Documentation
¶
Overview ¶
Package raft contains Raft consensus for WebMesh.
Package raft contains Raft consensus for WebMesh.
Index ¶
Constants ¶
const ( // DefaultListenPort is the default raft listen port DefaultListenPort = 9000 // DefaultListenAddress is the default raft listen address DefaultListenAddress = "[::]:9000" )
const ( Follower = raft.Follower Candidate = raft.Candidate Leader = raft.Leader Shutdown = raft.Shutdown )
Raft states.
const ( Voter = raft.Voter Nonvoter = raft.Nonvoter )
Raft suffrage states.
Variables ¶
var ( // ErrStarted is returned when the Raft node is already started. ErrStarted = errors.New("raft node already started") // ErrClosed is returned when the Raft node is already closed. ErrClosed = errors.New("raft node is closed") // ErrNoLeader is returned when there is no leader. ErrNoLeader = errors.New("no leader") // ErrAlreadyBootstrapped is returned when the Raft node is already bootstrapped. ErrAlreadyBootstrapped = transport.ErrAlreadyBootstrapped // ErrNotLeader is returned when the Raft node is not the leader. ErrNotLeader = raft.ErrNotLeader // ErrNotVoter is returned when the Raft node is not a voter. ErrNotVoter = raft.ErrNotVoter )
var DefaultDataDir = func() string { if runtime.GOOS == "windows" { return "C:\\ProgramData\\webmesh\\store" } return "/var/lib/webmesh/store" }()
DefaultDataDir is the default data directory.
var ErrNotRaftMember = errors.New("not a raft member")
ErrNotRaftMember is returned for methods that are only valid on raft members.
Functions ¶
func IsNoLeaderErr ¶ added in v0.6.0
IsNoLeaderErr returns true if the error is a raft.ErrNoLeader.
Types ¶
type BootstrapCallback ¶ added in v0.3.1
BootstrapCallback is a callback for when the cluster is bootstrapped. The isLeader flag is set to true if the node is the leader, and false otherwise.
type LeaderObservation ¶
type LeaderObservation = raft.LeaderObservation
LeaderObservation is an alias for raft.LeaderObservation.
type MonotonicLogStore ¶
MonotonicLogStore is a LogStore that is monotonic.
func (*MonotonicLogStore) IsMonotonic ¶
func (m *MonotonicLogStore) IsMonotonic() bool
IsMonotonic returns true if the log store is monotonic.
type Observation ¶
type Observation = raft.Observation
Observation is an alias for raft.Observation.
type ObservationCallback ¶ added in v0.3.1
type ObservationCallback func(ctx context.Context, obs Observation)
ObservationCallback is a callback that can be registered for when an observation is received.
type Options ¶
type Options struct { // NodeID is the node ID. NodeID string // DataDir is the directory to store data in. DataDir string // InMemory is if the store should be in memory. This should only be used for testing and ephemeral nodes. InMemory bool // ConnectionPoolCount is the number of connections to pool. If 0, no connection pooling is used. ConnectionPoolCount int // ConnectionTimeout is the timeout for connections. ConnectionTimeout time.Duration // HeartbeatTimeout is the timeout for heartbeats. HeartbeatTimeout time.Duration // ElectionTimeout is the timeout for elections. ElectionTimeout time.Duration // ApplyTimeout is the timeout for applying. ApplyTimeout time.Duration // CommitTimeout is the timeout for committing. CommitTimeout time.Duration // MaxAppendEntries is the maximum number of append entries. MaxAppendEntries int // LeaderLeaseTimeout is the timeout for leader leases. LeaderLeaseTimeout time.Duration // SnapshotInterval is the interval to take snapshots. SnapshotInterval time.Duration // SnapshotThreshold is the threshold to take snapshots. SnapshotThreshold uint64 // SnapshotRetention is the number of snapshots to retain. SnapshotRetention uint64 // ObserverChanBuffer is the buffer size for the observer channel. ObserverChanBuffer int // LogLevel is the log level for the raft backend. LogLevel string }
Options are the raft options.
func NewOptions ¶
NewOptions returns new raft options with sensible defaults.
type PeerObservation ¶
type PeerObservation = raft.PeerObservation
PeerObservation is an alias for raft.PeerObservation.
type Raft ¶
type Raft interface { // OnApply registers a callback for when a log is applied to the state machine. OnApply(cb fsm.ApplyCallback) // OnSnapshotRestore registers a callback for when a snapshot is restored. OnSnapshotRestore(cb SnapshotRestoreCallback) // OnObservation registers a callback for when an observation is received. OnObservation(cb ObservationCallback) // Start starts the Raft node. Start(ctx context.Context, opts StartOptions) error // ID returns the node ID. ID() string // Bootstrap attempts to bootstrap the Raft cluster. If the cluster is already // bootstrapped, ErrAlreadyBootstrapped is returned. Bootstrap(ctx context.Context) error // Storage returns the storage. This is only valid after Start is called. Storage() storage.MeshStorage // Configuration returns the current raft configuration. Configuration() (raft.Configuration, error) // LastIndex returns the last index sent to the Raft instance. LastIndex() uint64 // LastAppliedIndex returns the last applied index. LastAppliedIndex() uint64 // ListenPort returns the listen port. ListenPort() uint16 // LeaderID returns the leader ID. LeaderID() (string, error) // IsLeader returns true if the Raft node is the leader. IsLeader() bool // IsVoter returns true if the Raft node is a voter. IsVoter() bool // IsObserver returns true if the Raft node is an observer. IsObserver() bool // IsMember is a convenience method that returns true if the Raft node is a voter or observer. IsMember() bool // AddNonVoter adds a non-voting node to the cluster with timeout enforced by the context. AddNonVoter(ctx context.Context, id string, addr string) error // AddVoter adds a voting node to the cluster with timeout enforced by the context. AddVoter(ctx context.Context, id string, addr string) error // DemoteVoter demotes a voting node to a non-voting node with timeout enforced by the context. DemoteVoter(ctx context.Context, id string) error // RemoveServer removes a peer from the cluster with timeout enforced by the context. RemoveServer(ctx context.Context, id string, wait bool) error // Apply applies a raft log entry. Apply(ctx context.Context, log *v1.RaftLogEntry) (*v1.RaftApplyResponse, error) // Snapshot requests a raft snapshot. It returns a reader containing the contents // and metadata about the snapshot. Snapshot() (*raft.SnapshotMeta, io.ReadCloser, error) // Barrier issues a barrier request to the cluster. This is a no-op if the node is not the leader. Barrier(ctx context.Context, timeout time.Duration) (took time.Duration, err error) // Stop stops the Raft node. Stop(ctx context.Context) error }
Raft is the interface for Raft consensus and storage.
func NewPassthrough ¶ added in v0.2.0
NewPassthrough creates a new raft instance that is a no-op for most methods and uses the given Dialer for storage connections.
type SnapshotMeta ¶
type SnapshotMeta = raft.SnapshotMeta
SnapshotMeta is an alias for raft.SnapshotMeta.
type SnapshotRestoreCallback ¶ added in v0.3.2
type SnapshotRestoreCallback func(ctx context.Context, meta *raft.SnapshotMeta, data io.ReadCloser)
SnapshotRestoreCallback is a callback that can be registered for when a snapshot is restored.
type StartOptions ¶
type StartOptions struct { // Transport is the Raft transport. Transport transport.RaftTransport // MeshStorage is the mesh storage. MeshStorage storage.MeshStorage // RaftStorage is the Raft storage. RaftStorage storage.RaftStorage }
StartOptons are options for starting a Raft node.