public Vector completedJobs() {
Vector v = new Vector();
for (Iterator it = jobs.values().iterator(); it.hasNext(); ) {
JobInProgress jip = (JobInProgress) it.next();
JobStatus status = jip.getStatus();
if (status.getRunState() == JobStatus.SUCCEEDED) {
v.add(jip);
}
}
return v;
}
public Vector runningJobs() {
Vector v = new Vector();
for (Iterator it = jobs.values().iterator(); it.hasNext(); ) {
JobInProgress jip = (JobInProgress) it.next();
JobStatus status = jip.getStatus();
if (status.getRunState() == JobStatus.RUNNING) {
v.add(jip);
}
}
return v;
}
/**
* Check the ACLs for a user doing the passed operation.
*
* <ul>
* <li>If ACLs are disabled, allow all users.
* <li>Otherwise, if the operation is not a job operation(for eg. submit-job-to-queue), then
* allow only (a) clusterOwner(who started the cluster), (b) cluster administrators and (c)
* members of queue-submit-job-acl for the queue.
* <li>If the operation is a job operation, then allow only (a) jobOwner, (b) clusterOwner(who
* started the cluster), (c) cluster administrators, (d) members of queue admins acl for the
* queue and (e) members of job acl for the job operation
* </ul>
*
* @param job the job on which operation is requested
* @param callerUGI the user who is requesting the operation
* @param operation the operation for which authorization is needed
* @throws AccessControlException
*/
void checkAccess(JobInProgress job, UserGroupInformation callerUGI, Operation operation)
throws AccessControlException {
String queue = job.getProfile().getQueueName();
String jobId = job.getJobID().toString();
JobStatus jobStatus = job.getStatus();
String jobOwner = jobStatus.getUsername();
AccessControlList jobAcl = jobStatus.getJobACLs().get(operation.jobACLNeeded);
checkAccess(jobId, callerUGI, queue, operation, jobOwner, jobAcl);
}
/**
* Check the ACLs for a user doing the passed job operation.
*
* <ul>
* <li>If ACLs are disabled, allow all users.
* <li>Otherwise, allow only (a) jobOwner, (b) clusterOwner(who started the cluster), (c)
* cluster administrators, (d) members of job acl for the jobOperation
* </ul>
*
* @param jobStatus the status of the job
* @param callerUGI the user who is trying to perform the operation
* @param queue the job queue name
* @param operation the operation for which authorization is needed
*/
void checkAccess(
JobStatus jobStatus, UserGroupInformation callerUGI, String queue, Operation operation)
throws AccessControlException {
String jobId = jobStatus.getJobID().toString();
String jobOwner = jobStatus.getUsername();
AccessControlList jobAcl = jobStatus.getJobACLs().get(operation.jobACLNeeded);
// If acls are enabled, check if callerUGI is jobOwner, queue admin,
// cluster admin or part of job ACL
checkAccess(jobId, callerUGI, queue, operation, jobOwner, jobAcl);
}
public static JobStatus downgrade(org.apache.hadoop.mapreduce.JobStatus stat) {
JobStatus old =
new JobStatus(
JobID.downgrade(stat.getJobID()),
stat.getSetupProgress(),
stat.getMapProgress(),
stat.getReduceProgress(),
stat.getCleanupProgress(),
stat.getState().getValue(),
JobPriority.valueOf(stat.getPriority().name()),
stat.getUsername(),
stat.getJobName(),
stat.getJobFile(),
stat.getTrackingUrl());
old.setStartTime(stat.getStartTime());
old.setFinishTime(stat.getFinishTime());
old.setSchedulingInfo(stat.getSchedulingInfo());
old.setHistoryFile(stat.getHistoryFile());
return old;
}
private boolean isWait(JobInProgress job) {
long bookingTime = job.getJobConf().getLong(BOOKING_TIME, 0);
String[] dependencyJobs = job.getJobConf().getStrings(BOOKING_DEPENDENCY_JOBID, null);
boolean bookingTimeFilter = false;
boolean dependencyJobFilter = false;
if (bookingTime >= System.currentTimeMillis()) {
bookingTimeFilter = true;
}
if (null != dependencyJobs) {
for (String dependencyJob : dependencyJobs) {
