = cli.getOptionValue("compress");
if (null == codec) {
codec = DefaultCodec.class.getName();
} else if ("none".equals(codec)) {
codec = null;
} else if ("gzip".equals(codec)) {
codec = GzipCodec.class.getName();
} else {
try {
if (!CompressionCodec.class.isAssignableFrom(Class.forName(codec))) {
throw new IllegalArgumentException("Not a CompressionCodec: " + codec);
}
} catch (ClassNotFoundException e) {
throw new IllegalArgumentException("Not a CompressionCodec: " + codec);
}
}
if (null == codec) {
job.setBoolean("mapred.output.compress", false);
} else {
job.setBoolean("mapred.output.compress", true);
job.set("mapred.output.compression.type", "BLOCK");
job.set("mapred.output.compression.codec", codec);
try {
CompressionCodec instance = (CompressionCodec) Class.forName(codec).newInstance();
codecExtension = instance.getDefaultExtension();
} catch (Exception e) {
throw new AssertionError();
}
}
return true;
}
@Test
public void testReduceProcessor() throws Exception {
final String dagName = "mrdag0";
String mapVertexName = MultiStageMRConfigUtil.getInitialMapVertexName();
String reduceVertexName = MultiStageMRConfigUtil.getFinalReduceVertexName();
JobConf jobConf = new JobConf(defaultConf);
setUpJobConf(jobConf);
MRHelpers.translateVertexConfToTez(jobConf);
jobConf.setInt(MRJobConfig.APPLICATION_ATTEMPT_ID, 0);
jobConf.set(
MRFrameworkConfigs.TASK_LOCAL_RESOURCE_DIR,
new Path(workDir, "localized-resources").toUri().toString());
jobConf.setBoolean(MRJobConfig.MR_TEZ_SPLITS_VIA_EVENTS, false);
Path mapInput = new Path(workDir, "map0");
MapUtils.generateInputSplit(localFs, workDir, jobConf, mapInput);
InputSpec mapInputSpec =
new InputSpec(
"NullSrcVertex",
new InputDescriptor(MRInputLegacy.class.getName())
.setUserPayload(MRHelpers.createMRInputPayload(jobConf, null)),
1);
OutputSpec mapOutputSpec =
new OutputSpec(
"NullDestVertex",
new OutputDescriptor(LocalOnFileSorterOutput.class.getName())
.setUserPayload(TezUtils.createUserPayloadFromConf(jobConf)),
1);
// Run a map
LogicalIOProcessorRuntimeTask mapTask =
MapUtils.createLogicalTask(
localFs,
workDir,
jobConf,
0,
mapInput,
new TestUmbilical(),
dagName,
mapVertexName,
Collections.singletonList(mapInputSpec),
Collections.singletonList(mapOutputSpec));
mapTask.initialize();
mapTask.run();
mapTask.close();
LOG.info("Starting reduce...");
Token<JobTokenIdentifier> shuffleToken = new Token<JobTokenIdentifier>();
jobConf.setOutputFormat(SequenceFileOutputFormat.class);
jobConf.set(
MRFrameworkConfigs.TASK_LOCAL_RESOURCE_DIR,
new Path(workDir, "localized-resources").toUri().toString());
FileOutputFormat.setOutputPath(jobConf, new Path(workDir, "output"));
ProcessorDescriptor reduceProcessorDesc =
new ProcessorDescriptor(ReduceProcessor.class.getName())
.setUserPayload(TezUtils.createUserPayloadFromConf(jobConf));
InputSpec reduceInputSpec =
new InputSpec(
mapVertexName,
new InputDescriptor(LocalMergedInput.class.getName())
.setUserPayload(TezUtils.createUserPayloadFromConf(jobConf)),
1);
OutputSpec reduceOutputSpec =
new OutputSpec(
"NullDestinationVertex",
new OutputDescriptor(MROutputLegacy.class.getName())
.setUserPayload(TezUtils.createUserPayloadFromConf(jobConf)),
1);
// Now run a reduce
TaskSpec taskSpec =
new TaskSpec(
TezTestUtils.getMockTaskAttemptId(0, 1, 0, 0),
dagName,
reduceVertexName,
reduceProcessorDesc,
Collections.singletonList(reduceInputSpec),
Collections.singletonList(reduceOutputSpec),
null);
Map<String, ByteBuffer> serviceConsumerMetadata = new HashMap<String, ByteBuffer>();
serviceConsumerMetadata.put(
ShuffleUtils.SHUFFLE_HANDLER_SERVICE_ID, ShuffleUtils.convertJobTokenToBytes(shuffleToken));
LogicalIOProcessorRuntimeTask task =
new LogicalIOProcessorRuntimeTask(
taskSpec,
0,
jobConf,
new String[] {workDir.toString()},
new TestUmbilical(),
serviceConsumerMetadata,
HashMultimap.<String, String>create());
task.initialize();
task.run();
task.close();
// MRTask mrTask = (MRTask)t.getProcessor();
// TODO NEWTEZ Verify the partitioner has not been created
// Likely not applicable anymore.
