org.apache.hadoop.mapred
Class JobInProgress

java.lang.Object
  org.apache.hadoop.mapred.JobInProgress

public class JobInProgress
extends Object

JobInProgress maintains all the info for keeping a Job on the straight and narrow. It keeps its JobProfile and its latest JobStatus, plus a set of tables for doing bookkeeping of its Tasks. ***********************************************************

Nested Class Summary

static class

JobInProgress.Counter

Field Summary

protected Credentials

tokenStorage

Constructor Summary

protected

JobInProgress(JobID jobid, JobConf conf, JobTracker tracker) Create an almost empty JobInProgress, which can be used only for tests

Method Summary

void	cleanUpMetrics() Called when the job is complete
boolean	completedTask(org.apache.hadoop.mapred.TaskInProgress tip, TaskStatus status) A taskid assigned to this JobInProgress has reported in successfully.
int	desiredMaps()
int	desiredReduces()
int	desiredTasks() Return total number of map and reduce tasks desired by the job.
void	failedTask(org.apache.hadoop.mapred.TaskInProgress tip, TaskAttemptID taskid, String reason, TaskStatus.Phase phase, TaskStatus.State state, String trackerName) Fail a task with a given reason, but without a status object.
TaskStatus	findFinishedMap(int mapId) Find the details of someplace where a map has finished
protected org.apache.hadoop.mapred.TaskInProgress	findSpeculativeTask(Collection<org.apache.hadoop.mapred.TaskInProgress> list, TaskTrackerStatus ttStatus, double avgProgress, long currentTime, boolean shouldRemove) Find a speculative task
int	finishedMaps()
int	finishedReduces()
boolean	getCounters(Counters result) Returns the total job counters, by adding together the job, the map and the reduce counters.
long	getFinishTime()
Counters	getJobCounters() Returns the job-level counters.
JobID	getJobID()
String	getJobSubmitHostAddress()
String	getJobSubmitHostName()
long	getLaunchTime()
boolean	getMapCounters(Counters counters) Returns map phase counters by summing over all map tasks in progress.
int	getNumReservedTaskTrackersForMaps()
int	getNumReservedTaskTrackersForReduces()
long	getNumSchedulingOpportunities()
int	getNumSlotsPerTask(TaskType taskType)
JobPriority	getPriority()
JobProfile	getProfile()
org.apache.hadoop.mapred.QueueMetrics	getQueueMetrics() Get the QueueMetrics object associated with this job
boolean	getReduceCounters(Counters counters) Returns map phase counters by summing over all map tasks in progress.
Object	getSchedulingInfo()
long	getStartTime()
JobStatus	getStatus()
TaskCompletionEvent[]	getTaskCompletionEvents(int fromEventId, int maxEvents)
org.apache.hadoop.mapred.TaskInProgress	getTaskInProgress(TaskID tipid) Return the TaskInProgress that matches the tipid.
String	getUser() Get the user for the job
boolean	inited() Check if the job has been initialized.
void	initTasks() Construct the splits, etc.
void	kill() Kill the job and all its component tasks.
Task	obtainJobCleanupTask(TaskTrackerStatus tts, int clusterSize, int numUniqueHosts, boolean isMapSlot) Return a CleanupTask, if appropriate, to run on the given tasktracker
Task	obtainJobSetupTask(TaskTrackerStatus tts, int clusterSize, int numUniqueHosts, boolean isMapSlot) Return a SetupTask, if appropriate, to run on the given tasktracker
Task	obtainNewMapTask(TaskTrackerStatus tts, int clusterSize, int numUniqueHosts) Return a MapTask, if appropriate, to run on the given tasktracker
Task	obtainNewNodeLocalMapTask(TaskTrackerStatus tts, int clusterSize, int numUniqueHosts)
Task	obtainNewNodeOrRackLocalMapTask(TaskTrackerStatus tts, int clusterSize, int numUniqueHosts)
Task	obtainNewNonLocalMapTask(TaskTrackerStatus tts, int clusterSize, int numUniqueHosts)
Task	obtainNewReduceTask(TaskTrackerStatus tts, int clusterSize, int numUniqueHosts) Return a ReduceTask, if appropriate, to run on the given tasktracker.
Task	obtainTaskCleanupTask(TaskTrackerStatus tts, boolean isMapSlot)
void	overrideSchedulingOpportunities()
int	pendingMaps()
int	pendingReduces()
Vector<org.apache.hadoop.mapred.TaskInProgress>	reportCleanupTIPs(boolean shouldBeComplete) Return a vector of cleanup TaskInProgress objects
Vector<org.apache.hadoop.mapred.TaskInProgress>	reportSetupTIPs(boolean shouldBeComplete) Return a vector of setup TaskInProgress objects
Vector<org.apache.hadoop.mapred.TaskInProgress>	reportTasksInProgress(boolean shouldBeMap, boolean shouldBeComplete) Return a vector of completed TaskInProgress objects
void	reserveTaskTracker(TaskTracker taskTracker, TaskType type, int numSlots)
void	resetSchedulingOpportunities()
int	runningMaps()
int	runningReduces()
protected void	scheduleMap(org.apache.hadoop.mapred.TaskInProgress tip) Adds a map tip to the list of running maps.
boolean	scheduleOffSwitch(int numTaskTrackers) Check if we can schedule an off-switch task for this job.
protected void	scheduleReduce(org.apache.hadoop.mapred.TaskInProgress tip) Adds a reduce tip to the list of running reduces
boolean	scheduleReduces()
void	schedulingOpportunity()
void	setPriority(JobPriority priority)
void	setSchedulingInfo(Object schedulingInfo)
void	unreserveTaskTracker(TaskTracker taskTracker, TaskType type)
void	updateTaskStatus(org.apache.hadoop.mapred.TaskInProgress tip, TaskStatus status) Assuming JobTracker is locked on entry.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

