org.apache.tez.serviceplugins.api

Interface TaskCommunicatorContext

All Superinterfaces:

org.apache.tez.serviceplugins.api.ServicePluginContextBase

All Known Implementing Classes:

TaskCommunicatorContextImpl

public interface TaskCommunicatorContext
extends org.apache.tez.serviceplugins.api.ServicePluginContextBase

Method Summary

Modifier and Type	Method and Description
boolean	canCommit(org.apache.tez.dag.records.TezTaskAttemptID taskAttemptId) Check whether a running attempt can commit.
void	containerAlive(org.apache.hadoop.yarn.api.records.ContainerId containerId) Inform the framework that a container is alive.
org.apache.hadoop.security.Credentials	getAMCredentials() Get credentials associated with the AppMaster
org.apache.hadoop.yarn.api.records.ApplicationAttemptId	getApplicationAttemptId() Get the application attempt id for the running application.
String	getCurrentAppIdentifier() Get an identifier for the executing context of the DAG.
long	getDagStartTime() Get the start time for the currently executing DAG
long	getFirstAttemptStartTime(String vertexName, int taskIndex) Get the start time for the first attempt of the specified task
Iterable<String>	getInputVertexNames(String vertexName) Get the name of the Input vertices for the specified vertex.
int	getVertexCompletedTaskCount(String vertexName) Get the number of completed tasks for a given vertex
int	getVertexRunningTaskCount(String vertexName) Get the number of running tasks for a given vertex
int	getVertexTotalTaskCount(String vertexName) Get the total number of tasks in the given vertex
TaskHeartbeatResponse	heartbeat(TaskHeartbeatRequest request) Mechanism for a TaskCommunicator to provide updates on a running task, as well as receive new information which may need to be propagated to the task.
boolean	isKnownContainer(org.apache.hadoop.yarn.api.records.ContainerId containerId) Check whether the container is known by the framework.
void	registerForVertexStateUpdates(String vertexName, Set<org.apache.tez.dag.api.event.VertexState> stateSet) Register to get notifications on updates to the specified vertex.
void	taskAlive(org.apache.tez.dag.records.TezTaskAttemptID taskAttemptId) Inform the framework that a task is alive.
void	taskFailed(org.apache.tez.dag.records.TezTaskAttemptID taskAttemptId, org.apache.tez.runtime.api.TaskFailureType taskFailureType, org.apache.tez.serviceplugins.api.TaskAttemptEndReason taskAttemptEndReason, String diagnostics) Inform the framework that a task has failed.
void	taskKilled(org.apache.tez.dag.records.TezTaskAttemptID taskAttemptId, org.apache.tez.serviceplugins.api.TaskAttemptEndReason taskAttemptEndReason, String diagnostics) Inform the framework that a task has been killed
void	taskStartedRemotely(org.apache.tez.dag.records.TezTaskAttemptID taskAttemptId) Inform the framework that the task has started execution
void	taskStartedRemotely(org.apache.tez.dag.records.TezTaskAttemptID taskAttemptId, org.apache.hadoop.yarn.api.records.ContainerId containerId) Deprecated.
void	taskSubmitted(org.apache.tez.dag.records.TezTaskAttemptID taskAttemptId, org.apache.hadoop.yarn.api.records.ContainerId containerId) Inform the framework that the task has been submitted for execution.

Methods inherited from interface org.apache.tez.serviceplugins.api.ServicePluginContextBase

getCurrentDagInfo, getInitialUserPayload, reportError

Method Detail

getApplicationAttemptId

org.apache.hadoop.yarn.api.records.ApplicationAttemptId getApplicationAttemptId()

Get the application attempt id for the running application. Relevant when running under YARN

Returns:

the applicationAttemptId for the running app

getAMCredentials

org.apache.hadoop.security.Credentials getAMCredentials()

Get credentials associated with the AppMaster

Returns:

credentials

canCommit

boolean canCommit(org.apache.tez.dag.records.TezTaskAttemptID taskAttemptId)
           throws IOException

Check whether a running attempt can commit. This provides a leader election mechanism amongst multiple running attempts

Parameters:

taskAttemptId - the associated task attempt id

Returns:

whether the attempt can commit or not

Throws:

IOException

heartbeat

TaskHeartbeatResponse heartbeat(TaskHeartbeatRequest request)
                         throws IOException,
                                org.apache.tez.dag.api.TezException

Mechanism for a TaskCommunicator to provide updates on a running task, as well as receive new information which may need to be propagated to the task. This includes events generated by the task and events which need to be sent to the task This method must be invoked periodically to receive updates for a running task

Parameters:

request - the update from the running task.

Returns:

the response that is requried by the task.

