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;
}
/**
* 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 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;
}
@Override
public JobTrackerState getJobTrackerState(ClusterStatus clusterStatus) throws Exception {
switch (clusterStatus.getJobTrackerStatus()) {
case INITIALIZING:
return JobTrackerState.INITIALIZING;
case RUNNING:
return JobTrackerState.RUNNING;
default:
String errorMsg = "Unrecognized JobTracker state: " + clusterStatus.getJobTrackerStatus();
throw new Exception(errorMsg);
}
}
/** Return the status information about the Map-Reduce cluster */
public HiveClusterStatus getClusterStatus() throws HiveServerException, TException {
HiveClusterStatus hcs;
try {
ClusterStatus cs = driver.getClusterStatus();
JobTracker.State jbs = cs.getJobTrackerState();
// Convert the ClusterStatus to its Thrift equivalent: HiveClusterStatus
int state;
switch (jbs) {
case INITIALIZING:
state = JobTrackerState.INITIALIZING;
break;
case RUNNING:
state = JobTrackerState.RUNNING;
break;
default:
String errorMsg = "Unrecognized JobTracker state: " + jbs.toString();
throw new Exception(errorMsg);
}
hcs =
new HiveClusterStatus(
cs.getTaskTrackers(),
cs.getMapTasks(),
cs.getReduceTasks(),
cs.getMaxMapTasks(),
cs.getMaxReduceTasks(),
state);
} catch (Exception e) {
LOG.error(e.toString());
e.printStackTrace();
throw new HiveServerException("Unable to get cluster status: " + e.toString());
}
return hcs;
}
void startJobTracker(boolean wait) {
// Create the JobTracker
jobTracker = new JobTrackerRunner(conf);
jobTrackerThread = new Thread(jobTracker);
jobTrackerThread.start();
if (!wait) {
return;
}
while (jobTracker.isActive() && !jobTracker.isUp()) {
try { // let daemons get started
Thread.sleep(1000);
} catch (InterruptedException e) {
}
}
// is the jobtracker has started then wait for it to init
ClusterStatus status = null;
if (jobTracker.isUp()) {
status = jobTracker.getJobTracker().getClusterStatus(false);
while (jobTracker.isActive()
&& status.getJobTrackerState() == JobTracker.State.INITIALIZING) {
try {
LOG.info("JobTracker still initializing. Waiting.");
Thread.sleep(1000);
} catch (InterruptedException e) {
}
status = jobTracker.getJobTracker().getClusterStatus(false);
}
}
if (!jobTracker.isActive()) {
// return if jobtracker has crashed
return;
}
// Set the configuration for the task-trackers
this.jobTrackerPort = jobTracker.getJobTrackerPort();
this.jobTrackerInfoPort = jobTracker.getJobTrackerInfoPort();
}
@Override
public Task assignTask(TaskTrackerStatus tts, long currentTime, Collection<JobInProgress> visited)
throws IOException {
if (isRunnable()) {
visited.add(job);
TaskTrackerManager ttm = scheduler.taskTrackerManager;
ClusterStatus clusterStatus = ttm.getClusterStatus();
int numTaskTrackers = clusterStatus.getTaskTrackers();
if (taskType == TaskType.MAP) {
LocalityLevel localityLevel = scheduler.getAllowedLocalityLevel(job, currentTime);
scheduler.getEventLog().log("ALLOWED_LOC_LEVEL", job.getJobID(), localityLevel);
// obtainNewMapTask needs to be passed 1 + the desired locality level
return job.obtainNewMapTask(
tts, numTaskTrackers, ttm.getNumberOfUniqueHosts(), localityLevel.toCacheLevelCap());
} else {
return job.obtainNewReduceTask(tts, numTaskTrackers, ttm.getNumberOfUniqueHosts());
}
} else {
return null;
}
}
/** Wait until the system is idle. */
public void waitUntilIdle() {
waitTaskTrackers();
JobClient client;
try {
client = new JobClient(job);
ClusterStatus status = client.getClusterStatus();
while (status.getTaskTrackers() + numTrackerToExclude < taskTrackerList.size()) {
for (TaskTrackerRunner runner : taskTrackerList) {
if (runner.isDead) {
throw new RuntimeException("TaskTracker is dead");
}
}
Thread.sleep(1000);
status = client.getClusterStatus();
}
} catch (IOException ex) {
throw new RuntimeException(ex);
} catch (InterruptedException ex) {
throw new RuntimeException(ex);
}
}
/** When no input dir is specified, generate random data. */
protected static void confRandom(JobConf job) throws IOException {
// from RandomWriter
job.setInputFormat(RandomInputFormat.class);
job.setMapperClass(RandomMapOutput.class);
final ClusterStatus cluster = new JobClient(job).getClusterStatus();
int numMapsPerHost = job.getInt(RandomTextWriter.MAPS_PER_HOST, 10);
long numBytesToWritePerMap =
job.getLong(RandomTextWriter.BYTES_PER_MAP, 1 * 1024 * 1024 * 1024);
if (numBytesToWritePerMap == 0) {
throw new IOException("Cannot have " + RandomTextWriter.BYTES_PER_MAP + " set to 0");
}
long totalBytesToWrite =
job.getLong(
RandomTextWriter.TOTAL_BYTES,
numMapsPerHost * numBytesToWritePerMap * cluster.getTaskTrackers());
int numMaps = (int) (totalBytesToWrite / numBytesToWritePerMap);
if (numMaps == 0 && totalBytesToWrite > 0) {
numMaps = 1;
job.setLong(RandomTextWriter.BYTES_PER_MAP, totalBytesToWrite);
}
job.setNumMapTasks(numMaps);
}
private boolean exceededPadding(
boolean isMapTask, ClusterStatus clusterStatus, int maxTaskTrackerSlots) {
int numTaskTrackers = clusterStatus.getTaskTrackers();
int totalTasks = (isMapTask) ? clusterStatus.getMapTasks() : clusterStatus.getReduceTasks();
int totalTaskCapacity =
isMapTask ? clusterStatus.getMaxMapTasks() : clusterStatus.getMaxReduceTasks();
Collection<JobInProgress> jobQueue = jobQueueJobInProgressListener.getJobQueue();
boolean exceededPadding = false;
synchronized (jobQueue) {
int totalNeededTasks = 0;
for (JobInProgress job : jobQueue) {
if (job.getStatus().getRunState() != JobStatus.RUNNING || job.numReduceTasks == 0) {
continue;
}
//
// Beyond the highest-priority task, reserve a little
// room for failures and speculative executions; don't
// schedule tasks to the hilt.