// Assert.assertNull(mrTask.getPartitioner());
// Only a task commit happens, hence the data is still in the temporary directory.
Path reduceOutputDir =
new Path(
new Path(workDir, "output"),
"_temporary/0/" + IDConverter.toMRTaskIdForOutput(TezTestUtils.getMockTaskId(0, 1, 0)));
Path reduceOutputFile = new Path(reduceOutputDir, "part-v001-o000-00000");
SequenceFile.Reader reader = new SequenceFile.Reader(localFs, reduceOutputFile, jobConf);
LongWritable key = new LongWritable();
Text value = new Text();
long prev = Long.MIN_VALUE;
while (reader.next(key, value)) {
if (prev != Long.MIN_VALUE) {
Assert.assertTrue(prev < key.get());
prev = key.get();
}
}
reader.close();
}
/*
* Helper function to create Vertex from MapWork.
*/
private Vertex createVertex(
JobConf conf,
MapWork mapWork,
LocalResource appJarLr,
List<LocalResource> additionalLr,
FileSystem fs,
Path mrScratchDir,
Context ctx,
TezWork tezWork)
throws Exception {
Path tezDir = getTezDir(mrScratchDir);
// set up the operator plan
Utilities.setMapWork(conf, mapWork, mrScratchDir, false);
// create the directories FileSinkOperators need
Utilities.createTmpDirs(conf, mapWork);
// Tez ask us to call this even if there's no preceding vertex
MultiStageMRConfToTezTranslator.translateVertexConfToTez(conf, null);
// finally create the vertex
Vertex map = null;
// use tez to combine splits
boolean useTezGroupedSplits = false;
int numTasks = -1;
Class amSplitGeneratorClass = null;
InputSplitInfo inputSplitInfo = null;
Class inputFormatClass = conf.getClass("mapred.input.format.class", InputFormat.class);
boolean vertexHasCustomInput = false;
if (tezWork != null) {
for (BaseWork baseWork : tezWork.getParents(mapWork)) {
if (tezWork.getEdgeType(baseWork, mapWork) == EdgeType.CUSTOM_EDGE) {
vertexHasCustomInput = true;
}
}
}
if (vertexHasCustomInput) {
useTezGroupedSplits = false;
// grouping happens in execution phase. Setting the class to TezGroupedSplitsInputFormat
// here would cause pre-mature grouping which would be incorrect.
inputFormatClass = HiveInputFormat.class;
conf.setClass("mapred.input.format.class", HiveInputFormat.class, InputFormat.class);
// mapreduce.tez.input.initializer.serialize.event.payload should be set to false when using
// this plug-in to avoid getting a serialized event at run-time.
