raft

package
v0.0.0-...-678c32f Latest Latest
Warning

This package is not in the latest version of its module.

Go to latest
Published: Apr 11, 2020 License: Apache-2.0, MPL-2.0 Imports: 26 Imported by: 0

README

raft Build Status

raft is a Go library that manages a replicated log and can be used with an FSM to manage replicated state machines. It is a library for providing consensus.

The use cases for such a library are far-reaching as replicated state machines are a key component of many distributed systems. They enable building Consistent, Partition Tolerant (CP) systems, with limited fault tolerance as well.

Building

If you wish to build raft you'll need Go version 1.2+ installed.

Please check your installation with:

go version

Documentation

For complete documentation, see the associated Godoc.

To prevent complications with cgo, the primary backend MDBStore is in a separate repository, called raft-mdb. That is the recommended implementation for the LogStore and StableStore.

A pure Go backend using BoltDB is also available called raft-boltdb. It can also be used as a LogStore and StableStore.

Protocol

raft is based on "Raft: In Search of an Understandable Consensus Algorithm"

A high level overview of the Raft protocol is described below, but for details please read the full Raft paper followed by the raft source. Any questions about the raft protocol should be sent to the raft-dev mailing list.

Protocol Description

Raft nodes are always in one of three states: follower, candidate or leader. All nodes initially start out as a follower. In this state, nodes can accept log entries from a leader and cast votes. If no entries are received for some time, nodes self-promote to the candidate state. In the candidate state nodes request votes from their peers. If a candidate receives a quorum of votes, then it is promoted to a leader. The leader must accept new log entries and replicate to all the other followers. In addition, if stale reads are not acceptable, all queries must also be performed on the leader.

Once a cluster has a leader, it is able to accept new log entries. A client can request that a leader append a new log entry, which is an opaque binary blob to Raft. The leader then writes the entry to durable storage and attempts to replicate to a quorum of followers. Once the log entry is considered committed, it can be applied to a finite state machine. The finite state machine is application specific, and is implemented using an interface.

An obvious question relates to the unbounded nature of a replicated log. Raft provides a mechanism by which the current state is snapshotted, and the log is compacted. Because of the FSM abstraction, restoring the state of the FSM must result in the same state as a replay of old logs. This allows Raft to capture the FSM state at a point in time, and then remove all the logs that were used to reach that state. This is performed automatically without user intervention, and prevents unbounded disk usage as well as minimizing time spent replaying logs.

Lastly, there is the issue of updating the peer set when new servers are joining or existing servers are leaving. As long as a quorum of nodes is available, this is not an issue as Raft provides mechanisms to dynamically update the peer set. If a quorum of nodes is unavailable, then this becomes a very challenging issue. For example, suppose there are only 2 peers, A and B. The quorum size is also 2, meaning both nodes must agree to commit a log entry. If either A or B fails, it is now impossible to reach quorum. This means the cluster is unable to add, or remove a node, or commit any additional log entries. This results in unavailability. At this point, manual intervention would be required to remove either A or B, and to restart the remaining node in bootstrap mode.

A Raft cluster of 3 nodes can tolerate a single node failure, while a cluster of 5 can tolerate 2 node failures. The recommended configuration is to either run 3 or 5 raft servers. This maximizes availability without greatly sacrificing performance.

In terms of performance, Raft is comparable to Paxos. Assuming stable leadership, committing a log entry requires a single round trip to half of the cluster. Thus performance is bound by disk I/O and network latency.

Documentation

Index

Constants

View Source
const (

	// DefaultTimeoutScale is the default TimeoutScale in a NetworkTransport.
	DefaultTimeoutScale = 256 * 1024 // 256KB

)

Variables

View Source
var (
	// ErrTransportShutdown is returned when operations on a transport are
	// invoked after it's been terminated.
	ErrTransportShutdown = errors.New("transport shutdown")

	// ErrPipelineShutdown is returned when the pipeline is closed.
	ErrPipelineShutdown = errors.New("append pipeline closed")
)
View Source
var (

	// ErrLeader is returned when an operation can't be completed on a
	// leader node.
	ErrLeader = errors.New("node is the leader")

	// ErrNotLeader is returned when an operation can't be completed on a
	// follower or candidate node.
	ErrNotLeader = errors.New("node is not the leader")

	// ErrLeadershipLost is returned when a leader fails to commit a log entry
	// because it's been deposed in the process.
	ErrLeadershipLost = errors.New("leadership lost while committing log")

	// ErrRaftShutdown is returned when operations are requested against an
	// inactive Raft.
	ErrRaftShutdown = errors.New("raft is already shutdown")

	// ErrEnqueueTimeout is returned when a command fails due to a timeout.
	ErrEnqueueTimeout = errors.New("timed out enqueuing operation")

	// ErrKnownPeer is returned when trying to add a peer to the configuration
	// that already exists.
	ErrKnownPeer = errors.New("peer already known")

	// ErrUnknownPeer is returned when trying to remove a peer from the
	// configuration that doesn't exist.
	ErrUnknownPeer = errors.New("peer is unknown")

	// ErrNothingNewToSnapshot is returned when trying to create a snapshot
	// but there's nothing new commited to the FSM since we started.
	ErrNothingNewToSnapshot = errors.New("Nothing new to snapshot")
)
View Source
var (
	// ErrLogNotFound indicates a given log entry is not available.
	ErrLogNotFound = errors.New("log not found")

