org.apache.hadoop.mapreduce

Class OutputCommitter

java.lang.Object

org.apache.hadoop.mapreduce.OutputCommitter

Direct Known Subclasses:

OutputCommitter, PathOutputCommitter

@InterfaceAudience.Public
 @InterfaceStability.Stable
public abstract class OutputCommitter
extends Object

OutputCommitter describes the commit of task output for a Map-Reduce job.

The Map-Reduce framework relies on the OutputCommitter of the job to:

1. Setup the job during initialization. For example, create the temporary output directory for the job during the initialization of the job.
2. Cleanup the job after the job completion. For example, remove the temporary output directory after the job completion.
3. Setup the task temporary output.
4. Check whether a task needs a commit. This is to avoid the commit procedure if a task does not need commit.
5. Commit of the task output.
6. Discard the task commit.

The methods in this class can be called from several different processes and from several different contexts. It is important to know which process and which context each is called from. Each method should be marked accordingly in its documentation. It is also important to note that not all methods are guaranteed to be called once and only once. If a method is not guaranteed to have this property the output committer needs to handle this appropriately. Also note it will only be in rare situations where they may be called multiple times for the same task.

See Also:

FileOutputCommitter, JobContext, TaskAttemptContext

Constructor Summary

Constructors

Constructor and Description
OutputCommitter()

Method Summary

Modifier and Type	Method and Description
void	abortJob(JobContext jobContext, JobStatus.State state) For aborting an unsuccessful job's output.
abstract void	abortTask(TaskAttemptContext taskContext) Discard the task output.
void	cleanupJob(JobContext jobContext) Deprecated. Use commitJob(JobContext) and abortJob(JobContext, JobStatus.State) instead.
void	commitJob(JobContext jobContext) For committing job's output after successful job completion.
abstract void	commitTask(TaskAttemptContext taskContext) To promote the task's temporary output to final output location.
boolean	isCommitJobRepeatable(JobContext jobContext) Returns true if an in-progress job commit can be retried.
boolean	isRecoverySupported() Deprecated. Use isRecoverySupported(JobContext) instead.
boolean	isRecoverySupported(JobContext jobContext) Is task output recovery supported for restarting jobs? If task output recovery is supported, job restart can be done more efficiently.
abstract boolean	needsTaskCommit(TaskAttemptContext taskContext) Check whether task needs a commit.
void	recoverTask(TaskAttemptContext taskContext) Recover the task output.
abstract void	setupJob(JobContext jobContext) For the framework to setup the job output during initialization.
abstract void	setupTask(TaskAttemptContext taskContext) Sets up output for the task.

Methods inherited from class java.lang.Object

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

Constructor Detail

OutputCommitter

public OutputCommitter()

Method Detail

setupJob

public abstract void setupJob(JobContext jobContext)
                       throws IOException

For the framework to setup the job output during initialization. This is called from the application master process for the entire job. This will be called multiple times, once per job attempt.

Parameters:

jobContext - Context of the job whose output is being written.

Throws:

IOException - if temporary output could not be created

cleanupJob

@Deprecated
public void cleanupJob(JobContext jobContext)
                            throws IOException

Deprecated. Use commitJob(JobContext) and abortJob(JobContext, JobStatus.State) instead.

For cleaning up the job's output after job completion. This is called from the application master process for the entire job. This may be called multiple times.

Parameters:

jobContext - Context of the job whose output is being written.

Throws:

IOException

commitJob

public void commitJob(JobContext jobContext)
               throws IOException

For committing job's output after successful job completion. Note that this is invoked for jobs with final runstate as SUCCESSFUL. This is called from the application master process for the entire job. This is guaranteed to only be called once. If it throws an exception the entire job will fail.

Parameters:

jobContext - Context of the job whose output is being written.