//
totalNeededTasks += isMapTask ? job.desiredMaps() : job.desiredReduces();
int padding = 0;
if (numTaskTrackers > MIN_CLUSTER_SIZE_FOR_PADDING) {
padding = Math.min(maxTaskTrackerSlots, (int) (totalNeededTasks * padFraction));
}
if (totalTasks + padding >= totalTaskCapacity) {
exceededPadding = true;
break;
}
}
}
return exceededPadding;
}
@Override
public synchronized List<Task> assignTasks(TaskTracker taskTracker) throws IOException {
TaskTrackerStatus taskTrackerStatus = taskTracker.getStatus();
ClusterStatus clusterStatus = taskTrackerManager.getClusterStatus();
final int numTaskTrackers = clusterStatus.getTaskTrackers();
final int clusterMapCapacity = clusterStatus.getMaxMapTasks();
final int clusterReduceCapacity = clusterStatus.getMaxReduceTasks();
Collection<JobInProgress> jobQueue = jobQueueJobInProgressListener.getJobQueue();
//
// Get map + reduce counts for the current tracker.
//
final int trackerMapCapacity = taskTrackerStatus.getMaxMapSlots();
final int trackerReduceCapacity = taskTrackerStatus.getMaxReduceSlots();
final int trackerRunningMaps = taskTrackerStatus.countMapTasks();
final int trackerRunningReduces = taskTrackerStatus.countReduceTasks();
// Assigned tasks
List<Task> assignedTasks = new ArrayList<Task>();
//
// Compute (running + pending) map and reduce task numbers across pool
//
int remainingReduceLoad = 0;
int remainingMapLoad = 0;
synchronized (jobQueue) {
for (JobInProgress job : jobQueue) {
if (job.getStatus().getRunState() == JobStatus.RUNNING) {
remainingMapLoad += (job.desiredMaps() - job.finishedMaps());
if (job.scheduleReduces()) {
remainingReduceLoad += (job.desiredReduces() - job.finishedReduces());
}
}
}
}
// Compute the 'load factor' for maps and reduces
double mapLoadFactor = 0.0;
if (clusterMapCapacity > 0) {
mapLoadFactor = (double) remainingMapLoad / clusterMapCapacity;
}
double reduceLoadFactor = 0.0;
if (clusterReduceCapacity > 0) {
reduceLoadFactor = (double) remainingReduceLoad / clusterReduceCapacity;
}
//
// In the below steps, we allocate first map tasks (if appropriate),
// and then reduce tasks if appropriate. We go through all jobs
// in order of job arrival; jobs only get serviced if their
// predecessors are serviced, too.
//
//
// We assign tasks to the current taskTracker if the given machine
// has a workload that's less than the maximum load of that kind of
// task.
// However, if the cluster is close to getting loaded i.e. we don't
// have enough _padding_ for speculative executions etc., we only
// schedule the "highest priority" task i.e. the task from the job
// with the highest priority.
//
final int trackerCurrentMapCapacity =
Math.min((int) Math.ceil(mapLoadFactor * trackerMapCapacity), trackerMapCapacity);
int availableMapSlots = trackerCurrentMapCapacity - trackerRunningMaps;
boolean exceededMapPadding = false;
if (availableMapSlots > 0) {
exceededMapPadding = exceededPadding(true, clusterStatus, trackerMapCapacity);
}
int numLocalMaps = 0;
int numNonLocalMaps = 0;
boolean newIterationJob = false;
scheduleMaps:
for (int i = 0; i < availableMapSlots; ++i) {
synchronized (jobQueue) {
for (JobInProgress job : jobQueue) {
if (job.getStatus().getRunState() != JobStatus.RUNNING) {
continue;
}
if (job.getJobConf().isIterative()) {
String iterativeAppID = job.getJobConf().getIterativeAlgorithmID();
if (iterativeAppID.equals("none")) {
throw new IOException("please specify the iteration ID!");
}
String jointype = job.getJobConf().get("mapred.iterative.jointype");
// prepare the iterationid map and jobtask map
if (!this.tracker_mtask_map.containsKey(iterativeAppID)) {
// a new iterative algorithm
Map<String, LinkedList<Integer>> new_tracker_task_map =
new HashMap<String, LinkedList<Integer>>();
this.tracker_mtask_map.put(iterativeAppID, new_tracker_task_map);
Map<String, LinkedList<Integer>> new_tracker_rtask_map =
new HashMap<String, LinkedList<Integer>>();
this.tracker_rtask_map.put(iterativeAppID, new_tracker_rtask_map);
// record the first job of the series of jobs in the iterations
this.first_job_map.put(iterativeAppID, job.getJobID());
// record the list of jobs for a iteration
HashSet<JobID> jobs = new HashSet<JobID>();
jobs.add(job.getJobID());
this.iteration_jobs_map.put(iterativeAppID, jobs);
}
// this is the first job of the series of jobs
if (this.first_job_map.get(iterativeAppID).equals(job.getJobID())
&& job.getJobConf().isIterative()) {
LOG.info(job.getJobID() + " is the first iteration job");
newIterationJob = true;
}
// this is one of the following jobs, and prepare a assignment list for the assignment
if (!newIterationJob) {
LOG.info(job.getJobID() + " is not the first iteration job");
this.iteration_jobs_map.get(iterativeAppID).add(job.getJobID());
if (this.mtask_assign_map.get(job.getJobID()) == null) {
// prepare the map task assignment list
LOG.info("for job " + job.getJobID() + "'s assignment:");
Map<String, LinkedList<Integer>> map_task_assign =
new HashMap<String, LinkedList<Integer>>();
for (Map.Entry<String, LinkedList<Integer>> entry :
this.tracker_mtask_map.get(iterativeAppID).entrySet()) {
String tracker = entry.getKey();