conf.setBoolean("mapreduce.tez.input.initializer.serialize.event.payload", false);
} else {
// we'll set up tez to combine spits for us iff the input format
// is HiveInputFormat
if (inputFormatClass == HiveInputFormat.class) {
useTezGroupedSplits = true;
conf.setClass(
"mapred.input.format.class", TezGroupedSplitsInputFormat.class, InputFormat.class);
}
}
if (HiveConf.getBoolVar(conf, ConfVars.HIVE_AM_SPLIT_GENERATION)) {
// if we're generating the splits in the AM, we just need to set
// the correct plugin.
amSplitGeneratorClass = MRInputAMSplitGenerator.class;
} else {
// client side split generation means we have to compute them now
inputSplitInfo =
MRHelpers.generateInputSplits(
conf, new Path(tezDir, "split_" + mapWork.getName().replaceAll(" ", "_")));
numTasks = inputSplitInfo.getNumTasks();
}
byte[] serializedConf = MRHelpers.createUserPayloadFromConf(conf);
map =
new Vertex(
mapWork.getName(),
new ProcessorDescriptor(MapTezProcessor.class.getName()).setUserPayload(serializedConf),
numTasks,
getContainerResource(conf));
Map<String, String> environment = new HashMap<String, String>();
MRHelpers.updateEnvironmentForMRTasks(conf, environment, true);
map.setTaskEnvironment(environment);
map.setJavaOpts(getContainerJavaOpts(conf));
assert mapWork.getAliasToWork().keySet().size() == 1;
String alias = mapWork.getAliasToWork().keySet().iterator().next();
byte[] mrInput = null;
if (useTezGroupedSplits) {
mrInput =
MRHelpers.createMRInputPayloadWithGrouping(
serializedConf, HiveInputFormat.class.getName());
} else {
mrInput = MRHelpers.createMRInputPayload(serializedConf, null);
}
map.addInput(
alias,
new InputDescriptor(MRInputLegacy.class.getName()).setUserPayload(mrInput),
amSplitGeneratorClass);
Map<String, LocalResource> localResources = new HashMap<String, LocalResource>();
localResources.put(getBaseName(appJarLr), appJarLr);
for (LocalResource lr : additionalLr) {
localResources.put(getBaseName(lr), lr);
}
if (inputSplitInfo != null) {
// only relevant for client-side split generation
map.setTaskLocationsHint(inputSplitInfo.getTaskLocationHints());
MRHelpers.updateLocalResourcesForInputSplits(
FileSystem.get(conf), inputSplitInfo, localResources);
}
map.setTaskLocalResources(localResources);
return map;
}
public static void main(String[] args) throws Exception {
if (!validArgs(args)) {
printUsage();
return;
}
// These are the temp paths that are created on HDFS
String dir1 = "/user/miyuru/csrconverter-output";
String dir2 = "/user/miyuru/csrconverter-output-sorted";
// We first delete the temporary directories if they exist on the HDFS
FileSystem fs1 = FileSystem.get(new JobConf());
System.out.println("Deleting the dir : " + dir1);
if (fs1.exists(new Path(dir1))) {
fs1.delete(new Path(dir1), true);
}
System.out.println("Done deleting the dir : " + dir1);
System.out.println("Deleting the dir : " + dir2);
if (fs1.exists(new Path(dir2))) {
fs1.delete(new Path(dir2), true);
}
Path notinPath = new Path("/user/miyuru/notinverts/notinverts");
if (!fs1.exists(notinPath)) {
fs1.create(notinPath);
}
System.out.println("Done deleting the dir : " + dir2);
// Note on Aug 23 2014: Sometimes after this the mapReduce job hangs. need to see why.