	// ErrPipelineReplicationNotSupported can be returned by the transport to
	// signal that pipeline replication is not supported in general, and that
	// no error message should be produced.
	ErrPipelineReplicationNotSupported = errors.New("pipeline replication not supported")
)

Functions

func AddUniquePeer

func AddUniquePeer(peers []string, peer string) []string

AddUniquePeer is used to add a peer to a list of existing peers only if it is not already contained.

func DecodePeer

func DecodePeer(buf []byte) string

DecodePeer implements the Transport interface.

func EncodePeer

func EncodePeer(p string) []byte

func ExcludePeer

func ExcludePeer(peers []string, peer string) []string

ExcludePeer is used to exclude a single peer from a list of peers.

func NewInmemAddr

func NewInmemAddr() string

NewInmemAddr returns a new in-memory addr with a randomly generate UUID as the ID.

func PeerContained

func PeerContained(peers []string, peer string) bool

PeerContained checks if a given peer is contained in a list.

func ValidateConfig

func ValidateConfig(config *Config) error

ValidateConfig is used to validate a sane configuration

Types

type AppendEntriesRequest

type AppendEntriesRequest struct {
	// Provide the current term and leader
	Term   uint64
	Leader []byte

	// Provide the previous entries for integrity checking
	PrevLogEntry uint64
	PrevLogTerm  uint64

	// New entries to commit
	Entries []*Log

	// Commit index on the leader
	LeaderCommitIndex uint64
}

AppendEntriesRequest is the command used to append entries to the replicated log.

type AppendEntriesResponse

type AppendEntriesResponse struct {
	// Newer term if leader is out of date
	Term uint64

	// Last Log is a hint to help accelerate rebuilding slow nodes
	LastLog uint64

	// We may not succeed if we have a conflicting entry
	Success bool

	// There are scenarios where this request didn't succeed
	// but there's no need to wait/back-off the next attempt.
	NoRetryBackoff bool
}

AppendEntriesResponse is the response returned from an AppendEntriesRequest.

type AppendFuture

type AppendFuture interface {
	Future

	// Start returns the time that the append request was started.
	// It is always OK to call this method.
	Start() time.Time

	// Request holds the parameters of the AppendEntries call.
	// It is always OK to call this method.
	Request() *AppendEntriesRequest

	// Response holds the results of the AppendEntries call.
	// This method must only be called after the Error
	// method returns, and will only be valid on success.
	Response() *AppendEntriesResponse
}

AppendFuture is used to return information about a pipelined AppendEntries request.

type AppendPipeline

type AppendPipeline interface {
	// AppendEntries is used to add another request to the pipeline.
	// The send may block which is an effective form of back-pressure.
	AppendEntries(args *AppendEntriesRequest, resp *AppendEntriesResponse) (AppendFuture, error)

	// Consumer returns a channel that can be used to consume
	// response futures when they are ready.
	Consumer() <-chan AppendFuture

	// Close closes the pipeline and cancels all inflight RPCs
	Close() error
}

AppendPipeline is used for pipelining AppendEntries requests. It is used to increase the replication throughput by masking latency and better utilizing bandwidth.

type ApplyFuture

type ApplyFuture interface {
	Future

	// Response returns the FSM response as returned
	// by the FSM.Apply method. This must not be called
	// until after the Error method has returned.
	Response() interface{}

	// Index holds the index of the newly applied log entry.
	// This must not be called
	// until after the Error method has returned.
	Index() uint64
}

ApplyFuture is used for Apply() and can returns the FSM response.

type Config

type Config struct {
	// HeartbeatTimeout specifies the time in follower state without
	// a leader before we attempt an election.
	HeartbeatTimeout time.Duration

	// ElectionTimeout specifies the time in candidate state without
	// a leader before we attempt an election.
	ElectionTimeout time.Duration

	// CommitTimeout controls the time without an Apply() operation
	// before we heartbeat to ensure a timely commit. Due to random
	// staggering, may be delayed as much as 2x this value.
	CommitTimeout time.Duration

	// MaxAppendEntries controls the maximum number of append entries
	// to send at once. We want to strike a balance between efficiency
	// and avoiding waste if the follower is going to reject because of
	// an inconsistent log.
	MaxAppendEntries int

	// If we are a member of a cluster, and RemovePeer is invoked for the
	// local node, then we forget all peers and transition into the follower state.
	// If ShutdownOnRemove is is set, we additional shutdown Raft. Otherwise,
	// we can become a leader of a cluster containing only this node.
	ShutdownOnRemove bool

	// DisableBootstrapAfterElect is used to turn off EnableSingleNode
	// after the node is elected. This is used to prevent self-election
	// if the node is removed from the Raft cluster via RemovePeer. Setting
	// it to false will keep the bootstrap mode, allowing the node to self-elect
	// and potentially bootstrap a separate cluster.
	DisableBootstrapAfterElect bool

	// TrailingLogs controls how many logs we leave after a snapshot. This is
	// used so that we can quickly replay logs on a follower instead of being
	// forced to send an entire snapshot.
	TrailingLogs uint64

	// SnapshotInterval controls how often we check if we should perform a snapshot.
	// We randomly stagger between this value and 2x this value to avoid the entire
	// cluster from performing a snapshot at once.
	SnapshotInterval time.Duration

	// SnapshotThreshold controls how many outstanding logs there must be before
	// we perform a snapshot. This is to prevent excessive snapshots when we can
	// just replay a small set of logs.
	SnapshotThreshold uint64