LinkedList<Integer> taskids = entry.getValue();
LinkedList<Integer> copytaskids = new LinkedList<Integer>();
LOG.info("assign on tracker " + tracker);
for (int taskid : taskids) {
copytaskids.add(taskid);
LOG.info("task id " + taskid);
}
map_task_assign.put(tracker, copytaskids);
}
this.mtask_assign_map.put(job.getJobID(), map_task_assign);
// if one2one copy the map assign to reduce assign, the are with the same mapping
if (jointype.equals("one2one")) {
// prepare the reduce task assignment list
Map<String, LinkedList<Integer>> reduce_task_assign =
new HashMap<String, LinkedList<Integer>>();
for (Map.Entry<String, LinkedList<Integer>> entry :
this.tracker_mtask_map.get(iterativeAppID).entrySet()) {
String tracker = entry.getKey();
LinkedList<Integer> taskids = entry.getValue();
LinkedList<Integer> copytaskids = new LinkedList<Integer>();
for (int taskid : taskids) {
copytaskids.add(taskid);
}
reduce_task_assign.put(tracker, copytaskids);
}
this.tracker_rtask_map.put(iterativeAppID, reduce_task_assign);
}
// prepare the reduce task assignment list for all cases
Map<String, LinkedList<Integer>> reduce_task_assign =
new HashMap<String, LinkedList<Integer>>();
for (Map.Entry<String, LinkedList<Integer>> entry :
this.tracker_rtask_map.get(iterativeAppID).entrySet()) {
String tracker = entry.getKey();
LinkedList<Integer> taskids = entry.getValue();
LinkedList<Integer> copytaskids = new LinkedList<Integer>();
for (int taskid : taskids) {
copytaskids.add(taskid);
}
reduce_task_assign.put(tracker, copytaskids);
}
this.rtask_assign_map.put(job.getJobID(), reduce_task_assign);
}
}
Task t = null;
// the first iteration or following iteration
// if the first iteration: assign taskid by default (exception for the one2mul case,
// where we assign staring from 0,...,n)
// else if the following iterations: assign taskid based on the first iteration
// assignment
if (newIterationJob) {
/**
* the one2mul case should be carefully taken care, we want to assgin map0,map1,map2
* and reduce0 to a tracker, and assign map3,map4,map5 and reduce1 to another tracker
*/
if (jointype.equals("one2mul")
&& !tracker_rtask_map
.get(iterativeAppID)
.containsKey(taskTracker.getTrackerName())) {
// if contain the tracker, that means we have assigned tasks for this tracker
int scala = job.getJobConf().getInt("mapred.iterative.data.scala", 1);
// int mapsEachTracker = job.getJobConf().getNumMapTasks() / numTaskTrackers;
int reducersEachTracker = job.getJobConf().getNumReduceTasks() / numTaskTrackers;
if (job.getJobConf().getNumReduceTasks() % numTaskTrackers != 0)
throw new IOException(
"job.getJobConf().getNumReduceTasks() % numTaskTrackers != 0");
if (!this.tracker_mtask_map
.get(iterativeAppID)
.containsKey(taskTracker.getTrackerName())) {
LinkedList<Integer> tasklist = new LinkedList<Integer>();
this.tracker_mtask_map
.get(iterativeAppID)
.put(taskTracker.getTrackerName(), tasklist);
}
if (!this.tracker_rtask_map
.get(iterativeAppID)
.containsKey(taskTracker.getTrackerName())) {
LinkedList<Integer> tasklist = new LinkedList<Integer>();
this.tracker_rtask_map
.get(iterativeAppID)
.put(taskTracker.getTrackerName(), tasklist);
}
// for debugging
String debugout1 = "maps: ";
String debugout2 = "reduces: ";
int reduceOffsetId =
(tracker_rtask_map.get(iterativeAppID).size() - 1)
* reducersEachTracker; // the start reduce id
for (int count = 0; count < reducersEachTracker; count++) {
int reducepartitionid = reduceOffsetId + count;
debugout2 += reducepartitionid + " ";
tracker_rtask_map
.get(iterativeAppID)
.get(taskTracker.getTrackerName())
.add(reducepartitionid);
for (int count2 = 0; count2 < scala; count2++) {
int mappartitionid = reducepartitionid * scala + count2;
// int mapid = job.splitTaskMap.get(mappartitionid);
debugout1 += mappartitionid + " ";
this.tracker_mtask_map
.get(iterativeAppID)
.get(taskTracker.getTrackerName())
.add(mappartitionid);
}
}
// print out for debug
LOG.info(
"tracker "
+ taskTracker.getTrackerName()
+ " assigned tasks "
+ debugout1
+ " and "
+ debugout2);
// make the assignment list
String tracker = taskTracker.getTrackerName();
LinkedList<Integer> mtaskids =
this.tracker_mtask_map.get(iterativeAppID).get(taskTracker.getTrackerName());
LinkedList<Integer> mcopytaskids = new LinkedList<Integer>();
for (int taskid : mtaskids) {
mcopytaskids.add(taskid);
}
if (!mtask_assign_map.containsKey(job.getJobID())) {
Map<String, LinkedList<Integer>> map_task_assign =
new HashMap<String, LinkedList<Integer>>();
this.mtask_assign_map.put(job.getJobID(), map_task_assign);
}
this.mtask_assign_map.get(job.getJobID()).put(tracker, mcopytaskids);
// prepare the reduce task assignment list
LinkedList<Integer> rtaskids =
this.tracker_rtask_map.get(iterativeAppID).get(taskTracker.getTrackerName());
LinkedList<Integer> rcopytaskids = new LinkedList<Integer>();
for (int taskid : rtaskids) {
rcopytaskids.add(taskid);
}
if (!rtask_assign_map.containsKey(job.getJobID())) {
Map<String, LinkedList<Integer>> reduce_task_assign =
new HashMap<String, LinkedList<Integer>>();
this.rtask_assign_map.put(job.getJobID(), reduce_task_assign);
}
