/**
* Generate the requested number of file splits, with the filename set to the filename of the
* output file.
*/
public InputSplit[] getSplits(JobConf job, int numSplits) throws IOException {
/** 设置输入分片的个数* */
JobClient client = new JobClient(job);
ClusterStatus cluster = client.getClusterStatus();
/** 如果属性不存在 则返回默认的值 * */
int numMapsPerHost = job.getInt("test.randomwriter.maps_per_host", 10);
long numBytesToWritePerMap =
job.getLong("test.randomwrite.bytes_per_map", 1 * 1024 * 1024 * 1024);
if (numBytesToWritePerMap == 0) {
System.err.println("Cannot have test.randomwrite.bytes_per_map set to 0");
}
long totalBytesToWrite =
job.getLong(
"test.randomwrite.total_bytes",
numMapsPerHost * numBytesToWritePerMap * cluster.getTaskTrackers());
int numMaps = (int) (totalBytesToWrite / numBytesToWritePerMap);
if (numMaps == 0 && totalBytesToWrite > 0) {
numMaps = 1;
}
System.out.println("numMaps-------" + numMaps);
InputSplit[] result = new InputSplit[numMaps];
Path outDir = FileOutputFormat.getOutputPath(job);
for (int i = 0; i < result.length; ++i) {
result[i] = new FileSplit(new Path(outDir, "dummy-split-" + i), 0, 1, (String[]) null);
}
return result;
}
/**
* This method call when injected into a class will modify the output path, only if output is into
* HDFS
*
* @param job Job whose output path need to be changed
*/
public static void modifyOutputPath(JobConf job) {
Path path = org.apache.hadoop.mapred.FileOutputFormat.getOutputPath(job);
if (path == null) {
throw new IllegalArgumentException("Job Output path is null, expecting not null path value");
}
StringBuilder out = new StringBuilder(path.toString());
out.append(SEPARATOR_UNDERSCORE).append(System.currentTimeMillis());
org.apache.hadoop.mapred.FileOutputFormat.setOutputPath(job, new Path(out.toString()));
}
public RecordWriter<WritableComparable, Writable> getRecordWriter(
final FileSystem fs, JobConf job, String name, final Progressable progress)
throws IOException {
final Path segmentDumpFile = new Path(FileOutputFormat.getOutputPath(job), name);
// Get the old copy out of the way
if (fs.exists(segmentDumpFile)) fs.delete(segmentDumpFile, true);
final PrintStream printStream = new PrintStream(fs.create(segmentDumpFile));
return new RecordWriter<WritableComparable, Writable>() {
public synchronized void write(WritableComparable key, Writable value) throws IOException {
printStream.println(value);
}
public synchronized void close(Reporter reporter) throws IOException {
printStream.close();
}
};
}
public void configure(JobConf job) {
_jobConf = job;
crawlerCount =
job.getInt(CrawlEnvironment.PROPERTY_NUM_CRAWLERS, CrawlEnvironment.CRAWLERS.length);
partitionNumber = job.getInt("mapred.task.partition", -1);
try {
FileSystem fs = FileSystem.get(job);
Path workPath = FileOutputFormat.getOutputPath(job);
debugURLStream =
fs.create(new Path(workPath, "debugURLS-" + NUMBER_FORMAT.format(partitionNumber)));
urlDebugURLWriter = new OutputStreamWriter(debugURLStream, Charset.forName("UTF-8"));
_emittedURLSFilter = new URLFPBloomFilter(NUM_ELEMENTS, NUM_HASH_FUNCTIONS, NUM_BITS);
} catch (IOException e) {
LOG.error(StringUtils.stringifyException(e));
throw new RuntimeException(e);
}
}
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.");
}
static void checkRecords(Configuration defaults, Path sortInput, Path sortOutput)
throws IOException {
FileSystem inputfs = sortInput.getFileSystem(defaults);
FileSystem outputfs = sortOutput.getFileSystem(defaults);
FileSystem defaultfs = FileSystem.get(defaults);
JobConf jobConf = new JobConf(defaults, RecordStatsChecker.class);
jobConf.setJobName("sortvalidate-recordstats-checker");
int noSortReduceTasks = outputfs.listStatus(sortOutput, sortPathsFilter).length;
jobConf.setInt("sortvalidate.sort.reduce.tasks", noSortReduceTasks);
int noSortInputpaths = inputfs.listStatus(sortInput).length;
jobConf.setInputFormat(NonSplitableSequenceFileInputFormat.class);
jobConf.setOutputFormat(SequenceFileOutputFormat.class);
jobConf.setOutputKeyClass(IntWritable.class);
jobConf.setOutputValueClass(RecordStatsChecker.RecordStatsWritable.class);
jobConf.setMapperClass(Map.class);
jobConf.setCombinerClass(Reduce.class);
jobConf.setReducerClass(Reduce.class);
jobConf.setNumMapTasks(noSortReduceTasks);
jobConf.setNumReduceTasks(1);
FileInputFormat.setInputPaths(jobConf, sortInput);
FileInputFormat.addInputPath(jobConf, sortOutput);
Path outputPath = new Path("/tmp/sortvalidate/recordstatschecker");
if (defaultfs.exists(outputPath)) {
defaultfs.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.RecordStatsChecker: Validate sort "
+ "from "
+ inputPaths[0]
+ " ("
+ noSortInputpaths
+ " files), "
+ inputPaths[1]
+ " ("
+ noSortReduceTasks
+ " files) into "
+ FileOutputFormat.getOutputPath(jobConf)
+ " with 1 reducer.");
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.");
// Check to ensure that the statistics of the
// framework's sort-input and sort-output match
SequenceFile.Reader stats =
new SequenceFile.Reader(defaultfs, new Path(outputPath, "part-00000"), defaults);
IntWritable k1 = new IntWritable();
IntWritable k2 = new IntWritable();
RecordStatsWritable v1 = new RecordStatsWritable();
RecordStatsWritable v2 = new RecordStatsWritable();
if (!stats.next(k1, v1)) {
throw new IOException("Failed to read record #1 from reduce's output");
}
if (!stats.next(k2, v2)) {
throw new IOException("Failed to read record #2 from reduce's output");
}
if ((v1.getBytes() != v2.getBytes())
|| (v1.getRecords() != v2.getRecords())
|| v1.getChecksum() != v2.getChecksum()) {
throw new IOException(
"("
+ v1.getBytes()
+ ", "
+ v1.getRecords()
+ ", "
+ v1.getChecksum()
+ ") v/s ("
+ v2.getBytes()
+ ", "
+ v2.getRecords()
+ ", "
+ v2.getChecksum()
+ ")");
}
}