	// EnableSingleNode allows for a single node mode of operation. This
	// is false by default, which prevents a lone node from electing itself.
	// leader.
	EnableSingleNode bool

	// LeaderLeaseTimeout is used to control how long the "lease" lasts
	// for being the leader without being able to contact a quorum
	// of nodes. If we reach this interval without contact, we will
	// step down as leader.
	LeaderLeaseTimeout time.Duration

	// StartAsLeader forces Raft to start in the leader state. This should
	// never be used except for testing purposes, as it can cause a split-brain.
	StartAsLeader bool

	// NotifyCh is used to provide a channel that will be notified of leadership
	// changes. Raft will block writing to this channel, so it should either be
	// buffered or aggressively consumed.
	NotifyCh chan<- bool

	// LogOutput is used as a sink for logs, unless Logger is specified.
	// Defaults to os.Stderr.
	LogOutput io.Writer

	// Logger is a user-provided logger. If nil, a logger writing to LogOutput
	// is used.
	Logger *log.Logger
}

Config provides any necessary configuration to the Raft server

func DefaultConfig

func DefaultConfig() *Config

DefaultConfig returns a Config with usable defaults.

type DiscardSnapshotSink

type DiscardSnapshotSink struct{}

func (*DiscardSnapshotSink) Cancel

func (d *DiscardSnapshotSink) Cancel() error

func (*DiscardSnapshotSink) Close

func (d *DiscardSnapshotSink) Close() error

func (*DiscardSnapshotSink) ID

func (d *DiscardSnapshotSink) ID() string

func (*DiscardSnapshotSink) Write

func (d *DiscardSnapshotSink) Write(b []byte) (int, error)

type DiscardSnapshotStore

type DiscardSnapshotStore struct{}

DiscardSnapshotStore is used to successfully snapshot while always discarding the snapshot. This is useful for when the log should be truncated but no snapshot should be retained. This should never be used for production use, and is only suitable for testing.

func NewDiscardSnapshotStore

func NewDiscardSnapshotStore() *DiscardSnapshotStore

NewDiscardSnapshotStore is used to create a new DiscardSnapshotStore.

func (*DiscardSnapshotStore) Create

func (d *DiscardSnapshotStore) Create(index, term uint64, peers []byte) (SnapshotSink, error)

func (*DiscardSnapshotStore) List

func (d *DiscardSnapshotStore) List() ([]*SnapshotMeta, error)

func (*DiscardSnapshotStore) Open

type FSM

type FSM interface {
	// Apply log is invoked once a log entry is committed.
	// It returns a value which will be made available in the
	// ApplyFuture returned by Raft.Apply method if that
	// method was called on the same Raft node as the FSM.
	Apply(*Log) interface{}

	// Snapshot is used to support log compaction. This call should
	// return an FSMSnapshot which can be used to save a point-in-time
	// snapshot of the FSM. Apply and Snapshot are not called in multiple
	// threads, but Apply will be called concurrently with Persist. This means
	// the FSM should be implemented in a fashion that allows for concurrent
	// updates while a snapshot is happening.
	Snapshot() (FSMSnapshot, error)

	// Restore is used to restore an FSM from a snapshot. It is not called
	// concurrently with any other command. The FSM must discard all previous
	// state.
	Restore(io.ReadCloser) error
}

FSM provides an interface that can be implemented by clients to make use of the replicated log.

type FSMSnapshot

type FSMSnapshot interface {
	// Persist should dump all necessary state to the WriteCloser 'sink',
	// and call sink.Close() when finished or call sink.Cancel() on error.
	Persist(sink SnapshotSink) error

	// Release is invoked when we are finished with the snapshot.
	Release()
}

FSMSnapshot is returned by an FSM in response to a Snapshot It must be safe to invoke FSMSnapshot methods with concurrent calls to Apply.

type FileSnapshotSink

type FileSnapshotSink struct {
	// contains filtered or unexported fields
}

FileSnapshotSink implements SnapshotSink with a file.

func (*FileSnapshotSink) Cancel

func (s *FileSnapshotSink) Cancel() error

Cancel is used to indicate an unsuccessful end.

func (*FileSnapshotSink) Close

func (s *FileSnapshotSink) Close() error

Close is used to indicate a successful end.

func (*FileSnapshotSink) ID

func (s *FileSnapshotSink) ID() string

ID returns the ID of the snapshot, can be used with Open() after the snapshot is finalized.

func (*FileSnapshotSink) Write

func (s *FileSnapshotSink) Write(b []byte) (int, error)

Write is used to append to the state file. We write to the buffered IO object to reduce the amount of context switches.

type FileSnapshotStore

type FileSnapshotStore struct {
	// contains filtered or unexported fields
}

FileSnapshotStore implements the SnapshotStore interface and allows snapshots to be made on the local disk.

func NewFileSnapshotStore

func NewFileSnapshotStore(base string, retain int, logOutput io.Writer) (*FileSnapshotStore, error)

NewFileSnapshotStore creates a new FileSnapshotStore based on a base directory. The `retain` parameter controls how many snapshots are retained. Must be at least 1.

func NewFileSnapshotStoreWithLogger

func NewFileSnapshotStoreWithLogger(base string, retain int, logger *log.Logger) (*FileSnapshotStore, error)

NewFileSnapshotStoreWithLogger creates a new FileSnapshotStore based on a base directory. The `retain` parameter controls how many snapshots are retained. Must be at least 1.

func (*FileSnapshotStore) Create

func (f *FileSnapshotStore) Create(index, term uint64, peers []byte) (SnapshotSink, error)