JobStatus dependencyJobStatus = (JobStatus) finishJobStatus.get(dependencyJob);
if (null != dependencyJobStatus
&& dependencyJobStatus.getRunState() != JobStatus.SUCCEEDED) {
dependencyJobFilter = true;
}
}
}
if (bookingTimeFilter || dependencyJobFilter) return true;
else return false;
}
public synchronized boolean statusUpdate(
TaskAttemptID taskId, TaskStatus taskStatus, JvmContext context)
throws IOException, InterruptedException {
LOG.info(taskStatus.getStateString());
int taskIndex = mapIds.indexOf(taskId);
if (taskIndex >= 0) { // mapping
float numTasks = (float) this.numMapTasks;
partialMapProgress[taskIndex] = taskStatus.getProgress();
mapCounters[taskIndex] = taskStatus.getCounters();
float partialProgress = 0.0f;
for (float f : partialMapProgress) {
partialProgress += f;
}
status.setMapProgress(partialProgress / numTasks);
} else {
reduceCounters = taskStatus.getCounters();
status.setReduceProgress(taskStatus.getProgress());
}
// ignore phase
return true;
}
// for initTasks, update information from JobStory object
@Override
public synchronized void initTasks() throws IOException {
boolean loggingEnabled = LOG.isDebugEnabled();
if (loggingEnabled) {
LOG.debug("(initTasks@SJIP) Starting Initialization for " + jobId);
}
numMapTasks = jobStory.getNumberMaps();
numReduceTasks = jobStory.getNumberReduces();
JobHistory.JobInfo.logSubmitted(
getJobID(), conf, jobFile.toString(), this.startTime, hasRestarted());
if (loggingEnabled) {
LOG.debug("(initTasks@SJIP) Logged to job history for " + jobId);
}
// checkTaskLimits();
if (loggingEnabled) {
LOG.debug("(initTasks@SJIP) Checked task limits for " + jobId);
}
final String jobFile = "default";
splits = getRawSplits(jobStory.getInputSplits());
if (loggingEnabled) {
LOG.debug(
"(initTasks@SJIP) Created splits for job = "
+ jobId
+ " number of splits = "
+ splits.length);
}
// createMapTasks(jobFile, splits);
numMapTasks = splits.length;
maps = new TaskInProgress[numMapTasks];
for (int i = 0; i < numMapTasks; ++i) {
inputLength += splits[i].getDataLength();
maps[i] = new TaskInProgress(jobId, jobFile, splits[i], conf, this, i, numSlotsPerMap);
}
if (numMapTasks > 0) {
nonRunningMapCache = createCache(splits, maxLevel);
if (loggingEnabled) {
LOG.debug(
"initTasks:numMaps="
+ numMapTasks
+ " Size of nonRunningMapCache="
+ nonRunningMapCache.size()
+ " for "
+ jobId);
}
}
// set the launch time
this.launchTime = JobTracker.getClock().getTime();
// createReduceTasks(jobFile);
//
// Create reduce tasks
//
this.reduces = new TaskInProgress[numReduceTasks];
for (int i = 0; i < numReduceTasks; i++) {
reduces[i] =
new TaskInProgress(jobId, jobFile, numMapTasks, i, conf, this, numSlotsPerReduce);
nonRunningReduces.add(reduces[i]);
}
// Calculate the minimum number of maps to be complete before
// we should start scheduling reduces
completedMapsForReduceSlowstart =
(int)
Math.ceil(
(conf.getFloat(
"mapred.reduce.slowstart." + "completed.maps",
DEFAULT_COMPLETED_MAPS_PERCENT_FOR_REDUCE_SLOWSTART)
* numMapTasks));
tasksInited.set(true);
if (loggingEnabled) {
LOG.debug(
"Initializing job, nowstatus = " + JobStatus.getJobRunState(getStatus().getRunState()));
}
setupComplete();
if (loggingEnabled) {
LOG.debug(
"Initializing job, inited-status = "
+ JobStatus.getJobRunState(getStatus().getRunState()));
}
}
/** Indicate that one of the taskids in this TaskInProgress has failed. */
public void incompleteSubTask(TaskAttemptID taskid, JobStatus jobStatus) {
//
// Note the failure and its location
//
TaskStatus status = taskStatuses.get(taskid);
String trackerName;
String trackerHostName = null;
TaskStatus.State taskState = TaskStatus.State.FAILED;
if (status != null) {
trackerName = status.getTaskTracker();
trackerHostName = JobInProgressTraits.convertTrackerNameToHostName(trackerName);
// Check if the user manually KILLED/FAILED this task-attempt...
