io.druid.indexing.common.task

Interface Task

All Known Implementing Classes:

AbstractFixedIntervalTask, AbstractTask, AppendTask, ArchiveTask, ConvertSegmentBackwardsCompatibleTask, ConvertSegmentBackwardsCompatibleTask.SubTask, ConvertSegmentTask, ConvertSegmentTask.SubTask, HadoopConverterTask, HadoopConverterTask.ConverterSubTask, HadoopIndexTask, HadoopTask, IndexTask, KillTask, MergeTask, MergeTaskBase, MoveTask, NoopTask, RealtimeIndexTask, RestoreTask

public interface Task

Represents a task that can run on a worker. The general contracts surrounding Tasks are:

• Tasks must operate on a single datasource.
• Tasks should be immutable, since the task ID is used as a proxy for the task in many locations.
• Task IDs must be unique. This can be done by naming them using UUIDs or the current timestamp.
• Tasks are each part of a "task group", which is a set of tasks that can share interval locks. These are useful for producing sharded segments.
• Tasks do not need to explicitly release locks; they are released upon task completion. Tasks may choose to release locks early if they desire.

Method Summary

Modifier and Type	Method and Description
boolean	canRestore() Returns whether or not this task can restore its progress from its on-disk working directory.
String	getClasspathPrefix() Returns an extra classpath that should be prepended to the default classpath when running this task.
Map<String,Object>	getContext()
Object	getContextValue(String key)
String	getDataSource() Returns the datasource this task operates on.
String	getGroupId() Returns group ID of this task.
String	getId() Returns ID of this task.
String	getNodeType() Get the nodeType for if/when this task publishes on zookeeper.
<T> QueryRunner<T>	getQueryRunner(Query<T> query) Returns query runners for this task.
TaskResource	getTaskResource() Returns a TaskResource for this task.
String	getType() Returns a descriptive label for this task type.
boolean	isReady(TaskActionClient taskActionClient) Execute preflight actions for a task.
TaskStatus	run(TaskToolbox toolbox) Execute a task.
void	stopGracefully() Asks a task to arrange for its "run" method to exit promptly.

Method Detail

getId

String getId()

Returns ID of this task. Must be unique across all tasks ever created.

Returns:

task ID

getGroupId

String getGroupId()

Returns group ID of this task. Tasks with the same group ID can share locks. If tasks do not need to share locks, a common convention is to set group ID equal to task ID.

Returns:

task group ID

getTaskResource

TaskResource getTaskResource()

Returns a TaskResource for this task. Task resources define specific worker requirements a task may require.

Returns:

TaskResource for this task

getType

String getType()

Returns a descriptive label for this task type. Used for metrics emission and logging.

Returns:

task type label

getNodeType

String getNodeType()

Get the nodeType for if/when this task publishes on zookeeper.

Returns:

the nodeType to use when publishing the server to zookeeper. null if the task doesn't expect to publish to zookeeper.

getDataSource

String getDataSource()

Returns the datasource this task operates on. Each task can operate on only one datasource.

getQueryRunner

<T> QueryRunner<T> getQueryRunner(Query<T> query)

Returns query runners for this task. If this task is not meant to answer queries over its datasource, this method should return null.

Type Parameters:

T - query result type

Returns:

query runners for this task

getClasspathPrefix

String getClasspathPrefix()

Returns an extra classpath that should be prepended to the default classpath when running this task. If no extra classpath should be prepended, this should return null or the empty string.

isReady

boolean isReady(TaskActionClient taskActionClient)
         throws Exception

Execute preflight actions for a task. This can be used to acquire locks, check preconditions, and so on. The actions must be idempotent, since this method may be executed multiple times. This typically runs on the coordinator. If this method throws an exception, the task should be considered a failure.

This method must be idempotent, as it may be run multiple times per task.

Parameters:

taskActionClient - action client for this task (not the full toolbox)

Returns:

true if ready, false if not ready yet

Throws:

Exception - if the task should be considered a failure

canRestore

boolean canRestore()

Returns whether or not this task can restore its progress from its on-disk working directory. Restorable tasks may be started with a non-empty working directory. Tasks that exit uncleanly may still have a chance to attempt restores, meaning that restorable tasks should be able to deal with potentially partially written on-disk state.

stopGracefully

void stopGracefully()

Asks a task to arrange for its "run" method to exit promptly. This method will only be called if canRestore() returns true. Tasks that take too long to stop gracefully will be terminated with extreme prejudice.

run

TaskStatus run(TaskToolbox toolbox)
        throws Exception

Execute a task. This typically runs on a worker as determined by a TaskRunner, and will be run while holding a lock on our dataSource and implicit lock interval (if any). If this method throws an exception, the task should be considered a failure.

Parameters:

toolbox - Toolbox for this task

Returns:

Some kind of finished status (isRunnable must be false).

Throws:

Exception - if this task failed

getContext

Map<String,Object> getContext()

getContextValue

Object getContextValue(String key)