org.apache.tez.runtime.api

Interface TaskContext

All Known Subinterfaces:

InputContext, OutputContext, ProcessorContext

@InterfaceAudience.Public
public interface TaskContext

Base interface for Context classes used to initialize the Input, Output and Processor instances. This interface is not supposed to be implemented by users

Method Summary

Modifier and Type	Method and Description
void	fatalError(Throwable exception, String message) Deprecated. Replaced by reportFailure(TaskFailureType, Throwable, String) (FailureType, Throwable, String)} Note: To maintain compatibility, even though this method is named 'fatalError' - this method operates as reportFailure(TaskFailureType, Throwable, String) with the TaskFailureType set to TaskFailureType.NON_FATAL.
org.apache.hadoop.yarn.api.records.ApplicationId	getApplicationId() Get the ApplicationId for the running app
org.apache.hadoop.conf.Configuration	getContainerConfiguration() Get container configuration
TezCounters	getCounters()
int	getDAGAttemptNumber() Get the current DAG Attempt Number
int	getDagIdentifier() Get a numeric identifier for the dag to which the task belongs.
String	getDAGName() Get the name of the DAG
ExecutionContext	getExecutionContext() Get the context for the executor.
ObjectRegistry	getObjectRegistry() Returns a shared ObjectRegistry to hold user objects in memory between tasks.
ByteBuffer	getServiceConsumerMetaData(String serviceName) Returns meta-data for the specified service.
ByteBuffer	getServiceProviderMetaData(String serviceName) Return Provider meta-data for the specified service As an example, when the MR ShuffleHandler is used - this would return the shuffle port serialized as bytes
int	getTaskAttemptNumber() Get the current Task Attempt Number
int	getTaskIndex() Get the index of this Task among the tasks of this vertex
int	getTaskVertexIndex() Get the index of this task's vertex in the set of vertices in the DAG.
String	getTaskVertexName() Get the name of the Vertex in which the task is running
long	getTotalMemoryAvailableToTask() Gets the total memory available to all components of the running task.
String	getUniqueIdentifier() Returns an identifier which is unique to the specific Input, Processor or Output
UserPayload	getUserPayload() Get the User Payload for the Input/Output/Processor
int	getVertexParallelism() Get the vertex parallelism of the vertex to which this task belongs.
String[]	getWorkDirs() Get the work directories for the Input/Output/Processor
void	notifyProgress() Notifies the framework that progress is being made by this component.
void	reportFailure(org.apache.tez.runtime.api.TaskFailureType taskFailureType, Throwable exception, String message) Report an error to the framework.
void	requestInitialMemory(long size, MemoryUpdateCallback callbackHandler) Request a specific amount of memory during initialization (initialize(..*Context)) The requester is notified of allocation via the provided callback handler.
void	sendEvents(List<Event> events) Send Events to the AM and/or dependent Vertices

Method Detail

getApplicationId

org.apache.hadoop.yarn.api.records.ApplicationId getApplicationId()

Get the ApplicationId for the running app

Returns:

the ApplicationId

getDAGAttemptNumber

int getDAGAttemptNumber()

Get the current DAG Attempt Number

Returns:

DAG Attempt Number

getTaskIndex

int getTaskIndex()

Get the index of this Task among the tasks of this vertex

Returns:

Task Index

getTaskAttemptNumber

int getTaskAttemptNumber()

Get the current Task Attempt Number

Returns:

Task Attempt Number

getContainerConfiguration

org.apache.hadoop.conf.Configuration getContainerConfiguration()

Get container configuration

Returns:

Container configuration

getDAGName

String getDAGName()

Get the name of the DAG

Returns:

the DAG name

getTaskVertexName

String getTaskVertexName()

Get the name of the Vertex in which the task is running

Returns:

Vertex Name

getTaskVertexIndex

int getTaskVertexIndex()

Get the index of this task's vertex in the set of vertices in the DAG. This is consistent and valid across all tasks/vertices in the same DAG.

Returns:

index

getDagIdentifier

int getDagIdentifier()

Get a numeric identifier for the dag to which the task belongs. This will be unique within the running application.