this.rtask_assign_map.get(job.getJobID()).put(tracker, rcopytaskids);
// assign a map task for this tracker
Integer target = null;
try {
target =
this.mtask_assign_map
.get(job.getJobID())
.get(taskTracker.getTrackerName())
.peekFirst();
} catch (Exception e) {
e.printStackTrace();
}
if (target == null) {
// all have been assigned, no more work, maybe it should help others to process
LOG.info(
"all map tasks on tasktracker "
+ taskTracker.getTrackerName()
+ " have been processed");
break;
} else {
t =
job.obtainNewNodeOrRackLocalMapTask(
taskTrackerStatus,
numTaskTrackers,
taskTrackerManager.getNumberOfUniqueHosts(),
target);
}
} else {
t =
job.obtainNewNodeOrRackLocalMapTask(
taskTrackerStatus,
numTaskTrackers,
taskTrackerManager.getNumberOfUniqueHosts());
}
} else {
Integer target = null;
try {
target =
this.mtask_assign_map
.get(job.getJobID())
.get(taskTracker.getTrackerName())
.peekFirst();
} catch (Exception e) {
e.printStackTrace();
}
if (target == null) {
// all have been assigned, no more work, maybe it should help others to process
LOG.info(
"all map tasks on tasktracker "
+ taskTracker.getTrackerName()
+ " have been processed");
break;
} else {
t =
job.obtainNewNodeOrRackLocalMapTask(
taskTrackerStatus,
numTaskTrackers,
taskTrackerManager.getNumberOfUniqueHosts(),
target);
}
}
if (t != null) {
assignedTasks.add(t);
++numLocalMaps;
// new iteration job and the first task for a tasktracker
// for one2mul case, we don't need to record the assignment, since we already made the
// assignment list beforehand
if (!newIterationJob || jointype.equals("one2mul")) {
// poll, remove
this.mtask_assign_map
.get(job.getJobID())
.get(taskTracker.getTrackerName())
.pollFirst();
LOG.info("assigning task " + t.getTaskID() + " on " + taskTracker.getTrackerName());
} else {
// record the assignment list for map tasks
if (!this.tracker_mtask_map
.get(iterativeAppID)
.containsKey(taskTracker.getTrackerName())) {
LinkedList<Integer> tasklist = new LinkedList<Integer>();
this.tracker_mtask_map
.get(iterativeAppID)
.put(taskTracker.getTrackerName(), tasklist);
}
this.tracker_mtask_map
.get(iterativeAppID)
.get(taskTracker.getTrackerName())
.add(t.getTaskID().getTaskID().getId());
// prepare the reduce assignment, for mapping with reduce
if (jointype.equals("one2one")) {
// prepare the reduce assignment, for mapping with reduce
if (!first_job_reduces_map.containsKey(iterativeAppID)) {
Map<String, LinkedList<Integer>> tracker_reduce_map =
new HashMap<String, LinkedList<Integer>>();
first_job_reduces_map.put(iterativeAppID, tracker_reduce_map);
}
if (!first_job_reduces_map
.get(iterativeAppID)
.containsKey(taskTracker.getTrackerName())) {
LinkedList<Integer> reduces = new LinkedList<Integer>();
first_job_reduces_map
.get(iterativeAppID)
.put(taskTracker.getTrackerName(), reduces);
}
first_job_reduces_map
.get(iterativeAppID)
.get(taskTracker.getTrackerName())
.add(t.getTaskID().getTaskID().getId());
}
LOG.info("assigning task " + t.getTaskID() + " on " + taskTracker.getTrackerName());
}
// Don't assign map tasks to the hilt!
// Leave some free slots in the cluster for future task-failures,
// speculative tasks etc. beyond the highest priority job
if (exceededMapPadding) {
break scheduleMaps;
}
// Try all jobs again for the next Map task
break;
}
LOG.error("New Node Or Rack Local Map Task failed!");
if (newIterationJob) {
// Try to schedule a node-local or rack-local Map task
t =
job.obtainNewNonLocalMapTask(
taskTrackerStatus,
numTaskTrackers,
taskTrackerManager.getNumberOfUniqueHosts());
} else {
Integer target =
this.mtask_assign_map
.get(job.getJobID())
.get(taskTracker.getTrackerName())
.peekFirst();
if (target == null) {
// all have been assigned, no more work, maybe it should help others to process
LOG.info(
"all map tasks on tasktracker "
+ taskTracker.getTrackerName()
+ " have been processed");
break;
} else {
t =
job.obtainNewNonLocalMapTask(
taskTrackerStatus,
numTaskTrackers,
taskTrackerManager.getNumberOfUniqueHosts(),
target);
}
}
if (t != null) {
assignedTasks.add(t);
++numNonLocalMaps;
// new iteration job and the first task for a tasktracker
if (newIterationJob) {
if (!this.tracker_mtask_map
.get(iterativeAppID)
.containsKey(taskTracker.getTrackerName())) {
LinkedList<Integer> tasklist = new LinkedList<Integer>();
this.tracker_mtask_map
.get(iterativeAppID)
.put(taskTracker.getTrackerName(), tasklist);
}
this.tracker_mtask_map
.get(iterativeAppID)
.get(taskTracker.getTrackerName())
.add(t.getTaskID().getTaskID().getId());
} else {
// poll, remove
this.mtask_assign_map
.get(job.getJobID())
.get(taskTracker.getTrackerName())
.pollFirst();
}
// We assign at most 1 off-switch or speculative task
// This is to prevent TaskTrackers from stealing local-tasks
// from other TaskTrackers.
break scheduleMaps;
}
} else {
// not an iterative algorithm, normal schedule
Task t = null;
// Try to schedule a node-local or rack-local Map task
t =
job.obtainNewNodeOrRackLocalMapTask(
taskTrackerStatus,
numTaskTrackers,
taskTrackerManager.getNumberOfUniqueHosts());
if (t != null) {
assignedTasks.add(t);
++numLocalMaps;
// Don't assign map tasks to the hilt!
