/**
* 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;
}
private String[] getActiveServersList(JobContext context) {
String[] servers = null;
try {
JobClient jc = new JobClient((JobConf) context.getConfiguration());
ClusterStatus status = jc.getClusterStatus(true);
Collection<String> atc = status.getActiveTrackerNames();
servers = new String[atc.size()];
int s = 0;
for (String serverInfo : atc) {
// System.out.println("serverInfo:" + serverInfo);
StringTokenizer st = new StringTokenizer(serverInfo, ":");
String trackerName = st.nextToken();
// System.out.println("trackerName:" + trackerName);
StringTokenizer st1 = new StringTokenizer(trackerName, "_");
st1.nextToken();
servers[s++] = st1.nextToken();
}
} catch (IOException e) {
e.printStackTrace();
}
return servers;
}
/**
* This is the main routine for launching a distributed random write job. It runs 10 maps/node and
* each node writes 1 gig of data to a DFS file. The reduce doesn't do anything.
*
* @throws IOException
*/
public int run(String[] args) throws Exception {
if (args.length == 0) {
System.out.println("Usage: writer <out-dir>");
ToolRunner.printGenericCommandUsage(System.out);
return -1;
}
Path outDir = new Path(args[0]);
JobConf job = new JobConf(getConf());
job.setJarByClass(RandomWriter.class);
job.setJobName("random-writer");
FileOutputFormat.setOutputPath(job, outDir);
job.setOutputKeyClass(BytesWritable.class);
job.setOutputValueClass(BytesWritable.class);
job.setInputFormat(RandomInputFormat.class);
job.setMapperClass(Map.class);
job.setReducerClass(IdentityReducer.class);
job.setOutputFormat(SequenceFileOutputFormat.class);
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");
return -2;
}
long totalBytesToWrite =
job.getLong(
"test.randomwrite.total_bytes",
numMapsPerHost * numBytesToWritePerMap * cluster.getTaskTrackers());
int numMaps = (int) (totalBytesToWrite / numBytesToWritePerMap);
if (numMaps == 0 && totalBytesToWrite > 0) {
numMaps = 1;
job.setLong("test.randomwrite.bytes_per_map", totalBytesToWrite);
}
job.setNumMapTasks(numMaps);
/** 建议型的 * */
System.out.println("Running " + numMaps + " maps.");
// reducer NONE
job.setNumReduceTasks(0);
Date startTime = new Date();
System.out.println("Job started: " + startTime);
JobClient.runJob(job);
Date endTime = new Date();
System.out.println("Job ended: " + endTime);
System.out.println(
"The job took " + (endTime.getTime() - startTime.getTime()) / 1000 + " seconds.");
return 0;
}
/**
* The main driver for sort program. Invoke this method to submit the map/reduce job.
*
* @throws Exception When there is communication problems with the job tracker.
*/
public int run(String[] args) throws Exception {
Configuration conf = getConf();
JobClient client = new JobClient(conf);
ClusterStatus cluster = client.getClusterStatus();
int num_reduces = (int) (cluster.getMaxReduceTasks() * 0.9);
String join_reduces = conf.get(REDUCES_PER_HOST);
if (join_reduces != null) {
num_reduces = cluster.getTaskTrackers() * Integer.parseInt(join_reduces);
}
Job job = new Job(conf);
job.setJobName("join");
job.setJarByClass(Sort.class);
job.setMapperClass(Mapper.class);
job.setReducerClass(Reducer.class);
Class<? extends InputFormat> inputFormatClass = SequenceFileInputFormat.class;
Class<? extends OutputFormat> outputFormatClass = SequenceFileOutputFormat.class;
Class<? extends WritableComparable> outputKeyClass = BytesWritable.class;
Class<? extends Writable> outputValueClass = TupleWritable.class;
String op = "inner";
List<String> otherArgs = new ArrayList<String>();
for (int i = 0; i < args.length; ++i) {
try {
if ("-r".equals(args[i])) {
num_reduces = Integer.parseInt(args[++i]);
} else if ("-inFormat".equals(args[i])) {
inputFormatClass = Class.forName(args[++i]).asSubclass(InputFormat.class);
} else if ("-outFormat".equals(args[i])) {
outputFormatClass = Class.forName(args[++i]).asSubclass(OutputFormat.class);
} else if ("-outKey".equals(args[i])) {
outputKeyClass = Class.forName(args[++i]).asSubclass(WritableComparable.class);
} else if ("-outValue".equals(args[i])) {
outputValueClass = Class.forName(args[++i]).asSubclass(Writable.class);
} else if ("-joinOp".equals(args[i])) {
op = args[++i];
} else {
otherArgs.add(args[i]);
}
} catch (NumberFormatException except) {
System.out.println("ERROR: Integer expected instead of " + args[i]);
return printUsage();
} catch (ArrayIndexOutOfBoundsException except) {
System.out.println("ERROR: Required parameter missing from " + args[i - 1]);
return printUsage(); // exits
}
}
// Set user-supplied (possibly default) job configs
job.setNumReduceTasks(num_reduces);
if (otherArgs.size() < 2) {
System.out.println("ERROR: Wrong number of parameters: ");
return printUsage();
}
FileOutputFormat.setOutputPath(job, new Path(otherArgs.remove(otherArgs.size() - 1)));
List<Path> plist = new ArrayList<Path>(otherArgs.size());
for (String s : otherArgs) {
plist.add(new Path(s));
}
job.setInputFormatClass(CompositeInputFormat.class);
job.getConfiguration()
.set(
CompositeInputFormat.JOIN_EXPR,
CompositeInputFormat.compose(op, inputFormatClass, plist.toArray(new Path[0])));
job.setOutputFormatClass(outputFormatClass);
job.setOutputKeyClass(outputKeyClass);
job.setOutputValueClass(outputValueClass);
Date startTime = new Date();
System.out.println("Job started: " + startTime);
int ret = job.waitForCompletion(true) ? 0 : 1;
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.");
return ret;
}
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.");
}