cluster

Documentation

Overview

Package cluster manages the node's membership in the Hive cluster, including peer tracking, consistent hashing, data routing, and rebalancing.

Index

• Variables
• type Cluster
  • func NewCluster(cfg Config) (*Cluster, error)
  • func (m *Cluster) Exec(op transport.Op, key string, args ...[]byte) ([][]byte, error)
  • func (m *Cluster) KeyCount() int
  • func (m *Cluster) Peers() []PeerInfo
  • func (m *Cluster) Shutdown() error
  • func (m *Cluster) Used() int64
• type Config
• type NodeStatus
• type OpScope
• type PeerInfo

Variables

var ErrNotFound = errors.New("hive: not found")

ErrNotFound is returned when a requested key or field does not exist or has expired.

Types

type Cluster

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

Cluster owns the cluster state for this node.

func NewCluster

func NewCluster(cfg Config) (*Cluster, error)

NewCluster creates and starts a cluster Cluster. In clustered mode it binds a TCP server and contacts Seeds to join.

func (*Cluster) Exec

func (m *Cluster) Exec(op transport.Op, key string, args ...[]byte) ([][]byte, error)

Exec is the exported entry point for store files to run a cluster op. It delegates to dispatch, keeping routing logic internal to this package.

func (*Cluster) KeyCount

func (m *Cluster) KeyCount() int

KeyCount returns the number of live keys held locally by this node.

func (*Cluster) Peers

func (m *Cluster) Peers() []PeerInfo

Peers returns a snapshot of all known peers.

func (*Cluster) Shutdown

func (m *Cluster) Shutdown() error

Shutdown gracefully stops the node. Safe to call more than once.

func (*Cluster) Used

func (m *Cluster) Used() int64

Used returns this node's current estimated local byte usage.

type Config

type Config struct {
	NodeID               string
	BindAddr             string
	BindPort             int
	Seeds                []string
	RoutingTimeout       time.Duration
	ConnPoolSize         int
	ReplicationFactor    int
	MemLimit             uint64
	GossipInterval       time.Duration
	GossipFanout         int
	GossipTimeout        time.Duration
	RebalanceDebounce    time.Duration
	RebalanceBatchSize   int
	ReplicationQueueSize int
	ReplicationBatchSize int
	CleanupInterval      time.Duration
	Clustered            bool
	TLSConfig            *tls.Config
}

Config holds all configuration for the cluster manager.

type NodeStatus

type NodeStatus uint8

NodeStatus represents a peer's liveness state in the cluster.

const (
	NodeAlive NodeStatus = 0
	NodeDead  NodeStatus = 1
)

type OpScope

type OpScope uint8

OpScope describes how a cluster op is routed and replicated.

const (
	ScopeWrite OpScope = iota // route to primary owner, replicate to replicas
	ScopeRead                 // route to primary owner, no replication
	ScopeLocal                // always execute on the receiving node
)

type PeerInfo

type PeerInfo struct {
	NodeID            string
	Addr              string
	Status            NodeStatus
	Incarnation       uint64
	ReplicationFactor int
	MemLimit          uint64
	MemUsed           uint64
}

PeerInfo is the canonical peer representation used both as internal mutable state and as the read-only snapshot returned by Peers().