tokenStorage

protected Credentials tokenStorage

Constructor Detail

JobInProgress

protected JobInProgress(JobID jobid,
                        JobConf conf,
                        JobTracker tracker)
                 throws IOException

Create an almost empty JobInProgress, which can be used only for tests

Throws:

IOException

Method Detail

getQueueMetrics

public org.apache.hadoop.mapred.QueueMetrics getQueueMetrics()

Get the QueueMetrics object associated with this job

Returns:

QueueMetrics

cleanUpMetrics

public void cleanUpMetrics()

Called when the job is complete

inited

public boolean inited()

Check if the job has been initialized.

Returns:

true if the job has been initialized, false otherwise

getUser

public String getUser()

Get the user for the job

initTasks

public void initTasks()
               throws IOException,
                      org.apache.hadoop.mapred.JobInProgress.KillInterruptedException,
                      UnknownHostException

Construct the splits, etc. This is invoked from an async thread so that split-computation doesn't block anyone.

Throws:

IOException

org.apache.hadoop.mapred.JobInProgress.KillInterruptedException

UnknownHostException

getProfile

public JobProfile getProfile()

getStatus

public JobStatus getStatus()

getLaunchTime

public long getLaunchTime()

getStartTime

public long getStartTime()

getFinishTime

public long getFinishTime()

desiredMaps

public int desiredMaps()

finishedMaps

public int finishedMaps()

desiredReduces

public int desiredReduces()

runningMaps

public int runningMaps()

runningReduces

public int runningReduces()

finishedReduces

public int finishedReduces()

pendingMaps

public int pendingMaps()

pendingReduces

public int pendingReduces()

desiredTasks

public int desiredTasks()

Return total number of map and reduce tasks desired by the job.

Returns:

total number of map and reduce tasks desired by the job

getNumSlotsPerTask

public int getNumSlotsPerTask(TaskType taskType)

getPriority

public JobPriority getPriority()

setPriority

public void setPriority(JobPriority priority)

reportTasksInProgress

public Vector<org.apache.hadoop.mapred.TaskInProgress> reportTasksInProgress(boolean shouldBeMap,
                                                                             boolean shouldBeComplete)

Return a vector of completed TaskInProgress objects

reportCleanupTIPs

public Vector<org.apache.hadoop.mapred.TaskInProgress> reportCleanupTIPs(boolean shouldBeComplete)

Return a vector of cleanup TaskInProgress objects

reportSetupTIPs

public Vector<org.apache.hadoop.mapred.TaskInProgress> reportSetupTIPs(boolean shouldBeComplete)

Return a vector of setup TaskInProgress objects

updateTaskStatus

public void updateTaskStatus(org.apache.hadoop.mapred.TaskInProgress tip,
                             TaskStatus status)

Assuming JobTracker is locked on entry.

getJobCounters

public Counters getJobCounters()

Returns the job-level counters.

Returns:

the job-level counters.

getMapCounters

public boolean getMapCounters(Counters counters)

Returns map phase counters by summing over all map tasks in progress. This method returns true if counters are within limit or false.

getReduceCounters

public boolean getReduceCounters(Counters counters)

Returns map phase counters by summing over all map tasks in progress. This method returns true if counters are within limits and false otherwise.

getCounters

public boolean getCounters(Counters result)

Returns the total job counters, by adding together the job, the map and the reduce counters. This method returns true if counters are within limits and false otherwise.

obtainNewMapTask

public Task obtainNewMapTask(TaskTrackerStatus tts,
                             int clusterSize,
                             int numUniqueHosts)
                      throws IOException

Return a MapTask, if appropriate, to run on the given tasktracker

Throws:

IOException

obtainTaskCleanupTask

public Task obtainTaskCleanupTask(TaskTrackerStatus tts,
                                  boolean isMapSlot)
                           throws IOException

Throws:

IOException

obtainNewNodeLocalMapTask

public Task obtainNewNodeLocalMapTask(TaskTrackerStatus tts,
                                      int clusterSize,
                                      int numUniqueHosts)
                               throws IOException

Throws:

IOException

obtainNewNodeOrRackLocalMapTask

public Task obtainNewNodeOrRackLocalMapTask(TaskTrackerStatus tts,
                                            int clusterSize,
                                            int numUniqueHosts)
                                     throws IOException

Throws:

IOException

obtainNewNonLocalMapTask

public Task obtainNewNonLocalMapTask(TaskTrackerStatus tts,
                                     int clusterSize,
                                     int numUniqueHosts)
                              throws IOException

Throws:

IOException

schedulingOpportunity

public void schedulingOpportunity()

resetSchedulingOpportunities

public void resetSchedulingOpportunities()

getNumSchedulingOpportunities

public long getNumSchedulingOpportunities()

overrideSchedulingOpportunities

public void overrideSchedulingOpportunities()

scheduleOffSwitch

public boolean scheduleOffSwitch(int numTaskTrackers)

Check if we can schedule an off-switch task for this job.