Throws:

IOException

abortJob

public void abortJob(JobContext jobContext,
                     JobStatus.State state)
              throws IOException

For aborting an unsuccessful job's output. Note that this is invoked for jobs with final runstate as JobStatus.State.FAILED or JobStatus.State.KILLED. This is called from the application master process for the entire job. This may be called multiple times.

Parameters:

jobContext - Context of the job whose output is being written.

state - final runstate of the job

Throws:

IOException

setupTask

public abstract void setupTask(TaskAttemptContext taskContext)
                        throws IOException

Sets up output for the task. This is called from each individual task's process that will output to HDFS, and it is called just for that task. This may be called multiple times for the same task, but for different task attempts.

Parameters:

taskContext - Context of the task whose output is being written.

Throws:

IOException

needsTaskCommit

public abstract boolean needsTaskCommit(TaskAttemptContext taskContext)
                                 throws IOException

Check whether task needs a commit. This is called from each individual task's process that will output to HDFS, and it is called just for that task.

Parameters:

taskContext -

Returns:

true/false

Throws:

IOException

commitTask

public abstract void commitTask(TaskAttemptContext taskContext)
                         throws IOException

To promote the task's temporary output to final output location. If needsTaskCommit(TaskAttemptContext) returns true and this task is the task that the AM determines finished first, this method is called to commit an individual task's output. This is to mark that tasks output as complete, as commitJob(JobContext) will also be called later on if the entire job finished successfully. This is called from a task's process. This may be called multiple times for the same task, but different task attempts. It should be very rare for this to be called multiple times and requires odd networking failures to make this happen. In the future the Hadoop framework may eliminate this race.

Parameters:

taskContext - Context of the task whose output is being written.

Throws:

IOException - if commit is not successful.

abortTask

public abstract void abortTask(TaskAttemptContext taskContext)
                        throws IOException

Discard the task output. This is called from a task's process to clean up a single task's output that can not yet been committed. This may be called multiple times for the same task, but for different task attempts.

Parameters:

taskContext -

Throws:

IOException

isRecoverySupported

@Deprecated
public boolean isRecoverySupported()

Deprecated. Use isRecoverySupported(JobContext) instead.

Is task output recovery supported for restarting jobs? If task output recovery is supported, job restart can be done more efficiently.

Returns:

true if task output recovery is supported, false otherwise

See Also:

recoverTask(TaskAttemptContext)

isCommitJobRepeatable

public boolean isCommitJobRepeatable(JobContext jobContext)
                              throws IOException

Returns true if an in-progress job commit can be retried. If the MR AM is re-run then it will check this value to determine if it can retry an in-progress commit that was started by a previous version. Note that in rare scenarios, the previous AM version might still be running at that time, due to system anomalies. Hence if this method returns true then the retry commit operation should be able to run concurrently with the previous operation. If repeatable job commit is supported, job restart can tolerate previous AM failures during job commit. By default, it is not supported. Extended classes (like: FileOutputCommitter) should explicitly override it if provide support.

Parameters:

jobContext - Context of the job whose output is being written.

Returns:

true repeatable job commit is supported, false otherwise

Throws:

IOException

isRecoverySupported

public boolean isRecoverySupported(JobContext jobContext)
                            throws IOException

Is task output recovery supported for restarting jobs? If task output recovery is supported, job restart can be done more efficiently.

Parameters:

jobContext - Context of the job whose output is being written.

Returns:

true if task output recovery is supported, false otherwise

Throws:

IOException

See Also:

recoverTask(TaskAttemptContext)

recoverTask

public void recoverTask(TaskAttemptContext taskContext)
                 throws IOException

Recover the task output. The retry-count for the job will be passed via the MRJobConfig.APPLICATION_ATTEMPT_ID key in JobContext.getConfiguration() for the OutputCommitter. This is called from the application master process, but it is called individually for each task. If an exception is thrown the task will be attempted again. This may be called multiple times for the same task. But from different application attempts.

Parameters:

taskContext - Context of the task whose output is being recovered

Throws:

IOException