Throws:

IOException

org.apache.tez.dag.api.TezException

isKnownContainer

boolean isKnownContainer(org.apache.hadoop.yarn.api.records.ContainerId containerId)

Check whether the container is known by the framework. The state of this container is irrelevant

Parameters:

containerId - the relevant container id

Returns:

true if the container is known, false if it isn't

taskAlive

void taskAlive(org.apache.tez.dag.records.TezTaskAttemptID taskAttemptId)

Inform the framework that a task is alive. This needs to be invoked periodically to avoid the task attempt timing out. Invocations to heartbeat provides the same keep-alive functionality

Parameters:

taskAttemptId - the relevant task attempt

containerAlive

void containerAlive(org.apache.hadoop.yarn.api.records.ContainerId containerId)

Inform the framework that a container is alive. This need to be invoked periodically to avoid the container attempt timing out. Invocations to heartbeat provides the same keep-alive functionality

Parameters:

containerId - the relevant container id

taskSubmitted

void taskSubmitted(org.apache.tez.dag.records.TezTaskAttemptID taskAttemptId,
                   org.apache.hadoop.yarn.api.records.ContainerId containerId)

Inform the framework that the task has been submitted for execution. The expectation is that the implementing TaskCommunicator will inform the framework about task submission, followed by the task starting.

Parameters:

taskAttemptId - the relevant task attempt id

containerId - the containerId in which the task attempt is running

taskStartedRemotely

void taskStartedRemotely(org.apache.tez.dag.records.TezTaskAttemptID taskAttemptId)

Inform the framework that the task has started execution

Parameters:

taskAttemptId - the relevant task attempt id

taskStartedRemotely

@Deprecated
void taskStartedRemotely(org.apache.tez.dag.records.TezTaskAttemptID taskAttemptId,
                                     org.apache.hadoop.yarn.api.records.ContainerId containerId)

Deprecated.

Inform the framework that the task has started execution Use taskSubmitted(TezTaskAttemptID, ContainerId) and taskStartedRemotely(TezTaskAttemptID) instead

Parameters:

taskAttemptId - the relevant task attempt id

taskKilled

void taskKilled(org.apache.tez.dag.records.TezTaskAttemptID taskAttemptId,
                org.apache.tez.serviceplugins.api.TaskAttemptEndReason taskAttemptEndReason,
                @Nullable
                String diagnostics)

Inform the framework that a task has been killed

Parameters:

taskAttemptId - the relevant task attempt id

taskAttemptEndReason - the reason for the task attempt being killed

diagnostics - any diagnostics messages which are relevant to the task attempt kill

taskFailed

void taskFailed(org.apache.tez.dag.records.TezTaskAttemptID taskAttemptId,
                org.apache.tez.runtime.api.TaskFailureType taskFailureType,
                org.apache.tez.serviceplugins.api.TaskAttemptEndReason taskAttemptEndReason,
                @Nullable
                String diagnostics)

Inform the framework that a task has failed. This, at the moment, is always treated as a an error which will cause a retry of the task to be triggered, if there are enough retry attempts left.

Parameters:

taskAttemptId - the relevant task attempt id

taskFailureType - the type of the error

taskAttemptEndReason - the reason for the task failure

diagnostics - any diagnostics messages which are relevant to the task attempt failure

registerForVertexStateUpdates

void registerForVertexStateUpdates(String vertexName,
                                   @Nullable
                                   Set<org.apache.tez.dag.api.event.VertexState> stateSet)

Register to get notifications on updates to the specified vertex. Notifications will be sent via InputInitializer.onVertexStateUpdated(org.apache.tez.dag.api.event.VertexStateUpdate)

This method can only be invoked once. Duplicate invocations will result in an error.

Parameters:

vertexName - the vertex name for which notifications are required.

stateSet - the set of states for which notifications are required. null implies all

getCurrentAppIdentifier

String getCurrentAppIdentifier()

Get an identifier for the executing context of the DAG.

Returns:

a String identifier for the exeucting context.

getInputVertexNames

Iterable<String> getInputVertexNames(String vertexName)

Get the name of the Input vertices for the specified vertex. Root Inputs are not returned.

Parameters:

vertexName - the vertex for which source vertex names will be returned

Returns:

an Iterable containing the list of input vertices for the specified vertex

getVertexTotalTaskCount

int getVertexTotalTaskCount(String vertexName)

Get the total number of tasks in the given vertex

Parameters:

vertexName - the relevant vertex name

Returns:

total number of tasks in this vertex

getVertexCompletedTaskCount

int getVertexCompletedTaskCount(String vertexName)

Get the number of completed tasks for a given vertex

Parameters:

vertexName - the vertex name

Returns:

the number of completed tasks for the vertex

getVertexRunningTaskCount

int getVertexRunningTaskCount(String vertexName)

Get the number of running tasks for a given vertex

Parameters:

vertexName - the vertex name

Returns:

the number of running tasks for the vertex

getFirstAttemptStartTime

long getFirstAttemptStartTime(String vertexName,
                              int taskIndex)

Get the start time for the first attempt of the specified task

Parameters:

vertexName - the vertex to which the task belongs

taskIndex - the index of the task

Returns:

the start time for the first attempt of the task

getDagStartTime

long getDagStartTime()

Get the start time for the currently executing DAG

Returns:

time when the current dag started executing