Create is used to start a new snapshot

func (*FileSnapshotStore) List

func (f *FileSnapshotStore) List() ([]*SnapshotMeta, error)

List returns available snapshots in the store.

func (*FileSnapshotStore) Open

Open takes a snapshot ID and returns a ReadCloser for that snapshot.

func (*FileSnapshotStore) ReapSnapshots

func (f *FileSnapshotStore) ReapSnapshots() error

ReapSnapshots reaps any snapshots beyond the retain count.

type FilterFn

type FilterFn func(o *Observation) bool

FilterFn is a function that can be registered in order to filter observations. The function reports whether the observation should be included - if it returns false, the observation will be filtered out.

type Future

type Future interface {
	// Error blocks until the future arrives and then
	// returns the error status of the future.
	// This may be called any number of times - all
	// calls will return the same value.
	// Note that it is not OK to call this method
	// twice concurrently on the same Future instance.
	Error() error
}

Future is used to represent an action that may occur in the future.

type InmemStore

type InmemStore struct {
	// contains filtered or unexported fields
}

InmemStore implements the LogStore and StableStore interface. It should NOT EVER be used for production. It is used only for unit tests. Use the MDBStore implementation instead.

func NewInmemStore

func NewInmemStore() *InmemStore

NewInmemStore returns a new in-memory backend. Do not ever use for production. Only for testing.

func (*InmemStore) DeleteRange

func (i *InmemStore) DeleteRange(min, max uint64) error

DeleteRange implements the LogStore interface.

func (*InmemStore) FirstIndex

func (i *InmemStore) FirstIndex() (uint64, error)

FirstIndex implements the LogStore interface.

func (*InmemStore) Get

func (i *InmemStore) Get(key []byte) ([]byte, error)

Get implements the StableStore interface.

func (*InmemStore) GetLog

func (i *InmemStore) GetLog(index uint64, log *Log) error

GetLog implements the LogStore interface.

func (*InmemStore) GetUint64

func (i *InmemStore) GetUint64(key []byte) (uint64, error)

GetUint64 implements the StableStore interface.

func (*InmemStore) LastIndex

func (i *InmemStore) LastIndex() (uint64, error)

LastIndex implements the LogStore interface.

func (*InmemStore) Set

func (i *InmemStore) Set(key []byte, val []byte) error

Set implements the StableStore interface.

func (*InmemStore) SetUint64

func (i *InmemStore) SetUint64(key []byte, val uint64) error

SetUint64 implements the StableStore interface.

func (*InmemStore) StoreLog

func (i *InmemStore) StoreLog(log *Log) error

StoreLog implements the LogStore interface.

func (*InmemStore) StoreLogs

func (i *InmemStore) StoreLogs(logs []*Log) error

StoreLogs implements the LogStore interface.

type InmemTransport

type InmemTransport struct {
	sync.RWMutex
	// contains filtered or unexported fields
}

InmemTransport Implements the Transport interface, to allow Raft to be tested in-memory without going over a network.

func NewInmemTransport

func NewInmemTransport(addr string) (string, *InmemTransport)

NewInmemTransport is used to initialize a new transport and generates a random local address if none is specified

func (*InmemTransport) AppendEntries

func (i *InmemTransport) AppendEntries(target string, args *AppendEntriesRequest, resp *AppendEntriesResponse) error

AppendEntries implements the Transport interface.

func (*InmemTransport) AppendEntriesPipeline

func (i *InmemTransport) AppendEntriesPipeline(target string) (AppendPipeline, error)

AppendEntriesPipeline returns an interface that can be used to pipeline AppendEntries requests.

func (*InmemTransport) Close

func (i *InmemTransport) Close() error

Close is used to permanently disable the transport

func (*InmemTransport) Connect

func (i *InmemTransport) Connect(peer string, t Transport)

Connect is used to connect this transport to another transport for a given peer name. This allows for local routing.

func (*InmemTransport) Consumer

func (i *InmemTransport) Consumer() <-chan RPC

Consumer implements the Transport interface.

func (*InmemTransport) DecodePeer

func (i *InmemTransport) DecodePeer(buf []byte) string

DecodePeer implements the Transport interface. It wraps the UUID in an InmemAddr.

func (*InmemTransport) Disconnect

func (i *InmemTransport) Disconnect(peer string)

Disconnect is used to remove the ability to route to a given peer.

func (*InmemTransport) DisconnectAll

func (i *InmemTransport) DisconnectAll()

DisconnectAll is used to remove all routes to peers.

func (*InmemTransport) EncodePeer

func (i *InmemTransport) EncodePeer(p string) []byte

EncodePeer implements the Transport interface. It uses the UUID as the address directly.

func (*InmemTransport) InstallSnapshot

func (i *InmemTransport) InstallSnapshot(target string, args *InstallSnapshotRequest, resp *InstallSnapshotResponse, data io.Reader) error

InstallSnapshot implements the Transport interface.

func (*InmemTransport) LocalAddr

func (i *InmemTransport) LocalAddr() string

LocalAddr implements the Transport interface.

func (*InmemTransport) RequestVote

func (i *InmemTransport) RequestVote(target string, args *RequestVoteRequest, resp *RequestVoteResponse) error

RequestVote implements the Transport interface.

func (*InmemTransport) SetHeartbeatHandler

func (i *InmemTransport) SetHeartbeatHandler(cb func(RPC))