Boolean shouldFail = tasksToKill.remove(taskid);
if (shouldFail != null) {
if (status.getRunState() == TaskStatus.State.FAILED
|| status.getRunState() == TaskStatus.State.KILLED) {
taskState = (shouldFail) ? TaskStatus.State.FAILED : TaskStatus.State.KILLED;
} else {
taskState =
(shouldFail) ? TaskStatus.State.FAILED_UNCLEAN : TaskStatus.State.KILLED_UNCLEAN;
}
status.setRunState(taskState);
addDiagnosticInfo(taskid, "Task has been " + taskState + " by the user");
}
taskState = status.getRunState();
if (taskState != TaskStatus.State.FAILED
&& taskState != TaskStatus.State.KILLED
&& taskState != TaskStatus.State.FAILED_UNCLEAN
&& taskState != TaskStatus.State.KILLED_UNCLEAN) {
LOG.info(
"Task '"
+ taskid
+ "' running on '"
+ trackerName
+ "' in state: '"
+ taskState
+ "' being failed!");
status.setRunState(TaskStatus.State.FAILED);
taskState = TaskStatus.State.FAILED;
}
// tasktracker went down and failed time was not reported.
if (0 == status.getFinishTime()) {
status.setFinishTime(JobTracker.getClock().getTime());
}
}
this.activeTasks.remove(taskid);
// Since we do not fail completed reduces (whose outputs go to hdfs), we
// should note this failure only for completed maps, only if this taskid;
// completed this map. however if the job is done, there is no need to
// manipulate completed maps
if (this.isMapTask()
&& !jobSetup
&& !jobCleanup
&& isComplete(taskid)
&& jobStatus.getRunState() != JobStatus.SUCCEEDED) {
this.completes--;
// Reset the successfulTaskId since we don't have a SUCCESSFUL task now
resetSuccessfulTaskid();
}
// Note that there can be failures of tasks that are hosted on a machine
// that has not yet registered with restarted jobtracker
// recalculate the counts only if its a genuine failure
if (tasks.contains(taskid)) {
if (taskState == TaskStatus.State.FAILED) {
numTaskFailures++;
machinesWhereFailed.add(trackerHostName);
if (maxSkipRecords > 0) {
// skipping feature enabled
LOG.debug("TaskInProgress adding" + status.getNextRecordRange());
failedRanges.add(status.getNextRecordRange());
skipping = startSkipping();
}
} else if (taskState == TaskStatus.State.KILLED) {
numKilledTasks++;
}
}
if (numTaskFailures >= maxTaskAttempts) {
LOG.info("TaskInProgress " + getTIPId() + " has failed " + numTaskFailures + " times.");
kill();
}
}
@SuppressWarnings("unchecked")
@Override
public void run() {
JobID jobId = profile.getJobID();
JobContext jContext = new JobContextImpl(conf, jobId);
org.apache.hadoop.mapreduce.OutputCommitter outputCommitter = null;
try {
outputCommitter = createOutputCommitter(conf.getUseNewMapper(), jobId, conf);
} catch (Exception e) {
LOG.info("Failed to createOutputCommitter", e);
return;
}
try {
TaskSplitMetaInfo[] taskSplitMetaInfos =
SplitMetaInfoReader.readSplitMetaInfo(jobId, localFs, conf, systemJobDir);
int numReduceTasks = job.getNumReduceTasks();
if (numReduceTasks > 1 || numReduceTasks < 0) {
// we only allow 0 or 1 reducer in local mode
numReduceTasks = 1;
job.setNumReduceTasks(1);
}
outputCommitter.setupJob(jContext);
status.setSetupProgress(1.0f);
Map<TaskAttemptID, MapOutputFile> mapOutputFiles =
Collections.synchronizedMap(new HashMap<TaskAttemptID, MapOutputFile>());
List<MapTaskRunnable> taskRunnables =
getMapTaskRunnables(taskSplitMetaInfos, jobId, mapOutputFiles);
ExecutorService mapService = createMapExecutor(taskRunnables.size());
// Start populating the executor with work units.
