org.apache.flink.runtime.executiongraph.failover

Class FailureHandlingResult

java.lang.Object

org.apache.flink.runtime.executiongraph.failover.FailureHandlingResult

public class FailureHandlingResult
extends Object

Result containing the tasks to restart upon a task failure. Also contains the reason of the failure and the vertices to restart if the failure is recoverable (in contrast to non-recoverable failure type or restarting suppressed by restart strategy).

Method Summary

Modifier and Type	Method and Description
boolean	canRestart() Returns whether the restarting can be conducted.
Throwable	getError() Returns reason why the restarting cannot be conducted.
Optional<Execution>	getFailedExecution() Returns an Optional with the Execution causing this failure or an empty Optional if it's a global failure.
CompletableFuture<Map<String,String>>	getFailureLabels() Returns the labels future associated with the failure.
long	getRestartDelayMS() Returns the delay before the restarting.
long	getTimestamp() Returns the time of the failure.
Set<ExecutionVertexID>	getVerticesToRestart() Returns the tasks to restart.
boolean	isGlobalFailure() Checks if this failure was a global failure, i.e., coming from a "safety net" failover that involved all tasks and should reset also components like the coordinators.
boolean	isRootCause()
static FailureHandlingResult	restartable(Execution failedExecution, Throwable cause, long timestamp, CompletableFuture<Map<String,String>> failureLabels, Set<ExecutionVertexID> verticesToRestart, long restartDelayMS, boolean globalFailure, boolean isRootCause) Creates a result of a set of tasks to restart to recover from the failure.
static FailureHandlingResult	unrecoverable(Execution failedExecution, Throwable error, long timestamp, CompletableFuture<Map<String,String>> failureLabels, boolean globalFailure, boolean isRootCause) Creates a result that the failure is not recoverable and no restarting should be conducted.

Methods inherited from class java.lang.Object

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

Method Detail

getVerticesToRestart

public Set<ExecutionVertexID> getVerticesToRestart()

Returns the tasks to restart.

Returns:

the tasks to restart

getRestartDelayMS

public long getRestartDelayMS()

Returns the delay before the restarting.

Returns:

the delay before the restarting

getFailedExecution

public Optional<Execution> getFailedExecution()

Returns an Optional with the Execution causing this failure or an empty Optional if it's a global failure.

Returns:

The Optional with the failed Execution or an empty Optional if it's a global failure.

getError

@Nullable
public Throwable getError()

Returns reason why the restarting cannot be conducted.

Returns:

reason why the restarting cannot be conducted

getFailureLabels

public CompletableFuture<Map<String,String>> getFailureLabels()

Returns the labels future associated with the failure.

Returns:

the CompletableFuture Map of String labels

getTimestamp

public long getTimestamp()

Returns the time of the failure.

Returns:

The timestamp.

canRestart

public boolean canRestart()

Returns whether the restarting can be conducted.

Returns:

whether the restarting can be conducted

isGlobalFailure

public boolean isGlobalFailure()

Checks if this failure was a global failure, i.e., coming from a "safety net" failover that involved all tasks and should reset also components like the coordinators.

isRootCause

public boolean isRootCause()

Returns:

True means that the current failure is a new attempt, false means that there has been a failure before and has not been tried yet, and the current failure will be merged into the previous attempt, and these merged exceptions will be considered as the concurrent exceptions.

restartable

public static FailureHandlingResult restartable(@Nullable
                                                Execution failedExecution,
                                                @Nullable
                                                Throwable cause,
                                                long timestamp,
                                                CompletableFuture<Map<String,String>> failureLabels,
                                                Set<ExecutionVertexID> verticesToRestart,
                                                long restartDelayMS,
                                                boolean globalFailure,
                                                boolean isRootCause)

Creates a result of a set of tasks to restart to recover from the failure.

The result can be flagged to be from a global failure triggered by the scheduler, rather than from the failure of an individual task.

Parameters:

failedExecution - the Execution that the failure is originating from. Passing null as a value indicates that the failure was issued by Flink itself.

cause - The reason of the failure.

timestamp - The time of the failure.

failureLabels - Map of labels characterizing the failure produced by the FailureEnrichers.

verticesToRestart - containing task vertices to restart to recover from the failure. null indicates that the failure is not restartable.

restartDelayMS - indicate a delay before conducting the restart

Returns:

result of a set of tasks to restart to recover from the failure

unrecoverable

public static FailureHandlingResult unrecoverable(@Nullable
                                                  Execution failedExecution,
                                                  @Nonnull
                                                  Throwable error,
                                                  long timestamp,
                                                  CompletableFuture<Map<String,String>> failureLabels,
                                                  boolean globalFailure,
                                                  boolean isRootCause)

Creates a result that the failure is not recoverable and no restarting should be conducted.

The result can be flagged to be from a global failure triggered by the scheduler, rather than from the failure of an individual task.

Parameters:

failedExecution - the Execution that the failure is originating from. Passing null as a value indicates that the failure was issued by Flink itself.

error - reason why the failure is not recoverable

timestamp - The time of the failure.

failureLabels - Map of labels characterizing the failure produced by the FailureEnrichers.

Returns:

result indicating the failure is not recoverable