SetHeartbeatHandler is used to set optional fast-path for heartbeats, not supported for this transport.

type InstallSnapshotRequest

type InstallSnapshotRequest struct {
	Term   uint64
	Leader []byte

	// These are the last index/term included in the snapshot
	LastLogIndex uint64
	LastLogTerm  uint64

	// Peer Set in the snapshot
	Peers []byte

	// Size of the snapshot
	Size int64
}

InstallSnapshotRequest is the command sent to a Raft peer to bootstrap its log (and state machine) from a snapshot on another peer.

type InstallSnapshotResponse

type InstallSnapshotResponse struct {
	Term    uint64
	Success bool
}

InstallSnapshotResponse is the response returned from an InstallSnapshotRequest.

type JSONPeers

type JSONPeers struct {
	// contains filtered or unexported fields
}

JSONPeers is used to provide peer persistence on disk in the form of a JSON file. This allows human operators to manipulate the file.

func NewJSONPeers

func NewJSONPeers(base string, trans Transport) *JSONPeers

NewJSONPeers creates a new JSONPeers store. Requires a transport to handle the serialization of network addresses.

func (*JSONPeers) Peers

func (j *JSONPeers) Peers() ([]string, error)

Peers implements the PeerStore interface.

func (*JSONPeers) SetPeers

func (j *JSONPeers) SetPeers(peers []string) error

SetPeers implements the PeerStore interface.

type LeaderObservation

type LeaderObservation struct {
	Leader string
}

LeaderObservation is used in Observation.Data when leadership changes.

type Log

type Log struct {
	// Index holds the index of the log entry.
	Index uint64

	// Term holds the election term of the log entry.
	Term uint64

	// Type holds the type of the log entry.
	Type LogType

	// Data holds the log entry's type-specific data.
	Data []byte
	// contains filtered or unexported fields
}

Log entries are replicated to all members of the Raft cluster and form the heart of the replicated state machine.

type LogCache

type LogCache struct {
	// contains filtered or unexported fields
}

LogCache wraps any LogStore implementation to provide an in-memory ring buffer. This is used to cache access to the recently written entries. For implementations that do not cache themselves, this can provide a substantial boost by avoiding disk I/O on recent entries.

func NewLogCache

func NewLogCache(capacity int, store LogStore) (*LogCache, error)

NewLogCache is used to create a new LogCache with the given capacity and backend store.

func (*LogCache) DeleteRange

func (c *LogCache) DeleteRange(min, max uint64) error

func (*LogCache) FirstIndex

func (c *LogCache) FirstIndex() (uint64, error)

func (*LogCache) GetLog

func (c *LogCache) GetLog(idx uint64, log *Log) error

func (*LogCache) LastIndex

func (c *LogCache) LastIndex() (uint64, error)

func (*LogCache) StoreLog

func (c *LogCache) StoreLog(log *Log) error

func (*LogCache) StoreLogs

func (c *LogCache) StoreLogs(logs []*Log) error

type LogStore

type LogStore interface {
	// FirstIndex returns the first index written. 0 for no entries.
	FirstIndex() (uint64, error)

	// LastIndex returns the last index written. 0 for no entries.
	LastIndex() (uint64, error)

	// GetLog gets a log entry at a given index.
	GetLog(index uint64, log *Log) error

	// StoreLog stores a log entry.
	StoreLog(log *Log) error

	// StoreLogs stores multiple log entries.
	StoreLogs(logs []*Log) error

	// DeleteRange deletes a range of log entries. The range is inclusive.
	DeleteRange(min, max uint64) error
}

LogStore is used to provide an interface for storing and retrieving logs in a durable fashion.

type LogType

type LogType uint8

LogType describes various types of log entries.

const (
	// LogCommand is applied to a user FSM.
	LogCommand LogType = iota

	// LogNoop is used to assert leadership.
	LogNoop

	// LogAddPeer is used to add a new peer.
	LogAddPeer

	// LogRemovePeer is used to remove an existing peer.
	LogRemovePeer

	// LogBarrier is used to ensure all preceding operations have been
	// applied to the FSM. It is similar to LogNoop, but instead of returning
	// once committed, it only returns once the FSM manager acks it. Otherwise
	// it is possible there are operations committed but not yet applied to
	// the FSM.
	LogBarrier
)

type LoopbackTransport

type LoopbackTransport interface {
	Transport // Embedded transport reference
	WithPeers // Embedded peer management
	WithClose // with a close routine
}

LoopbackTransport is an interface that provides a loopback transport suitable for testing e.g. InmemTransport. It's there so we don't have to rewrite tests.

type NetworkTransport

type NetworkTransport struct {
	TimeoutScale int
	// contains filtered or unexported fields
}

NetworkTransport provides a network based transport that can be used to communicate with Raft on remote machines. It requires an underlying stream layer to provide a stream abstraction, which can be simple TCP, TLS, etc.

This transport is very simple and lightweight. Each RPC request is framed by sending a byte that indicates the message type, followed by the MsgPack encoded request.

The response is an error string followed by the response object, both are encoded using MsgPack.

