Package org.elasticsearch.cluster.coordination

Class LeaderChecker

java.lang.Object

org.elasticsearch.cluster.coordination.LeaderChecker

public class LeaderChecker
extends java.lang.Object

The LeaderChecker is responsible for allowing followers to check that the currently elected leader is still connected and healthy. We are fairly lenient, possibly allowing multiple checks to fail before considering the leader to be faulty, to allow for the leader to temporarily stand down on occasion, e.g. if it needs to move to a higher term. On deciding that the leader has failed a follower will become a candidate and attempt to become a leader itself.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	LeaderChecker.LeaderCheckRequest

Field Summary

Fields

Modifier and Type	Field	Description
static java.lang.String	LEADER_CHECK_ACTION_NAME
static Setting<TimeValue>	LEADER_CHECK_INTERVAL_SETTING
static Setting<java.lang.Integer>	LEADER_CHECK_RETRY_COUNT_SETTING
static Setting<TimeValue>	LEADER_CHECK_TIMEOUT_SETTING

Constructor Summary

Constructors

Constructor	Description
LeaderChecker(Settings settings, TransportService transportService, java.lang.Runnable onLeaderFailure)

Method Summary

Modifier and Type	Method	Description
DiscoveryNode	leader()
void	setCurrentNodes(DiscoveryNodes discoveryNodes)	Update the "known" discovery nodes.
void	updateLeader(DiscoveryNode leader)	Starts and / or stops a leader checker for the given leader.

Methods inherited from class java.lang.Object

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

Field Detail

LEADER_CHECK_ACTION_NAME

public static final java.lang.String LEADER_CHECK_ACTION_NAME

See Also:

Constant Field Values

LEADER_CHECK_INTERVAL_SETTING

public static final Setting<TimeValue> LEADER_CHECK_INTERVAL_SETTING

LEADER_CHECK_TIMEOUT_SETTING

public static final Setting<TimeValue> LEADER_CHECK_TIMEOUT_SETTING

LEADER_CHECK_RETRY_COUNT_SETTING

public static final Setting<java.lang.Integer> LEADER_CHECK_RETRY_COUNT_SETTING

Constructor Detail

LeaderChecker

public LeaderChecker(Settings settings,
                     TransportService transportService,
                     java.lang.Runnable onLeaderFailure)

Method Detail

leader

public DiscoveryNode leader()

updateLeader

public void updateLeader(@Nullable
                         DiscoveryNode leader)

Starts and / or stops a leader checker for the given leader. Should only be called after successfully joining this leader.

Parameters:

leader - the node to be checked as leader, or null if checks should be disabled

setCurrentNodes

public void setCurrentNodes(DiscoveryNodes discoveryNodes)

Update the "known" discovery nodes. Should be called on the leader before a new cluster state is published to reflect the new publication targets, and also called if a leader becomes a non-leader. TODO if heartbeats can make nodes become followers then this needs to be called before a heartbeat is sent to a new node too.

isLocalNodeElectedMaster() should reflect whether this node is a leader, and nodeExists() should indicate whether nodes are known publication targets or not.