Module org.jgroups.raft

Package org.jgroups.raft.util

Class RequestTable<T>

java.lang.Object

org.jgroups.raft.util.RequestTable<T>

public class RequestTable<T>
extends Object

Keeps track of AppendEntriesRequest messages and responses.

Each AppendEntry request is keyed by the index at which it was inserted at the leader. The values (RequestEntry) contain the responses from followers. When a response is added, and the majority has been reached, add() returns true and the key/value pair will be removed. (subsequent responses will be ignored). On a majority, the commit index is advanced.

Only created on leader

Since:

0.1

Author:

Bela Ban

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	RequestTable.Entry<T>

Field Summary

Fields

Modifier and Type	Field	Description
protected ArrayRingBuffer<RequestTable.Entry<T>>	requests

Constructor Summary

Constructors

Constructor	Description
RequestTable()

Method Summary

Modifier and Type	Method	Description
boolean	add(long index, T sender, Supplier<Integer> majority)	Adds a response to the response set.
void	create(long index, T vote, CompletableFuture<byte[]> future, Supplier<Integer> majority)
void	create(long index, T vote, CompletableFuture<byte[]> future, Supplier<Integer> majority, Options opts)
void	destroy(Throwable t)	Completes all uncommitted requests with the provided exception.
boolean	isCommitted(long index)	Whether or not the entry at index is committed
void	notifyAndRemove(long index, byte[] response)	Notifies the CompletableFuture and then removes the entry for index
RequestTable.Entry<T>	remove(long index)
int	size()	number of requests being processed
String	toString()

Methods inherited from class java.lang.Object

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

Field Details

requests

protected ArrayRingBuffer<RequestTable.Entry<T>> requests

Constructor Details

RequestTable

public RequestTable()

Method Details

create

public void create(long index,
 T vote,
 CompletableFuture<byte[]> future,
 Supplier<Integer> majority)

create

public void create(long index,
 T vote,
 CompletableFuture<byte[]> future,
 Supplier<Integer> majority,
 Options opts)

destroy

public void destroy(Throwable t)

Completes all uncommitted requests with the provided exception.

This method should be invoked before setting the instance to null when the leader steps down. This provides a more responsive completion of requests to the users, instead of having requests time out. Internal operations, such as membership changes, do not have a timeout associated, which would hang and any subsequent changes would not complete.

Parameters:

t - : Throwable to complete the requests exceptionally.

add

public boolean add(long index,
 T sender,
 Supplier<Integer> majority)

Adds a response to the response set. If the majority has been reached, returns true

Returns:

True if a majority has been reached, false otherwise. Note that this is done exactly once

isCommitted

public boolean isCommitted(long index)

Whether or not the entry at index is committed

size

public int size()

number of requests being processed

remove

public RequestTable.Entry<T> remove(long index)

notifyAndRemove

public void notifyAndRemove(long index,
 byte[] response)

Notifies the CompletableFuture and then removes the entry for index

toString

public String toString()

Overrides:

toString in class Object