// Leave some free slots in the cluster for future task-failures,
// speculative tasks etc. beyond the highest priority job
if (exceededMapPadding) {
break scheduleMaps;
}
// Try all jobs again for the next Map task
break;
}
// Try to schedule a node-local or rack-local Map task
t =
job.obtainNewNonLocalMapTask(
taskTrackerStatus,
numTaskTrackers,
taskTrackerManager.getNumberOfUniqueHosts());
if (t != null) {
assignedTasks.add(t);
++numNonLocalMaps;
// We assign at most 1 off-switch or speculative task
// This is to prevent TaskTrackers from stealing local-tasks
// from other TaskTrackers.
break scheduleMaps;
}
}
}
}
}
int assignedMaps = assignedTasks.size();
//
// Same thing, but for reduce tasks
// However we _never_ assign more than 1 reduce task per heartbeat
//
/** should maintain the reduce task location for the termination check */
final int trackerCurrentReduceCapacity =
Math.min((int) Math.ceil(reduceLoadFactor * trackerReduceCapacity), trackerReduceCapacity);
final int availableReduceSlots =
Math.min((trackerCurrentReduceCapacity - trackerRunningReduces), 1);
boolean exceededReducePadding = false;
// LOG.info("availableReduceSlots " + availableReduceSlots);
if (availableReduceSlots > 0) {
exceededReducePadding = exceededPadding(false, clusterStatus, trackerReduceCapacity);
synchronized (jobQueue) {
for (JobInProgress job : jobQueue) {
LOG.info("job " + job.getJobID());
if (job.getStatus().getRunState() != JobStatus.RUNNING || job.numReduceTasks == 0) {
LOG.info("have to continue " + job.getStatus().getRunState());
continue;
}
Task t = null;
if (job.getJobConf().isIterative()) {
String iterativeAppID = job.getJobConf().getIterativeAlgorithmID();
if (iterativeAppID.equals("none")) {
throw new IOException("please specify the iteration ID!");
}
String jointype = job.getJobConf().get("mapred.iterative.jointype");
if (jointype.equals("one2one")) {
// one-to-one or one-to-mul jobs
if (this.first_job_map.get(iterativeAppID).equals(job.getJobID())
&& job.getJobConf().isIterative()) {
LOG.info(job.getJobID() + " is the first iteration job for reduce");
newIterationJob = true;
}
Integer target = null;
if (newIterationJob) {
if (first_job_reduces_map.get(iterativeAppID) == null) {
throw new IOException(
"I think something is wrong since the tasktracker never receive "
+ "a map task with iterativeapp id "
+ iterativeAppID);
}
if (first_job_reduces_map.get(iterativeAppID).get(taskTracker.getTrackerName())
== null) {
throw new IOException(
"I think something is wrong since the tasktracker never receive "
+ "a map task with iterativeapp id "
+ iterativeAppID
+ " from "
+ taskTracker.getTrackerName());
}
target =
this.first_job_reduces_map
.get(iterativeAppID)
.get(taskTracker.getTrackerName())
.pollFirst();
} else {
// the task assignment has already been processed during the map task assignment, so
// never use tracker_rtask_map
target =
this.rtask_assign_map
.get(job.getJobID())
.get(taskTracker.getTrackerName())
.pollFirst();
}
if (target == null) {
// all have been assigned, no more work, maybe it should help others to process
LOG.info(
"all reduce tasks on tasktracker "
+ taskTracker.getTrackerName()
+ " have been processed");
break;
} else {
t =
job.obtainNewReduceTask(
taskTrackerStatus,
numTaskTrackers,
taskTrackerManager.getNumberOfUniqueHosts(),
target);
}
} else if (jointype.equals("one2mul")) {
Integer target =
this.rtask_assign_map
.get(job.getJobID())
.get(taskTracker.getTrackerName())
.pollFirst();
if (target == null) {
// all have been assigned, no more work, maybe it should help others to process
LOG.info(
"all reduce tasks on tasktracker "
+ taskTracker.getTrackerName()
+ " have been processed");
break;
} else {
t =
job.obtainNewReduceTask(
taskTrackerStatus,
numTaskTrackers,
taskTrackerManager.getNumberOfUniqueHosts(),
target);
}
} else {
// one-to-all case, assign tasks in the first iteration job, and remember this mapping
// this is the first job of the series of jobs
if (this.first_job_map.get(iterativeAppID).equals(job.getJobID())) {
LOG.info(job.getJobID() + " is the first iteration job for reduce");
newIterationJob = true;
}
/*
//this is one of the following jobs, and prepare a assignment list for the assignment
else{
LOG.info(job.getJobID() + " is not the first iteration job for reduce");
if(this.rtask_assign_map.get(job.getJobID()) == null){
//prepare the map task assignment list
Map<String, LinkedList<Integer>> reduce_task_assign = new HashMap<String, LinkedList<Integer>>();
for(Map.Entry<String, LinkedList<Integer>> entry : this.tracker_rtask_map.get(iterativeAppID).entrySet()){
String tracker = entry.getKey();
LinkedList<Integer> taskids = entry.getValue();
LinkedList<Integer> copytaskids = new LinkedList<Integer>();
for(int taskid : taskids){
copytaskids.add(taskid);
}
reduce_task_assign.put(tracker, copytaskids);
}
this.rtask_assign_map.put(job.getJobID(), reduce_task_assign);
}
}
*/
// the first iteration or following iteration
// if the first iteration: assign taskid by default
// else if the following iterations: assign taskid based on the first iteration
// assignment
if (newIterationJob) {
t =
job.obtainNewReduceTask(
taskTrackerStatus,
numTaskTrackers,
taskTrackerManager.getNumberOfUniqueHosts());
if (t != null) {
if (!this.tracker_rtask_map
.get(iterativeAppID)
.containsKey(taskTracker.getTrackerName())) {
LinkedList<Integer> tasklist = new LinkedList<Integer>();
this.tracker_rtask_map
.get(iterativeAppID)
.put(taskTracker.getTrackerName(), tasklist);
}
this.tracker_rtask_map
.get(iterativeAppID)
.get(taskTracker.getTrackerName())
.add(t.getTaskID().getTaskID().getId());
LOG.info(
"assigning reduce task "
+ t.getTaskID()
+ " on "
+ taskTracker.getTrackerName());
}
} else {
Integer target =
this.rtask_assign_map
.get(job.getJobID())
.get(taskTracker.getTrackerName())
.peekFirst();
if (target == null) {
// all have been assigned, no more work, maybe it should help others to process
LOG.info(
"all map tasks on tasktracker "
+ taskTracker.getTrackerName()
+ " have been processed");
break;
} else {
t =
job.obtainNewReduceTask(
taskTrackerStatus,
numTaskTrackers,
taskTrackerManager.getNumberOfUniqueHosts(),
target);
}
if (t != null) {
// poll, remove
this.rtask_assign_map
.get(job.getJobID())
.get(taskTracker.getTrackerName())
.pollFirst();
LOG.info(
"assigning reduce task "
+ t.getTaskID()
+ " on "
+ taskTracker.getTrackerName());
}
}
}
} else {
t =
job.obtainNewReduceTask(
taskTrackerStatus,
numTaskTrackers,
taskTrackerManager.getNumberOfUniqueHosts());
}
LOG.info("try to assign new task " + t);
if (t != null) {
assignedTasks.add(t);
break;
}
// Don't assign reduce tasks to the hilt!