InstallSnapshot is special, in that after the RPC request we stream the entire state. That socket is not re-used as the connection state is not known if there is an error.

func NewNetworkTransport

func NewNetworkTransport(
	stream StreamLayer,
	maxPool int,
	timeout time.Duration,
	logOutput io.Writer,
) *NetworkTransport

NewNetworkTransport creates a new network transport with the given dialer and listener. The maxPool controls how many connections we will pool. The timeout is used to apply I/O deadlines. For InstallSnapshot, we multiply the timeout by (SnapshotSize / TimeoutScale).

func NewNetworkTransportWithLogger

func NewNetworkTransportWithLogger(
	stream StreamLayer,
	maxPool int,
	timeout time.Duration,
	logger *log.Logger,
) *NetworkTransport

NewNetworkTransportWithLogger creates a new network transport with the given dialer and listener. The maxPool controls how many connections we will pool. The timeout is used to apply I/O deadlines. For InstallSnapshot, we multiply the timeout by (SnapshotSize / TimeoutScale).

func NewTCPTransport

func NewTCPTransport(
	bindAddr string,
	advertise net.Addr,
	maxPool int,
	timeout time.Duration,
	logOutput io.Writer, listener net.Listener,
) (*NetworkTransport, error)

NewTCPTransport returns a NetworkTransport that is built on top of a TCP streaming transport layer.

func NewTCPTransportWithLogger

func NewTCPTransportWithLogger(
	bindAddr string,
	advertise net.Addr,
	maxPool int,
	timeout time.Duration,
	logger *log.Logger,
	listener net.Listener,
) (*NetworkTransport, error)

NewTCPTransportWithLogger returns a NetworkTransport that is built on top of a TCP streaming transport layer, with log output going to the supplied Logger

func (*NetworkTransport) AppendEntries

func (n *NetworkTransport) AppendEntries(target string, args *AppendEntriesRequest, resp *AppendEntriesResponse) error

AppendEntries implements the Transport interface.

func (*NetworkTransport) AppendEntriesPipeline

func (n *NetworkTransport) AppendEntriesPipeline(target string) (AppendPipeline, error)

AppendEntriesPipeline returns an interface that can be used to pipeline AppendEntries requests.

func (*NetworkTransport) Close

func (n *NetworkTransport) Close() error

Close is used to stop the network transport.

func (*NetworkTransport) Consumer

func (n *NetworkTransport) Consumer() <-chan RPC

Consumer implements the Transport interface.

func (*NetworkTransport) DecodePeer

func (n *NetworkTransport) DecodePeer(buf []byte) string

DecodePeer implements the Transport interface.

func (*NetworkTransport) EncodePeer

func (n *NetworkTransport) EncodePeer(p string) []byte

EncodePeer implements the Transport interface.

func (*NetworkTransport) InstallSnapshot

func (n *NetworkTransport) InstallSnapshot(target string, args *InstallSnapshotRequest, resp *InstallSnapshotResponse, data io.Reader) error

InstallSnapshot implements the Transport interface.

func (*NetworkTransport) IsShutdown

func (n *NetworkTransport) IsShutdown() bool

IsShutdown is used to check if the transport is shutdown.

func (*NetworkTransport) LocalAddr

func (n *NetworkTransport) LocalAddr() string

LocalAddr implements the Transport interface.

func (*NetworkTransport) RequestVote

func (n *NetworkTransport) RequestVote(target string, args *RequestVoteRequest, resp *RequestVoteResponse) error

RequestVote implements the Transport interface.

func (*NetworkTransport) SetHeartbeatHandler

func (n *NetworkTransport) SetHeartbeatHandler(cb func(rpc RPC))

SetHeartbeatHandler is used to setup a heartbeat handler as a fast-pass. This is to avoid head-of-line blocking from disk IO.

type Observation

type Observation struct {
	// Raft holds the Raft instance generating the observation.
	Raft *Raft
	// Data holds observation-specific data. Possible types are
	// *RequestVoteRequest, RaftState and LeaderObservation.
	Data interface{}
}

Observation is sent along the given channel to observers when an event occurs.

type Observer

type Observer struct {
	// contains filtered or unexported fields
}

Observer describes what to do with a given observation.

func NewObserver

func NewObserver(channel chan Observation, blocking bool, filter FilterFn) *Observer

NewObserver creates a new observer that can be registered to make observations on a Raft instance. Observations will be sent on the given channel if they satisfy the given filter.

If blocking is true, the observer will block when it can't send on the channel, otherwise it may discard events.

func (*Observer) GetNumDropped

func (or *Observer) GetNumDropped() uint64

GetNumDropped returns the number of dropped observations due to blocking.

func (*Observer) GetNumObserved

func (or *Observer) GetNumObserved() uint64

GetNumObserved returns the number of observations.

type PeerStore

type PeerStore interface {
	// Peers returns the list of known peers.
	Peers() ([]string, error)

	// SetPeers sets the list of known peers. This is invoked when a peer is
	// added or removed.
	SetPeers([]string) error
}

PeerStore provides an interface for persistent storage and retrieval of peers. We use a separate interface than StableStore since the peers may need to be edited by a human operator. For example, in a two node cluster, the failure of either node requires human intervention since consensus is impossible.

type RPC

type RPC struct {
	Command  interface{}
	Reader   io.Reader // Set only for InstallSnapshot
	RespChan chan<- RPCResponse
}

RPC has a command, and provides a response mechanism.

func (*RPC) Respond

func (r *RPC) Respond(resp interface{}, err error)

Respond is used to respond with a response, error or both

type RPCResponse

type RPCResponse struct {
	Response interface{}
	Error    error
}

RPCResponse captures both a response and a potential error.

type Raft

type Raft struct {
	// contains filtered or unexported fields
}

Raft implements a Raft node.

func NewRaft

func NewRaft(conf *Config, fsm FSM, logs LogStore, stable StableStore, snaps SnapshotStore,
	peerStore PeerStore, localAddr string, trans Transport) (*Raft, error)