Returns:

the dag identifier

getCounters

TezCounters getCounters()

sendEvents

void sendEvents(List<Event> events)

Send Events to the AM and/or dependent Vertices

Parameters:

events - Events to be sent

getUserPayload

UserPayload getUserPayload()

Get the User Payload for the Input/Output/Processor

Returns:

User Payload

getWorkDirs

String[] getWorkDirs()

Get the work directories for the Input/Output/Processor

Returns:

an array of work dirs

getUniqueIdentifier

String getUniqueIdentifier()

Returns an identifier which is unique to the specific Input, Processor or Output

Returns:

a unique identifier

getObjectRegistry

ObjectRegistry getObjectRegistry()

Returns a shared ObjectRegistry to hold user objects in memory between tasks.

Returns:

ObjectRegistry

notifyProgress

void notifyProgress()

Notifies the framework that progress is being made by this component. This is used to identify hung components that are not making progress. Must be called periodically until processing has completed for this component. Care must be taken to call this when real progress has been made. Simply calling this continuously from a thread without regard to real work may prevent identification of hung components and delay/stall job completion.

fatalError

@Deprecated
void fatalError(@Nullable
                            Throwable exception,
                            @Nullable
                            String message)

Deprecated. Replaced by reportFailure(TaskFailureType, Throwable, String) (FailureType, Throwable, String)} Note: To maintain compatibility, even though this method is named 'fatalError' - this method operates as reportFailure(TaskFailureType, Throwable, String) with the TaskFailureType set to TaskFailureType.NON_FATAL.

Report an error to the framework. This will cause the taskAttempt to fail, and should not be used to report errors which can be handled locally in the TaskAttempt. A new TaskAttempt will be launched depending upon how many retries are available for the task.

Parameters:

exception - an exception representing the error

message - a diagnostic message which may be associated with the error

reportFailure

void reportFailure(org.apache.tez.runtime.api.TaskFailureType taskFailureType,
                   @Nullable
                   Throwable exception,
                   @Nullable
                   String message)

Report an error to the framework. This will cause the entire task to be terminated.

Parameters:

taskFailureType - the type of the error

exception - any exception that may be associated with the error

message - a diagnostic message which may be associated with the error

getServiceConsumerMetaData

ByteBuffer getServiceConsumerMetaData(String serviceName)

Returns meta-data for the specified service. As an example, when the MR ShuffleHandler is used - this would return the jobToken serialized as bytes

Parameters:

serviceName - the name of the service for which meta-data is required

Returns:

a ByteBuffer representing the meta-data

getServiceProviderMetaData

@Nullable
ByteBuffer getServiceProviderMetaData(String serviceName)

Return Provider meta-data for the specified service As an example, when the MR ShuffleHandler is used - this would return the shuffle port serialized as bytes

Parameters:

serviceName - the name of the service for which provider meta-data is required

Returns:

a ByteBuffer representing the meta-data

requestInitialMemory

void requestInitialMemory(long size,
                          MemoryUpdateCallback callbackHandler)

Request a specific amount of memory during initialization (initialize(..*Context)) The requester is notified of allocation via the provided callback handler. Currently, (post TEZ-668) the caller will be informed about the available memory after initialization (I/P/O initialize(...)), and before the start/run invocation. There will be no other invocations on the callback. This method can be called only once by any component. Calling it multiple times from within the same component will result in an error. Each Input / Output must request memory. For Inputs / Outputs which do not have a specific ask, a null callback handler can be specified with a request size of 0.

Parameters:

size - request size in bytes.

callbackHandler - the callback handler to be invoked once memory is assigned

getTotalMemoryAvailableToTask

long getTotalMemoryAvailableToTask()

Gets the total memory available to all components of the running task. This values will always be constant, and does not factor in any allocations.

Returns:

the total available memory for all components of the task

getVertexParallelism

int getVertexParallelism()

Get the vertex parallelism of the vertex to which this task belongs.

Returns:

Parallelism of the current vertex.

getExecutionContext

ExecutionContext getExecutionContext()

Get the context for the executor. This may be shared across multiple tasks

Returns:

the execution context