VertexCounterClient.setDefaultGraphID(args[3], args[2]);
// First job creates the inverted index
JobConf conf = new JobConf(CSRConverter.class);
conf.set("org.acacia.partitioner.hbase.zookeeper.quorum", args[1]);
conf.set("org.acacia.partitioner.hbase.table", args[2]);
conf.set("org.acacia.partitioner.hbase.contacthost", args[3]);
conf.setOutputKeyClass(LongWritable.class);
conf.setOutputValueClass(Text.class);
// conf.setMapperClass(InvertedMapper.class);
conf.setReducerClass(InvertedReducer.class);
// conf.setInputFormat(TextInputFormat.class);
conf.setInputFormat(NLinesInputFormat.class);
conf.setOutputFormat(TextOutputFormat.class);
// FileInputFormat.setInputPaths(conf, new Path(args[0]));
MultipleInputs.addInputPath(
conf, new Path(args[0]), NLinesInputFormat.class, InvertedMapper.class);
MultipleInputs.addInputPath(
conf,
new Path("/user/miyuru/notinverts/notinverts"),
TextInputFormat.class,
InvertedMapper.class);
FileOutputFormat.setOutputPath(conf, new Path(dir1));
// Also for the moment we turn-off the speculative execution
conf.setBoolean("mapred.map.tasks.speculative.execution", false);
conf.setBoolean("mapred.reduce.tasks.speculative.execution", false);
conf.setNumMapTasks(96);
conf.setNumReduceTasks(96);
conf.setPartitionerClass(VertexPartitioner.class);
conf.set("vertex-count", args[4]);
conf.set("zero-flag", args[5]);
Job job = new Job(conf, "csr_inverter");
job.setSortComparatorClass(SortComparator.class);
job.waitForCompletion(true);
}
@Override
public void localizeConfiguration(JobConf jobConf) throws IOException, InterruptedException {
super.localizeConfiguration(jobConf);
jobConf.setBoolean(JobContext.TASK_ISMAP, true);
}
public void merge(Path out, Path[] segs, boolean filter, long slice) throws Exception {
String segmentName = Generator.generateSegmentName();
if (LOG.isInfoEnabled()) {
LOG.info("Merging " + segs.length + " segments to " + out + "/" + segmentName);
}
JobConf job = new NutchJob(getConf());
job.setJobName("mergesegs " + out + "/" + segmentName);
job.setBoolean("segment.merger.filter", filter);
job.setLong("segment.merger.slice", slice);
job.set("segment.merger.segmentName", segmentName);
FileSystem fs = FileSystem.get(getConf());
// prepare the minimal common set of input dirs
boolean g = true;
boolean f = true;
boolean p = true;
boolean c = true;
boolean pd = true;
boolean pt = true;
for (int i = 0; i < segs.length; i++) {
if (!fs.exists(segs[i])) {
if (LOG.isWarnEnabled()) {
LOG.warn("Input dir " + segs[i] + " doesn't exist, skipping.");
}
segs[i] = null;
continue;
}
if (LOG.isInfoEnabled()) {
LOG.info("SegmentMerger: adding " + segs[i]);
}
Path cDir = new Path(segs[i], Content.DIR_NAME);
Path gDir = new Path(segs[i], CrawlDatum.GENERATE_DIR_NAME);
Path fDir = new Path(segs[i], CrawlDatum.FETCH_DIR_NAME);
Path pDir = new Path(segs[i], CrawlDatum.PARSE_DIR_NAME);
Path pdDir = new Path(segs[i], ParseData.DIR_NAME);
Path ptDir = new Path(segs[i], ParseText.DIR_NAME);
c = c && fs.exists(cDir);
g = g && fs.exists(gDir);
f = f && fs.exists(fDir);
p = p && fs.exists(pDir);
pd = pd && fs.exists(pdDir);
pt = pt && fs.exists(ptDir);
}
StringBuffer sb = new StringBuffer();
if (c) sb.append(" " + Content.DIR_NAME);
if (g) sb.append(" " + CrawlDatum.GENERATE_DIR_NAME);
if (f) sb.append(" " + CrawlDatum.FETCH_DIR_NAME);
if (p) sb.append(" " + CrawlDatum.PARSE_DIR_NAME);
if (pd) sb.append(" " + ParseData.DIR_NAME);
if (pt) sb.append(" " + ParseText.DIR_NAME);
if (LOG.isInfoEnabled()) {