NewRaft is used to construct a new Raft node. It takes a configuration, as well as implementations of various interfaces that are required. If we have any old state, such as snapshots, logs, peers, etc, all those will be restored when creating the Raft node.

func (*Raft) AddPeer

func (r *Raft) AddPeer(peer string) Future

AddPeer is used to add a new peer into the cluster. This must be run on the leader or it will fail.

func (*Raft) AppliedIndex

func (r *Raft) AppliedIndex() uint64

AppliedIndex returns the last index applied to the FSM. This is generally lagging behind the last index, especially for indexes that are persisted but have not yet been considered committed by the leader. NOTE - this reflects the last index that was sent to the application's FSM over the apply channel but DOES NOT mean that the application's FSM has yet consumed it and applied it to its internal state. Thus, the application's state may lag behind this index.

func (*Raft) Apply

func (r *Raft) Apply(cmd []byte, timeout time.Duration) ApplyFuture

Apply is used to apply a command to the FSM in a highly consistent manner. This returns a future that can be used to wait on the application. An optional timeout can be provided to limit the amount of time we wait for the command to be started. This must be run on the leader or it will fail.

func (*Raft) Barrier

func (r *Raft) Barrier(timeout time.Duration) Future

Barrier is used to issue a command that blocks until all preceeding operations have been applied to the FSM. It can be used to ensure the FSM reflects all queued writes. An optional timeout can be provided to limit the amount of time we wait for the command to be started. This must be run on the leader or it will fail.

func (*Raft) DeregisterObserver

func (r *Raft) DeregisterObserver(or *Observer)

DeregisterObserver deregisters an observer.

func (*Raft) LastContact

func (r *Raft) LastContact() time.Time

LastContact returns the time of last contact by a leader. This only makes sense if we are currently a follower.

func (*Raft) LastIndex

func (r *Raft) LastIndex() uint64

LastIndex returns the last index in stable storage, either from the last log or from the last snapshot.

func (*Raft) Leader

func (r *Raft) Leader() string

Leader is used to return the current leader of the cluster. It may return empty string if there is no current leader or the leader is unknown.

func (*Raft) LeaderCh

func (r *Raft) LeaderCh() <-chan bool

LeaderCh is used to get a channel which delivers signals on acquiring or losing leadership. It sends true if we become the leader, and false if we lose it. The channel is not buffered, and does not block on writes.

func (*Raft) Peers

func (r *Raft) Peers() []string

func (*Raft) RegisterObserver

func (r *Raft) RegisterObserver(or *Observer)

RegisterObserver registers a new observer.

func (*Raft) RemovePeer

func (r *Raft) RemovePeer(peer string) Future

RemovePeer is used to remove a peer from the cluster. If the current leader is being removed, it will cause a new election to occur. This must be run on the leader or it will fail.

func (*Raft) SetPeers

func (r *Raft) SetPeers(p []string) Future

SetPeers is used to forcibly replace the set of internal peers and the peerstore with the ones specified. This can be considered unsafe.

func (*Raft) Shutdown

func (r *Raft) Shutdown() Future

Shutdown is used to stop the Raft background routines. This is not a graceful operation. Provides a future that can be used to block until all background routines have exited.

func (*Raft) Snapshot

func (r *Raft) Snapshot() Future

Snapshot is used to manually force Raft to take a snapshot. Returns a future that can be used to block until complete.

func (*Raft) State

func (r *Raft) State() RaftState

State is used to return the current raft state.

func (*Raft) Stats

func (r *Raft) Stats() map[string]string

Stats is used to return a map of various internal stats. This should only be used for informative purposes or debugging.

Keys are: "state", "term", "last_log_index", "last_log_term", "commit_index", "applied_index", "fsm_pending", "last_snapshot_index", "last_snapshot_term", "num_peers" and "last_contact".

The value of "state" is a numerical value representing a RaftState const.

The value of "last_contact" is either "never" if there has been no contact with a leader, "0" if the node is in the leader state, or the time since last contact with a leader formatted as a string.

All other values are uint64s, formatted as strings.

func (*Raft) String

func (r *Raft) String() string

func (*Raft) VerifyLeader

func (r *Raft) VerifyLeader() Future

VerifyLeader is used to ensure the current node is still the leader. This can be done to prevent stale reads when a new leader has potentially been elected.

type RaftState

type RaftState uint32

RaftState captures the state of a Raft node: Follower, Candidate, Leader, or Shutdown.

const (
	// Follower is the initial state of a Raft node.
	Follower RaftState = iota

	// Candidate is one of the valid states of a Raft node.
	Candidate

	// Leader is one of the valid states of a Raft node.
	Leader

	// Shutdown is the terminal state of a Raft node.
	Shutdown
)

func (RaftState) String

func (s RaftState) String() string

type RequestVoteRequest

type RequestVoteRequest struct {
	// Provide the term and our id
	Term      uint64
	Candidate []byte

	// Used to ensure safety
	LastLogIndex uint64
	LastLogTerm  uint64
}

RequestVoteRequest is the command used by a candidate to ask a Raft peer for a vote in an election.

type RequestVoteResponse

type RequestVoteResponse struct {
	// Newer term if leader is out of date
	Term uint64

	// Return the peers, so that a node can shutdown on removal
	Peers []byte

	// Is the vote granted
	Granted bool
}

RequestVoteResponse is the response returned from a RequestVoteRequest.

