public interface TaskCoordinator
StreamTask implementations
 to allow the user code to request actions from the Samza framework, including committing the current checkpoints
 to configured org.apache.samza.checkpoint.CheckpointManager or shutting down the task or all tasks within
 a container.
 This interface may evolve over time.
| Modifier and Type | Interface and Description | 
|---|---|
| static class  | TaskCoordinator.RequestScopeA task can make requests to the Samza framework while processing messages, such as
  commit(RequestScope)andshutdown(RequestScope). | 
| Modifier and Type | Method and Description | 
|---|---|
| void | commit(TaskCoordinator.RequestScope scope)Requests that Samza should write out a checkpoint, from which a task can restart
 after failure. | 
| void | shutdown(TaskCoordinator.RequestScope scope)Requests that the container should be shut down. | 
void commit(TaskCoordinator.RequestScope scope)
If CURRENT_TASK is given, a checkpoint is only written for the
 current task. If ALL_TASKS_IN_CONTAINER is given, a checkpoint is
 written for all tasks in the current container.
 
Note that if you also have also configured your job to commit in regular
 intervals (using the task.commit.ms property), those time-based
 commits are not affected by calling this method. Any commits you request explicitly
 are in addition to timer-based commits. You can set task.commit.ms=-1
 if you don't want commits to happen automatically.
scope - Which tasks are being asked to commit.void shutdown(TaskCoordinator.RequestScope scope)
If CURRENT_TASK is given, that indicates a willingness of the current
 task to shut down. All tasks in the container (including the one that requested
 shutdown) will continue processing messages. Only when every task in the container
 has called shutdown(CURRENT_TASK), the container is shut down. Once a
 task has called shutdown(CURRENT_TASK), it cannot change its mind
 (i.e. it cannot revoke its willingness to shut down).
 
If ALL_TASKS_IN_CONTAINER is given, the container will shut down
 immediately after it has finished processing the current message. Any buffers of
 pending writes are flushed, but no further messages will be processed in this
 container.
scope - The approach we should use for shutting down the container.