// They may begin running immediately (in other threads).
for (Runnable r : taskRunnables) {
mapService.submit(r);
}
try {
mapService.shutdown(); // Instructs queue to drain.
// Wait for tasks to finish; do not use a time-based timeout.
// (See http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6179024)
LOG.info("Waiting for map tasks");
mapService.awaitTermination(Long.MAX_VALUE, TimeUnit.NANOSECONDS);
} catch (InterruptedException ie) {
// Cancel all threads.
mapService.shutdownNow();
throw ie;
}
LOG.info("Map task executor complete.");
// After waiting for the map tasks to complete, if any of these
// have thrown an exception, rethrow it now in the main thread context.
for (MapTaskRunnable r : taskRunnables) {
if (r.storedException != null) {
throw new Exception(r.storedException);
}
}
TaskAttemptID reduceId = new TaskAttemptID(new TaskID(jobId, false, 0), 0);
try {
if (numReduceTasks > 0) {
ReduceTask reduce =
new ReduceTask(systemJobFile.toString(), reduceId, 0, mapIds.size(), 1);
reduce.setUser(UserGroupInformation.getCurrentUser().getShortUserName());
JobConf localConf = new JobConf(job);
localConf.set("mapreduce.jobtracker.address", "local");
TaskRunner.setupChildMapredLocalDirs(reduce, localConf);
// move map output to reduce input
for (int i = 0; i < mapIds.size(); i++) {
if (!this.isInterrupted()) {
TaskAttemptID mapId = mapIds.get(i);
Path mapOut = mapOutputFiles.get(mapId).getOutputFile();
MapOutputFile localOutputFile = new MapOutputFile();
localOutputFile.setConf(localConf);
Path reduceIn =
localOutputFile.getInputFileForWrite(
mapId.getTaskID(), localFs.getFileStatus(mapOut).getLen());
if (!localFs.mkdirs(reduceIn.getParent())) {
throw new IOException(
"Mkdirs failed to create " + reduceIn.getParent().toString());
}
if (!localFs.rename(mapOut, reduceIn))
throw new IOException("Couldn't rename " + mapOut);
} else {
throw new InterruptedException();
}
}
if (!this.isInterrupted()) {
reduce.setJobFile(localJobFile.toString());
localConf.setUser(reduce.getUser());
reduce.localizeConfiguration(localConf);
reduce.setConf(localConf);
reduce_tasks += 1;
myMetrics.launchReduce(reduce.getTaskID());
reduce.run(localConf, this);
myMetrics.completeReduce(reduce.getTaskID());
reduce_tasks -= 1;
} else {
throw new InterruptedException();
}
}
} finally {
for (MapOutputFile output : mapOutputFiles.values()) {
output.removeAll();
}
}
// delete the temporary directory in output directory
outputCommitter.commitJob(jContext);
status.setCleanupProgress(1.0f);
if (killed) {
this.status.setRunState(JobStatus.KILLED);
} else {
this.status.setRunState(JobStatus.SUCCEEDED);
}
JobEndNotifier.localRunnerNotification(job, status);
} catch (Throwable t) {
try {
outputCommitter.abortJob(jContext, org.apache.hadoop.mapreduce.JobStatus.State.FAILED);
} catch (IOException ioe) {
LOG.info("Error cleaning up job:" + id);
}
status.setCleanupProgress(1.0f);
if (killed) {
this.status.setRunState(JobStatus.KILLED);
} else {
this.status.setRunState(JobStatus.FAILED);
}
LOG.warn(id, t);
JobEndNotifier.localRunnerNotification(job, status);
} finally {
try {
fs.delete(systemJobFile.getParent(), true); // delete submit dir
localFs.delete(localJobFile, true); // delete local copy
// Cleanup distributed cache
taskDistributedCacheManager.release();
trackerDistributedCacheManager.purgeCache();
} catch (IOException e) {
LOG.warn("Error cleaning up " + id + ": " + e);
}
}
}