// Leave some free slots in the cluster for future task-failures,
// speculative tasks etc. beyond the highest priority job
if (exceededReducePadding) {
break;
}
}
}
}
if (LOG.isDebugEnabled()) {
LOG.debug(
"Task assignments for "
+ taskTrackerStatus.getTrackerName()
+ " --> "
+ "["
+ mapLoadFactor
+ ", "
+ trackerMapCapacity
+ ", "
+ trackerCurrentMapCapacity
+ ", "
+ trackerRunningMaps
+ "] -> ["
+ (trackerCurrentMapCapacity - trackerRunningMaps)
+ ", "
+ assignedMaps
+ " ("
+ numLocalMaps
+ ", "
+ numNonLocalMaps
+ ")] ["
+ reduceLoadFactor
+ ", "
+ trackerReduceCapacity
+ ", "
+ trackerCurrentReduceCapacity
+ ","
+ trackerRunningReduces
+ "] -> ["
+ (trackerCurrentReduceCapacity - trackerRunningReduces)
+ ", "
+ (assignedTasks.size() - assignedMaps)
+ "]");
}
return assignedTasks;
}
/*
* TODO:
* For Elf: need to change the major schedule logic, scheduling need
* to be *datacenter-aware*
* */
@Override
public synchronized List<Task> assignTasks(TaskTracker taskTracker) throws IOException {
TaskTrackerStatus taskTrackerStatus = taskTracker.getStatus();
ClusterStatus clusterStatus = taskTrackerManager.getClusterStatus();
final int numTaskTrackers = clusterStatus.getTaskTrackers();
final int clusterMapCapacity = clusterStatus.getMaxMapTasks();
final int clusterReduceCapacity = clusterStatus.getMaxReduceTasks();
Collection<JobInProgress> jobQueue = jobQueueJobInProgressListener.getJobQueue();
//
// Get map + reduce counts for the current tracker.
//
final int trackerMapCapacity = taskTrackerStatus.getMaxMapSlots();
final int trackerReduceCapacity = taskTrackerStatus.getMaxReduceSlots();
final int trackerRunningMaps = taskTrackerStatus.countMapTasks();
final int trackerRunningReduces = taskTrackerStatus.countReduceTasks();
// Assigned tasks
List<Task> assignedTasks = new ArrayList<Task>();
//
// Compute (running + pending) map and reduce task numbers across pool
//
int remainingReduceLoad = 0;
int remainingMapLoad = 0;
synchronized (jobQueue) {
for (JobInProgress job : jobQueue) {
if (job.getStatus().getRunState() == JobStatus.RUNNING) {
remainingMapLoad += (job.desiredMaps() - job.finishedMaps());
if (job.scheduleReduces()) {
remainingReduceLoad += (job.desiredReduces() - job.finishedReduces());
}
}
}
}
// Compute the 'load factor' for maps and reduces
double mapLoadFactor = 0.0;
if (clusterMapCapacity > 0) {
mapLoadFactor = (double) remainingMapLoad / clusterMapCapacity;
}
double reduceLoadFactor = 0.0;
if (clusterReduceCapacity > 0) {
reduceLoadFactor = (double) remainingReduceLoad / clusterReduceCapacity;
}
//
// In the below steps, we allocate first map tasks (if appropriate),
// and then reduce tasks if appropriate. We go through all jobs
// in order of job arrival; jobs only get serviced if their
// predecessors are serviced, too.
//
//
// We assign tasks to the current taskTracker if the given machine
// has a workload that's less than the maximum load of that kind of
// task.
// However, if the cluster is close to getting loaded i.e. we don't
// have enough _padding_ for speculative executions etc., we only
// schedule the "highest priority" task i.e. the task from the job
// with the highest priority.
//
final int trackerCurrentMapCapacity =
Math.min((int) Math.ceil(mapLoadFactor * trackerMapCapacity), trackerMapCapacity);
int availableMapSlots = trackerCurrentMapCapacity - trackerRunningMaps;
boolean exceededMapPadding = false;
if (availableMapSlots > 0) {
exceededMapPadding = exceededPadding(true, clusterStatus, trackerMapCapacity);
}
int numLocalMaps = 0;
int numNonLocalMaps = 0;
scheduleMaps:
// TODO: for Elf
// The main schedule logic here, outer for loop is for every slot, inner loop is for each job
for (int i = 0; i < availableMapSlots; ++i) {
synchronized (jobQueue) {
for (JobInProgress job : jobQueue) {
if (job.getStatus().getRunState() != JobStatus.RUNNING) {
continue;
}
Task t = null;
// Try to schedule a node-local or rack-local Map task
t =
job.obtainNewLocalMapTask(
taskTrackerStatus, numTaskTrackers, taskTrackerManager.getNumberOfUniqueHosts());
if (t != null) {
assignedTasks.add(t);
++numLocalMaps;
// Don't assign map tasks to the hilt!