Parameters:

numTaskTrackers - number of tasktrackers

Returns:

true if we can schedule off-switch, false otherwise We check the number of missed opportunities for the job. If it has 'waited' long enough we go ahead and schedule.

obtainJobCleanupTask

public Task obtainJobCleanupTask(TaskTrackerStatus tts,
                                 int clusterSize,
                                 int numUniqueHosts,
                                 boolean isMapSlot)
                          throws IOException

Return a CleanupTask, if appropriate, to run on the given tasktracker

Throws:

IOException

obtainJobSetupTask

public Task obtainJobSetupTask(TaskTrackerStatus tts,
                               int clusterSize,
                               int numUniqueHosts,
                               boolean isMapSlot)
                        throws IOException

Return a SetupTask, if appropriate, to run on the given tasktracker

Throws:

IOException

scheduleReduces

public boolean scheduleReduces()

obtainNewReduceTask

public Task obtainNewReduceTask(TaskTrackerStatus tts,
                                int clusterSize,
                                int numUniqueHosts)
                         throws IOException

Return a ReduceTask, if appropriate, to run on the given tasktracker. We don't have cache-sensitivity for reduce tasks, as they work on temporary MapRed files.

Throws:

IOException

reserveTaskTracker

public void reserveTaskTracker(TaskTracker taskTracker,
                               TaskType type,
                               int numSlots)

unreserveTaskTracker

public void unreserveTaskTracker(TaskTracker taskTracker,
                                 TaskType type)

getNumReservedTaskTrackersForMaps

public int getNumReservedTaskTrackersForMaps()

getNumReservedTaskTrackersForReduces

public int getNumReservedTaskTrackersForReduces()

scheduleMap

protected void scheduleMap(org.apache.hadoop.mapred.TaskInProgress tip)

Adds a map tip to the list of running maps.

Parameters:

tip - the tip that needs to be scheduled as running

scheduleReduce

protected void scheduleReduce(org.apache.hadoop.mapred.TaskInProgress tip)

Adds a reduce tip to the list of running reduces

Parameters:

tip - the tip that needs to be scheduled as running

findSpeculativeTask

protected org.apache.hadoop.mapred.TaskInProgress findSpeculativeTask(Collection<org.apache.hadoop.mapred.TaskInProgress> list,
                                                                      TaskTrackerStatus ttStatus,
                                                                      double avgProgress,
                                                                      long currentTime,
                                                                      boolean shouldRemove)

Find a speculative task

Parameters:

list - a list of tips

ttStatus - status of the tracker that has requested a tip

avgProgress - the average progress for speculation

currentTime - current time in milliseconds

shouldRemove - whether to remove the tips

Returns:

a tip that can be speculated on the tracker

completedTask

public boolean completedTask(org.apache.hadoop.mapred.TaskInProgress tip,
                             TaskStatus status)

A taskid assigned to this JobInProgress has reported in successfully.

kill

public void kill()

Kill the job and all its component tasks. This method should be called from jobtracker and should return fast as it locks the jobtracker.

failedTask

public void failedTask(org.apache.hadoop.mapred.TaskInProgress tip,
                       TaskAttemptID taskid,
                       String reason,
                       TaskStatus.Phase phase,
                       TaskStatus.State state,
                       String trackerName)

Fail a task with a given reason, but without a status object. Assuming JobTracker is locked on entry.

Parameters:

tip - The task's tip

taskid - The task id

reason - The reason that the task failed

trackerName - The task tracker the task failed on

getTaskInProgress

public org.apache.hadoop.mapred.TaskInProgress getTaskInProgress(TaskID tipid)

Return the TaskInProgress that matches the tipid.

findFinishedMap

public TaskStatus findFinishedMap(int mapId)

Find the details of someplace where a map has finished

Parameters:

mapId - the id of the map

Returns:

the task status of the completed task

getTaskCompletionEvents

public TaskCompletionEvent[] getTaskCompletionEvents(int fromEventId,
                                                     int maxEvents)

getJobID

public JobID getJobID()

Returns:

The JobID of this JobInProgress.

getJobSubmitHostName

public String getJobSubmitHostName()

Returns:

submitHostName of this JobInProgress.

getJobSubmitHostAddress

public String getJobSubmitHostAddress()

Returns:

submitHostAddress of this JobInProgress.

getSchedulingInfo

public Object getSchedulingInfo()

setSchedulingInfo

public void setSchedulingInfo(Object schedulingInfo)