LOG.info("SegmentMerger: using segment data from:" + sb.toString());
}
for (int i = 0; i < segs.length; i++) {
if (segs[i] == null) continue;
if (g) {
Path gDir = new Path(segs[i], CrawlDatum.GENERATE_DIR_NAME);
job.addInputPath(gDir);
}
if (c) {
Path cDir = new Path(segs[i], Content.DIR_NAME);
job.addInputPath(cDir);
}
if (f) {
Path fDir = new Path(segs[i], CrawlDatum.FETCH_DIR_NAME);
job.addInputPath(fDir);
}
if (p) {
Path pDir = new Path(segs[i], CrawlDatum.PARSE_DIR_NAME);
job.addInputPath(pDir);
}
if (pd) {
Path pdDir = new Path(segs[i], ParseData.DIR_NAME);
job.addInputPath(pdDir);
}
if (pt) {
Path ptDir = new Path(segs[i], ParseText.DIR_NAME);
job.addInputPath(ptDir);
}
}
job.setInputFormat(ObjectInputFormat.class);
job.setMapperClass(SegmentMerger.class);
job.setReducerClass(SegmentMerger.class);
job.setOutputPath(out);
job.setOutputKeyClass(Text.class);
job.setOutputValueClass(MetaWrapper.class);
job.setOutputFormat(SegmentOutputFormat.class);
setConf(job);
JobClient.runJob(job);
}
/** Runs this tool. */
@SuppressWarnings("deprecation")
public int run(String[] args) throws Exception {
JobConf job = new JobConf(getConf(), Docnos2Titles.class);
// Read commandline arguments
CommandLine cmdline = parseArgs(args);
if (cmdline == null) {
printUsage();
}
String eCollectionPath = cmdline.getOptionValue(ECOLLECTION_OPTION);
String fCollectionPath = cmdline.getOptionValue(FCOLLECTION_OPTION);
String pwsimOutputPath = cmdline.getOptionValue(PWSIM_OPTION);
String titlePairsPath = cmdline.getOptionValue(OUTPUT_PATH_OPTION);
String eLang = cmdline.getOptionValue(ELANG_OPTION);
String fLang = cmdline.getOptionValue(FLANG_OPTION);
String samplesFile = cmdline.getOptionValue(SAMPLEDOCNOS_OPTION);
job.setJobName("Docnos2Titles_" + fLang + "-" + eLang);
FileInputFormat.addInputPaths(job, eCollectionPath);
FileInputFormat.addInputPaths(job, fCollectionPath);
FileOutputFormat.setOutputPath(job, new Path(titlePairsPath));
DistributedCache.addCacheFile(new URI(pwsimOutputPath), job);
DistributedCache.addCacheFile(new URI(samplesFile), job);
job.set("eLang", eLang);
job.set("fLang", fLang);
job.set("PwsimPairs", pwsimOutputPath);
job.set("Ivory.SampleFile", samplesFile);
job.setInt("mapred.task.timeout", 60000000);
job.set("mapreduce.map.memory.mb", "3000");
job.set("mapreduce.map.java.opts", "-Xmx3000m");
job.setBoolean("mapred.map.tasks.speculative.execution", false);
job.setBoolean("mapred.reduce.tasks.speculative.execution", false);
job.setNumMapTasks(100);
job.setNumReduceTasks(1);
job.setInt("mapred.min.split.size", 2000000000);
job.setFloat("mapred.reduce.slowstart.completed.maps", 0.9f);
job.setInputFormat(SequenceFileInputFormat.class);
job.setOutputFormat(TextOutputFormat.class);
job.setMapOutputKeyClass(PairOfInts.class);
job.setMapOutputValueClass(PairOfIntString.class);
job.setOutputKeyClass(Text.class);
job.setOutputValueClass(Text.class);
job.setMapperClass(MyMapper.class);
job.setReducerClass(MyReducer.class);
sLogger.info("Running job " + job.getJobName() + "...");
sLogger.info("E-collection path: " + eCollectionPath);
sLogger.info("F-collection path: " + fCollectionPath);
sLogger.info("Pwsim output path: " + pwsimOutputPath);
sLogger.info("Output path: " + titlePairsPath);
sLogger.info("Sample file?: " + ((samplesFile != null) ? samplesFile : "none"));
long startTime = System.currentTimeMillis();
JobClient.runJob(job);
System.out.println(
"Job finished in " + (System.currentTimeMillis() - startTime) / 1000.0 + " seconds");
return 0;
}
/**
* Run a compactor job.