// Leave some free slots in the cluster for future task-failures,
// speculative tasks etc. beyond the highest priority job
if (exceededMapPadding) {
break scheduleMaps;
}
// Try all jobs again for the next Map task
// Note: it's FIFO here: next time in the inner for loop the head-of-queue
// will still be chosen
break;
}
// If no locality for this job, try launching non-local
// Try to schedule a node-local or rack-local Map task --> original comments
// FIXME: is the above comment correct? seems should be non-local task
t =
job.obtainNewNonLocalMapTask(
taskTrackerStatus, numTaskTrackers, taskTrackerManager.getNumberOfUniqueHosts());
if (t != null) {
assignedTasks.add(t);
++numNonLocalMaps;
// We assign at most 1 off-switch or speculative task
// This is to prevent TaskTrackers from stealing local-tasks
// from other TaskTrackers.
break scheduleMaps;
}
}
}
}
int assignedMaps = assignedTasks.size();
//
// Same thing, but for reduce tasks
// However we _never_ assign more than 1 reduce task per heartbeat
//
final int trackerCurrentReduceCapacity =
Math.min((int) Math.ceil(reduceLoadFactor * trackerReduceCapacity), trackerReduceCapacity);
final int availableReduceSlots =
Math.min((trackerCurrentReduceCapacity - trackerRunningReduces), 1);
boolean exceededReducePadding = false;
if (availableReduceSlots > 0) {
exceededReducePadding = exceededPadding(false, clusterStatus, trackerReduceCapacity);
synchronized (jobQueue) {
for (JobInProgress job : jobQueue) {
if (job.getStatus().getRunState() != JobStatus.RUNNING || job.numReduceTasks == 0) {
continue;
}
Task t =
job.obtainNewReduceTask(
taskTrackerStatus, numTaskTrackers, taskTrackerManager.getNumberOfUniqueHosts());
if (t != null) {
assignedTasks.add(t);
break;
}
// Don't assign reduce tasks to the hilt!
// Leave some free slots in the cluster for future task-failures,
// speculative tasks etc. beyond the highest priority job
if (exceededReducePadding) {
break;
}
}
}
}
if (LOG.isDebugEnabled()) {
LOG.debug(
"Task assignments for "
+ taskTrackerStatus.getTrackerName()
+ " --> "
+ "["
+ mapLoadFactor
+ ", "
+ trackerMapCapacity
+ ", "
+ trackerCurrentMapCapacity
+ ", "
+ trackerRunningMaps
+ "] -> ["
+ (trackerCurrentMapCapacity - trackerRunningMaps)
+ ", "
+ assignedMaps
+ " ("
+ numLocalMaps
+ ", "
+ numNonLocalMaps
+ ")] ["
+ reduceLoadFactor
+ ", "
+ trackerReduceCapacity
+ ", "
+ trackerCurrentReduceCapacity
+ ","
+ trackerRunningReduces
+ "] -> ["
+ (trackerCurrentReduceCapacity - trackerRunningReduces)
+ ", "
+ (assignedTasks.size() - assignedMaps)
+ "]");
}
return assignedTasks;
}
/**
* Performs a range query using MapReduce
*
* @param fs
* @param inputFile
* @param queryRange
* @param shape
* @param output
* @return
* @throws IOException
*/
public static long rangeQueryMapReduce(
FileSystem fs,
Path inputFile,
Path userOutputPath,
Shape queryShape,
Shape shape,
boolean overwrite,
boolean background,
QueryInput query)
throws IOException {
JobConf job = new JobConf(FileMBR.class);
FileSystem outFs = inputFile.getFileSystem(job);
Path outputPath = userOutputPath;
if (outputPath == null) {
do {
outputPath =
new Path(
inputFile.toUri().getPath() + ".rangequery_" + (int) (Math.random() * 1000000));
} while (outFs.exists(outputPath));
} else {
if (outFs.exists(outputPath)) {
if (overwrite) {
outFs.delete(outputPath, true);
} else {
throw new RuntimeException("Output path already exists and -overwrite flag is not set");
}
}
}
job.setJobName("RangeQuery");
job.setClass(SpatialSite.FilterClass, RangeFilter.class, BlockFilter.class);
RangeFilter.setQueryRange(job, queryShape); // Set query range for
// filter
ClusterStatus clusterStatus = new JobClient(job).getClusterStatus();
job.setNumMapTasks(clusterStatus.getMaxMapTasks() * 5);
job.setNumReduceTasks(3);
// Decide which map function to use depending on how blocks are indexed
// And also which input format to use
if (SpatialSite.isRTree(fs, inputFile)) {
// RTree indexed file
LOG.info("Searching an RTree indexed file");
job.setInputFormat(RTreeInputFormat.class);
} else {
// A file with no local index
LOG.info("Searching a non local-indexed file");
job.setInputFormat(ShapeInputFormat.class);
}
GlobalIndex<Partition> gIndex = SpatialSite.getGlobalIndex(fs, inputFile);
// if (gIndex != null && gIndex.isReplicated()){
// job.setMapperClass(RangeQueryMap.class);
Class<?> OutputKey = NullWritable.class;
try {
Class<?> c = shape.getClass();
Field f = c.getDeclaredField(query.field);
f.setAccessible(true);
if (f.getType().equals(Integer.TYPE)) {
OutputKey = IntWritable.class;
} else if (f.getType().equals(Double.TYPE)) {
OutputKey = DoubleWritable.class;
} else if (f.getType().equals(Long.TYPE)) {
OutputKey = LongWritable.class;
}
} catch (SecurityException e) {
e.printStackTrace();
} catch (NoSuchFieldException e) {
// TODO Auto-generated catch block
e.printStackTrace();
}
job.setMapOutputKeyClass(OutputKey);
switch (query.type) {
case Distinct:
job.setMapperClass(DistinctQueryMap.class);
job.setReducerClass(DistinctQueryReduce.class);
job.setMapOutputValueClass(NullWritable.class);
