org.apache.flink.runtime.executiongraph.failover

Class RestartPipelinedRegionFailoverStrategy

java.lang.Object

org.apache.flink.runtime.executiongraph.failover.RestartPipelinedRegionFailoverStrategy

All Implemented Interfaces:

FailoverStrategy

public class RestartPipelinedRegionFailoverStrategy
extends Object
implements FailoverStrategy

A failover strategy that proposes to restart involved regions when a vertex fails. A region is defined by this strategy as tasks that communicate via pipelined data exchange.

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	RestartPipelinedRegionFailoverStrategy.Factory The factory to instantiate RestartPipelinedRegionFailoverStrategy.

Constructor Summary

Constructors

Constructor and Description
RestartPipelinedRegionFailoverStrategy(SchedulingTopology topology) Creates a new failover strategy to restart pipelined regions that works on the given topology.
RestartPipelinedRegionFailoverStrategy(SchedulingTopology topology, ResultPartitionAvailabilityChecker resultPartitionAvailabilityChecker) Creates a new failover strategy to restart pipelined regions that works on the given topology.

Method Summary

Modifier and Type	Method and Description
SchedulingPipelinedRegion	getFailoverRegion(ExecutionVertexID vertexID) Returns the failover region that contains the given execution vertex.
Set<ExecutionVertexID>	getTasksNeedingRestart(ExecutionVertexID executionVertexId, Throwable cause) Returns a set of IDs corresponding to the set of vertices that should be restarted.

Methods inherited from class java.lang.Object

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

Constructor Detail

RestartPipelinedRegionFailoverStrategy

@VisibleForTesting
public RestartPipelinedRegionFailoverStrategy(SchedulingTopology topology)

Creates a new failover strategy to restart pipelined regions that works on the given topology. The result partitions are always considered to be available if no data consumption error happens.

Parameters:

topology - containing info about all the vertices and result partitions

RestartPipelinedRegionFailoverStrategy

public RestartPipelinedRegionFailoverStrategy(SchedulingTopology topology,
                                              ResultPartitionAvailabilityChecker resultPartitionAvailabilityChecker)

Creates a new failover strategy to restart pipelined regions that works on the given topology.

Parameters:

topology - containing info about all the vertices and result partitions

resultPartitionAvailabilityChecker - helps to query result partition availability

Method Detail

getTasksNeedingRestart

public Set<ExecutionVertexID> getTasksNeedingRestart(ExecutionVertexID executionVertexId,
                                                     Throwable cause)

Returns a set of IDs corresponding to the set of vertices that should be restarted. In this strategy, all task vertices in 'involved' regions are proposed to be restarted. The 'involved' regions are calculated with rules below: 1. The region containing the failed task is always involved 2. If an input result partition of an involved region is not available, i.e. Missing or Corrupted, the region containing the partition producer task is involved 3. If a region is involved, all of its consumer regions are involved

Specified by:

getTasksNeedingRestart in interface FailoverStrategy

Parameters:

executionVertexId - ID of the failed task

cause - cause of the failure

Returns:

set of IDs of vertices to restart

getFailoverRegion

@VisibleForTesting
public SchedulingPipelinedRegion getFailoverRegion(ExecutionVertexID vertexID)

Returns the failover region that contains the given execution vertex.

Returns:

the failover region that contains the given execution vertex