lb.cluster — org.clojars.pgdad/clj-ebpf-lb 0.10.2

Public API for distributed state sharing.

This namespace provides the public interface for cluster functionality:

• Cluster lifecycle (start!/stop!)
• State broadcasting
• Membership queries
• Event subscription

Configuration example: {:cluster {:enabled true :node-id "auto" :bind-address "0.0.0.0" :bind-port 7946 :seeds ["192.168.1.10:7946" "192.168.1.11:7946"] :gossip-interval-ms 200 :gossip-fanout 2 :push-pull-interval-ms 10000 :sync-health true :sync-circuit-breaker true :sync-drain true :sync-conntrack true}}

alive-nodes clj

(alive-nodes)

Get set of alive node IDs (including self).

all-nodes clj

(all-nodes)

Get map of all known nodes: node-id -> NodeInfo.

broadcast! clj

(broadcast! state-type key value)

Broadcast a state change to the cluster.

Args:

• state-type: Keyword identifying state type (e.g., :health, :circuit-breaker)
• key: Unique identifier within state type (e.g., target-id)
• value: State value (must be serializable as EDN)

Returns the SyncableState record that was broadcast, or nil if cluster not running.

broadcast-many! clj

(broadcast-many! states)

Broadcast multiple state changes to the cluster.

Args:

• states: Vector of SyncableState records

Use this for batch updates to reduce network overhead.

cluster-size clj

(cluster-size)

Get number of alive nodes in the cluster.

get-all-nodes clj

(get-all-nodes)

Get map of all known nodes: node-id -> NodeInfo.

get-config clj

(get-config)

Get current cluster configuration.

get-node-id clj

(get-node-id)

Get this node's unique identifier.

IStateProvider clj protocol

Protocol for state providers. Import from lb.cluster.protocol for implementation.

local-node clj

(local-node)

Get this node's NodeInfo record.

make-syncable-state clj

Create a SyncableState record. (make-syncable-state state-type key value version source-node)

next-version clj

Get next Lamport timestamp for versioning.

node-alive? clj

(node-alive? node-id)

Check if a specific node is considered alive.

node-dead? clj

(node-dead? node-id)

Check if a specific node is confirmed dead.

node-id clj

(node-id)

Get this node's unique identifier.

node-suspected? clj

(node-suspected? node-id)

Check if a specific node is suspected of failure.

register-provider! clj

(register-provider! provider)

Register a state provider for synchronization.

A state provider must implement lb.cluster.protocol/IStateProvider:

• provider-type: Return keyword identifying provider (e.g., :health)
• get-sync-state: Return vector of SyncableState records
• get-state-digest: Return map of [type key] -> version
• apply-remote-state: Apply received remote states
• on-node-failure: Handle node failure (for failover)

running? clj

(running?)

Check if cluster mode is running.

start! clj

(start! config)

Start cluster mode with the given configuration.

Config options: :enabled - Enable/disable cluster mode (default false) :node-id - Node identifier, "auto" generates UUID (default "auto") :bind-address - Address to bind gossip transport (default "0.0.0.0") :bind-port - Port for gossip communication (default 7946) :seeds - List of seed node addresses ["ip:port", ...]

Gossip tuning: :gossip-interval-ms - Gossip tick interval (default 200) :gossip-fanout - Peers to gossip to per tick (default 2) :push-pull-interval-ms - Full state sync interval (default 10000)

Failure detection: :ping-interval-ms - Probe interval (default 1000) :ping-timeout-ms - Probe timeout (default 500) :suspicion-mult - Suspicion multiplier (default 3)

State sync: :sync-health - Sync health status (default true) :sync-circuit-breaker - Sync circuit breaker state (default true) :sync-drain - Sync drain coordination (default true) :sync-conntrack - Sync connection tracking (default true)

Returns {:node-id <id> :address <addr>} on success, nil if disabled/already running.

stats clj

(stats)

Get cluster statistics.

Returns map with:

• :status - :stopped, :starting, :running, :stopping
• :membership - {:node-id, :alive-count, :suspected-count, :dead-count}
• :gossip - {:running?, :provider-count, :config}
• :config - Current cluster configuration

stop! clj

(stop!)

Stop cluster mode, gracefully leaving the cluster.

subscribe! clj

(subscribe! callback)

Subscribe to cluster events.

Callback receives ClusterEvent records with:

• :event-type - One of: :node-join, :node-leave, :node-suspect, :node-alive, :node-dead, :cluster-started, :cluster-stopped
• :node-id - Node related to event (if applicable)
• :timestamp - When event occurred
• :data - Event-specific data

Returns an unsubscribe function.

unregister-provider! clj

(unregister-provider! type)

Unregister a state provider by type.