break;
case Distribution:
job.setMapperClass(DistributionQueryMap.class);
job.setReducerClass(DistributionQueryReduce.class);
job.setMapOutputValueClass(IntWritable.class);
break;
default:
break;
}
// }
// else
// job.setMapperClass(RangeQueryMapNoDupAvoidance.class);
// Set query range for the map function
job.set(QUERY_SHAPE_CLASS, queryShape.getClass().getName());
job.set(QUERY_SHAPE, queryShape.toText(new Text()).toString());
job.set(QUERY_FIELD, query.field);
// Set shape class for the SpatialInputFormat
SpatialSite.setShapeClass(job, shape.getClass());
job.setOutputFormat(TextOutputFormat.class);
ShapeInputFormat.setInputPaths(job, inputFile);
TextOutputFormat.setOutputPath(job, outputPath);
// Submit the job
if (!background) {
RunningJob runningJob = JobClient.runJob(job);
Counters counters = runningJob.getCounters();
Counter outputRecordCounter = counters.findCounter(Task.Counter.MAP_OUTPUT_RECORDS);
final long resultCount = outputRecordCounter.getValue();
// If outputPath not set by user, automatically delete it
if (userOutputPath == null) outFs.delete(outputPath, true);
return resultCount;
} else {
JobClient jc = new JobClient(job);
lastRunningJob = jc.submitJob(job);
return -1;
}
}
/**
* 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;
}
public void _jspService(HttpServletRequest request, HttpServletResponse response)
throws java.io.IOException, ServletException {
PageContext pageContext = null;
HttpSession session = null;
ServletContext application = null;
ServletConfig config = null;
JspWriter out = null;
Object page = this;
JspWriter _jspx_out = null;
PageContext _jspx_page_context = null;
try {
response.setContentType("text/html; charset=UTF-8");
pageContext = _jspxFactory.getPageContext(this, request, response, null, true, 8192, true);
_jspx_page_context = pageContext;
application = pageContext.getServletContext();
config = pageContext.getServletConfig();
session = pageContext.getSession();
out = pageContext.getOut();
_jspx_out = out;
_jspx_resourceInjector =
(org.apache.jasper.runtime.ResourceInjector)
application.getAttribute("com.sun.appserv.jsp.resource.injector");
out.write('\n');
JobTracker tracker = (JobTracker) application.getAttribute("job.tracker");
ClusterStatus status = tracker.getClusterStatus();
String trackerName = StringUtils.simpleHostname(tracker.getJobTrackerMachine());
out.write("\n<html>\n<head>\n<title>");
out.print(trackerName);
out.write(
" Hadoop Locality Statistics</title>\n<link rel=\"stylesheet\" type=\"text/css\" href=\"/static/hadoop.css\">\n</head>\n<body>\n<h1>");
out.print(trackerName);
out.write(" Hadoop Locality Statistics</h1>\n\n<b>State:</b> ");
out.print(status.getJobTrackerState());
out.write("<br>\n<b>Started:</b> ");
out.print(new Date(tracker.getStartTime()));
out.write("<br>\n<b>Version:</b> ");
out.print(VersionInfo.getVersion());
out.write(",\n r");
out.print(VersionInfo.getRevision());
out.write("<br>\n<b>Compiled:</b> ");
out.print(VersionInfo.getDate());
out.write(" by\n ");
out.print(VersionInfo.getUser());
out.write("<br>\n<b>Identifier:</b> ");
out.print(tracker.getTrackerIdentifier());
out.write("<br>\n\n<hr>\n\n");
Collection<JobInProgress> jobs = new ArrayList<JobInProgress>();
jobs.addAll(tracker.completedJobs());
jobs.addAll(tracker.runningJobs());
jobs.addAll(tracker.failedJobs());
int dataLocalMaps = 0;
int rackLocalMaps = 0;
int totalMaps = 0;
int totalReduces = 0;
for (JobInProgress job : jobs) {
Counters counters = job.getCounters();
dataLocalMaps += counters.getCounter(JobInProgress.Counter.DATA_LOCAL_MAPS);
rackLocalMaps += counters.getCounter(JobInProgress.Counter.RACK_LOCAL_MAPS);
totalMaps += counters.getCounter(JobInProgress.Counter.TOTAL_LAUNCHED_MAPS);
totalReduces += counters.getCounter(JobInProgress.Counter.TOTAL_LAUNCHED_REDUCES);
}
int dataLocalMapPct = totalMaps == 0 ? 0 : (100 * dataLocalMaps) / totalMaps;
int rackLocalMapPct = totalMaps == 0 ? 0 : (100 * rackLocalMaps) / totalMaps;
int dataRackLocalMapPct =
totalMaps == 0 ? 0 : (100 * (dataLocalMaps + rackLocalMaps)) / totalMaps;
out.write("\n<p>\n<b>Data Local Maps:</b> ");
out.print(dataLocalMaps);
out.write(' ');
out.write('(');
out.print(dataLocalMapPct);
out.write("%) <br>\n<b>Rack Local Maps:</b> ");
out.print(rackLocalMaps);
out.write(' ');
out.write('(');
out.print(rackLocalMapPct);
out.write("%) <br>\n<b>Data or Rack Local:</b> ");
out.print(dataLocalMaps + rackLocalMaps);
out.write(' ');
out.write('(');
out.print(dataRackLocalMapPct);
out.write("%) <br>\n<b>Total Maps:</b> ");
out.print(totalMaps);
out.write(" <br>\n<b>Total Reduces:</b> ");
out.print(totalReduces);
out.write(" <br>\n</p>\n\n");
out.println(ServletUtil.htmlFooter());
out.write('\n');
} catch (Throwable t) {
if (!(t instanceof SkipPageException)) {
out = _jspx_out;
if (out != null && out.getBufferSize() != 0) out.clearBuffer();
if (_jspx_page_context != null) _jspx_page_context.handlePageException(t);
}
} finally {
_jspxFactory.releasePageContext(_jspx_page_context);
}
}
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.");
}