*
* @param conf Hive configuration file
* @param jobName name to run this job with
* @param t metastore table
* @param sd metastore storage descriptor
* @param txns list of valid transactions
* @param isMajor is this a major compaction?
* @throws java.io.IOException if the job fails
*/
void run(
HiveConf conf,
String jobName,
Table t,
StorageDescriptor sd,
ValidTxnList txns,
boolean isMajor,
Worker.StatsUpdater su)
throws IOException {
JobConf job = new JobConf(conf);
job.setJobName(jobName);
job.setOutputKeyClass(NullWritable.class);
job.setOutputValueClass(NullWritable.class);
job.setJarByClass(CompactorMR.class);
LOG.debug("User jar set to " + job.getJar());
job.setMapperClass(CompactorMap.class);
job.setNumReduceTasks(0);
job.setInputFormat(CompactorInputFormat.class);
job.setOutputFormat(NullOutputFormat.class);
job.setOutputCommitter(CompactorOutputCommitter.class);
String queueName = conf.getVar(HiveConf.ConfVars.COMPACTOR_JOB_QUEUE);
if (queueName != null && queueName.length() > 0) {
job.setQueueName(queueName);
}
job.set(FINAL_LOCATION, sd.getLocation());
job.set(TMP_LOCATION, sd.getLocation() + "/" + TMPDIR + "_" + UUID.randomUUID().toString());
job.set(INPUT_FORMAT_CLASS_NAME, sd.getInputFormat());
job.set(OUTPUT_FORMAT_CLASS_NAME, sd.getOutputFormat());
job.setBoolean(IS_MAJOR, isMajor);
job.setBoolean(IS_COMPRESSED, sd.isCompressed());
job.set(TABLE_PROPS, new StringableMap(t.getParameters()).toString());
job.setInt(NUM_BUCKETS, sd.getNumBuckets());
job.set(ValidTxnList.VALID_TXNS_KEY, txns.toString());
setColumnTypes(job, sd.getCols());
// Figure out and encode what files we need to read. We do this here (rather than in
// getSplits below) because as part of this we discover our minimum and maximum transactions,
// and discovering that in getSplits is too late as we then have no way to pass it to our
// mapper.
AcidUtils.Directory dir = AcidUtils.getAcidState(new Path(sd.getLocation()), conf, txns, false);
StringableList dirsToSearch = new StringableList();
Path baseDir = null;
if (isMajor) {
// There may not be a base dir if the partition was empty before inserts or if this
// partition is just now being converted to ACID.
baseDir = dir.getBaseDirectory();
if (baseDir == null) {
List<HdfsFileStatusWithId> originalFiles = dir.getOriginalFiles();
if (!(originalFiles == null) && !(originalFiles.size() == 0)) {
// There are original format files
for (HdfsFileStatusWithId stat : originalFiles) {
Path path = stat.getFileStatus().getPath();
dirsToSearch.add(path);
LOG.debug("Adding original file " + path + " to dirs to search");
}
// Set base to the location so that the input format reads the original files.
baseDir = new Path(sd.getLocation());
}
} else {
// add our base to the list of directories to search for files in.