type SnapshotMeta

type SnapshotMeta struct {
	ID    string // ID is opaque to the store, and is used for opening
	Index uint64
	Term  uint64
	Peers []byte
	Size  int64
}

SnapshotMeta is for metadata of a snapshot.

type SnapshotSink

type SnapshotSink interface {
	io.WriteCloser
	ID() string
	Cancel() error
}

SnapshotSink is returned by StartSnapshot. The FSM will Write state to the sink and call Close on completion. On error, Cancel will be invoked.

type SnapshotStore

type SnapshotStore interface {
	// Create is used to begin a snapshot at a given index and term,
	// with the current peer set already encoded.
	Create(index, term uint64, peers []byte) (SnapshotSink, error)

	// List is used to list the available snapshots in the store.
	// It should return then in descending order, with the highest index first.
	List() ([]*SnapshotMeta, error)

	// Open takes a snapshot ID and provides a ReadCloser. Once close is
	// called it is assumed the snapshot is no longer needed.
	Open(id string) (*SnapshotMeta, io.ReadCloser, error)
}

SnapshotStore interface is used to allow for flexible implementations of snapshot storage and retrieval. For example, a client could implement a shared state store such as S3, allowing new nodes to restore snapshots without streaming from the leader.

type StableStore

type StableStore interface {
	Set(key []byte, val []byte) error

	// Get returns the value for key, or an empty byte slice if key was not found.
	Get(key []byte) ([]byte, error)

	SetUint64(key []byte, val uint64) error

	// GetUint64 returns the uint64 value for key, or 0 if key was not found.
	GetUint64(key []byte) (uint64, error)
}

StableStore is used to provide stable storage of key configurations to ensure safety.

type StaticPeers

type StaticPeers struct {
	StaticPeers []string
	// contains filtered or unexported fields
}

StaticPeers is used to provide a static list of peers.

func (*StaticPeers) Peers

func (s *StaticPeers) Peers() ([]string, error)

Peers implements the PeerStore interface.

func (*StaticPeers) SetPeers

func (s *StaticPeers) SetPeers(p []string) error

SetPeers implements the PeerStore interface.

type StreamLayer

type StreamLayer interface {
	net.Listener

	// Dial is used to create a new outgoing connection
	Dial(address string, timeout time.Duration) (net.Conn, error)
}

StreamLayer is used with the NetworkTransport to provide the low level stream abstraction.

type TCPStreamLayer

type TCPStreamLayer struct {
	// contains filtered or unexported fields
}

TCPStreamLayer implements StreamLayer interface for plain TCP.

func (*TCPStreamLayer) Accept

func (t *TCPStreamLayer) Accept() (c net.Conn, err error)

Accept implements the net.Listener interface.

func (*TCPStreamLayer) Addr

func (t *TCPStreamLayer) Addr() net.Addr

Addr implements the net.Listener interface.

func (*TCPStreamLayer) Close

func (t *TCPStreamLayer) Close() (err error)

Close implements the net.Listener interface.

func (*TCPStreamLayer) Dial

func (t *TCPStreamLayer) Dial(address string, timeout time.Duration) (net.Conn, error)

Dial implements the StreamLayer interface.

type Transport

type Transport interface {
	// Consumer returns a channel that can be used to
	// consume and respond to RPC requests.
	Consumer() <-chan RPC

	// LocalAddr is used to return our local address to distinguish from our peers.
	LocalAddr() string

	// AppendEntriesPipeline returns an interface that can be used to pipeline
	// AppendEntries requests.
	AppendEntriesPipeline(target string) (AppendPipeline, error)

	// AppendEntries sends the appropriate RPC to the target node.
	AppendEntries(target string, args *AppendEntriesRequest, resp *AppendEntriesResponse) error

	// RequestVote sends the appropriate RPC to the target node.
	RequestVote(target string, args *RequestVoteRequest, resp *RequestVoteResponse) error

	// InstallSnapshot is used to push a snapshot down to a follower. The data is read from
	// the ReadCloser and streamed to the client.
	InstallSnapshot(target string, args *InstallSnapshotRequest, resp *InstallSnapshotResponse, data io.Reader) error

	// EncodePeer is used to serialize a peer name.
	EncodePeer(string) []byte

	// DecodePeer is used to deserialize a peer name.
	DecodePeer([]byte) string

	// SetHeartbeatHandler is used to setup a heartbeat handler
	// as a fast-pass. This is to avoid head-of-line blocking from
	// disk IO. If a Transport does not support this, it can simply
	// ignore the call, and push the heartbeat onto the Consumer channel.
	SetHeartbeatHandler(cb func(rpc RPC))
}

Transport provides an interface for network transports to allow Raft to communicate with other nodes.

type WithClose

type WithClose interface {
	// Close permanently closes a transport, stopping
	// any associated goroutines and freeing other resources.
	Close() error
}

WithClose is an interface that a transport may provide which allows a transport to be shut down cleanly when a Raft instance shuts down.

It is defined separately from Transport as unfortunately it wasn't in the original interface specification.

type WithPeers

type WithPeers interface {
	Connect(peer string, t Transport) // Connect a peer
	Disconnect(peer string)           // Disconnect a given peer
	DisconnectAll()                   // Disconnect all peers, possibly to reconnect them later
}

WithPeers is an interface that a transport may provide which allows for connection and disconnection. Unless the transport is a loopback transport, the transport specified to "Connect" is likely to be nil.

Directories

Path Synopsis

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL