private static DBConfiguration setOutput(JobConf job, String tableName) {
job.setOutputFormat(DBOutputFormat.class);
job.setReduceSpeculativeExecution(false);
DBConfiguration dbConf = new DBConfiguration(job);
dbConf.setOutputTableName(tableName);
return dbConf;
}
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;
}
public JobBuilder speculativeReducerExecution() throws IOException {
_jobConf.setReduceSpeculativeExecution(true);
return this;
}
public JobBuilder speculativeReducerExecution(boolean enable) throws IOException {
_jobConf.setReduceSpeculativeExecution(enable);
return this;
}
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);
}
}