LOG.debug("Adding base directory " + baseDir + " to dirs to search");
dirsToSearch.add(baseDir);
}
}
List<AcidUtils.ParsedDelta> parsedDeltas = dir.getCurrentDirectories();
if (parsedDeltas == null || parsedDeltas.size() == 0) {
// Seriously, no deltas? Can't compact that.
LOG.error("No delta files found to compact in " + sd.getLocation());
return;
}
StringableList deltaDirs = new StringableList();
long minTxn = Long.MAX_VALUE;
long maxTxn = Long.MIN_VALUE;
for (AcidUtils.ParsedDelta delta : parsedDeltas) {
LOG.debug("Adding delta " + delta.getPath() + " to directories to search");
dirsToSearch.add(delta.getPath());
deltaDirs.add(delta.getPath());
minTxn = Math.min(minTxn, delta.getMinTransaction());
maxTxn = Math.max(maxTxn, delta.getMaxTransaction());
}
if (baseDir != null) job.set(BASE_DIR, baseDir.toString());
job.set(DELTA_DIRS, deltaDirs.toString());
job.set(DIRS_TO_SEARCH, dirsToSearch.toString());
job.setLong(MIN_TXN, minTxn);
job.setLong(MAX_TXN, maxTxn);
LOG.debug("Setting minimum transaction to " + minTxn);
LOG.debug("Setting maximume transaction to " + maxTxn);
RunningJob rj = JobClient.runJob(job);
LOG.info(
"Submitted "
+ (isMajor ? CompactionType.MAJOR : CompactionType.MINOR)
+ " compaction job '"
+ jobName
+ "' with jobID="
+ rj.getID()
+ " to "
+ job.getQueueName()
+ " queue. "
+ "(current delta dirs count="
+ dir.getCurrentDirectories().size()
+ ", obsolete delta dirs count="
+ dir.getObsolete());
rj.waitForCompletion();
su.gatherStats();
}
public void indexSolr(
String solrUrl,
Path crawlDb,
Path linkDb,
List<Path> segments,
boolean noCommit,
boolean deleteGone,
String solrParams,
boolean filter,
boolean normalize)
throws IOException {
filter = false;
SimpleDateFormat sdf = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss");
long start = System.currentTimeMillis();
LOG.info("SolrIndexer: starting at " + sdf.format(start));
final JobConf job = new NutchJob(getConf());
job.setJobName("index-solr " + solrUrl);
LOG.info("SolrIndexer: deleting gone documents: " + deleteGone);
LOG.info("SolrIndexer: URL filtering: " + filter);
LOG.info("SolrIndexer: URL normalizing: " + normalize);
IndexerMapReduce.initMRJob(crawlDb, linkDb, segments, job);
job.set(SolrConstants.SERVER_URL, solrUrl);
job.setBoolean(IndexerMapReduce.INDEXER_DELETE, deleteGone);
job.setBoolean(IndexerMapReduce.URL_FILTERING, filter);
job.setBoolean(IndexerMapReduce.URL_NORMALIZING, normalize);
if (solrParams != null) {
job.set(SolrConstants.PARAMS, solrParams);
}
NutchIndexWriterFactory.addClassToConf(job, SolrWriter.class);
job.setReduceSpeculativeExecution(false);
final Path tmp = new Path("tmp_" + System.currentTimeMillis() + "-" + new Random().nextInt());
FileOutputFormat.setOutputPath(job, tmp);
try {
JobClient.runJob(job);
// do the commits once and for all the reducers in one go
SolrServer solr = SolrUtils.getCommonsHttpSolrServer(job);
if (!noCommit) {
solr.commit();
}
long end = System.currentTimeMillis();
LOG.info(
"SolrIndexer: finished at "
+ sdf.format(end)
+ ", elapsed: "
+ TimingUtil.elapsedTime(start, end));
} catch (Exception e) {
LOG.error(e.toString());
} finally {
FileSystem.get(job).delete(tmp, true);
}
}
