Struct ControllerConfig 

pub struct ControllerConfig {

    pub client_dispatch_queue_capacity: ConnectionDispatchQueueCapacity,
    pub client_frame_max: ClientFrameMax,
    pub node_id: NodeId,
    pub bootstrap_servers: Vec<String>,
    pub directory_id: Uuid,
    pub auto_join: bool,
    pub observer_lag_bound: u64,
    pub initial_voters: VoterSet,
    pub controller_listen_addr: SocketAddr,
    pub log_dir: PathBuf,
    pub election_timeout: Time,
    pub fetch_timeout: Time,
    pub heartbeat_interval: Option<Time>,
    pub controller_fetch_miss_limit: ControllerFetchMissLimit,
    pub metadata_raft_command_queue_capacity: MetadataRaftCommandQueueCapacity,
    pub metadata_raft_fetch_max: MetadataRaftFetchMax,
    pub client_id: String,
    pub bootstrap_mode: BootstrapMode,
    pub bootstrap_records: Vec<MetadataRecord>,
    pub default_min_insync_replicas: i32,
    pub cluster_id: Option<Uuid>,
    pub dialer: Option<Arc<dyn OutboundDialer>>,
    pub handshake: Option<Arc<dyn RaftListenerHandshake>>,
    pub shard_router: Option<Arc<dyn RaftShardRouter>>,
    pub admin_router: Option<Arc<dyn ControllerAdminRouter>>,
    pub unstable_api_versions: UnstableApiVersions,
    pub unstable_feature_versions: UnstableFeatureVersions,
    pub listener_limits: ListenerLimits,
    pub max_bytes_between_snapshots: ByteSize,
    pub max_snapshot_interval: Time,
    pub snapshot_interval_records: u64,
    pub metadata_snapshot_fetch_max: ByteSize,
    pub metadata_log: MetadataLogConfig,
    pub event_queue_recorder: Arc<dyn EventQueueRecorder>,
}

Fields

client_dispatch_queue_capacity: ConnectionDispatchQueueCapacity

Capacity used by outbound controller client connections.

client_frame_max: ClientFrameMax

Maximum frame size used by outbound controller client connections.

node_id: NodeId

bootstrap_servers: Vec<String>

Endpoints used only to discover the leader at cold start (KIP-853 dynamic).

directory_id: Uuid

This replica’s stable directory id (generated at format time).

auto_join: bool

Issue AddVoter for self once caught up as an observer.

observer_lag_bound: u64

Max allowed lag (in log entries) for an observer to be promotable.

initial_voters: VoterSet

Initial voter set for the bootstrapping node only; empty for joiners.

controller_listen_addr: SocketAddr

log_dir: PathBuf

The metadata log directory, Kafka’s metadata.log.dir. The controller keeps the metadata partition in its METADATA_PARTITION_DIR subdirectory.

election_timeout: Time

Kafka’s controller.quorum.election.timeout.ms: the base of the randomized election timer.

fetch_timeout: Time

Kafka’s controller.quorum.fetch.timeout.ms: the follower and observer fetch watchdog, and the base of the leader’s check-quorum window.

heartbeat_interval: Option<Time>

Explicit heartbeat cadence. None preserves the derived fetch_timeout / 3 behavior.

controller_fetch_miss_limit: ControllerFetchMissLimit

metadata_raft_command_queue_capacity: MetadataRaftCommandQueueCapacity

metadata_raft_fetch_max: MetadataRaftFetchMax

client_id: String

bootstrap_mode: BootstrapMode

bootstrap_records: Vec<MetadataRecord>

The bootstrap metadata of a static quorum: Kafka’s bootstrap.checkpoint, which krabka-format writes as bootstrap.records.bin.

The active controller writes these records once, when it activates on a metadata log that holds no metadata.version. A bootstrap checkpoint at offset 0 and epoch 0 that holds metadata records replaces them, as Kafka’s QuorumController.handleLoadBootstrap does. An empty list writes nothing.

default_min_insync_replicas: i32

This node’s static min.insync.replicas: Kafka’s ConfigurationControlManager.getStaticallyConfiguredMinInsyncReplicas. When the bootstrap records finalize eligible.leader.replicas.version above 0, the active controller writes it as the cluster-level min.insync.replicas together with them, as Kafka’s ActivationRecordsGenerator.recordsForEmptyLog does. Default: 1.

cluster_id: Option<Uuid>

Cluster UUID applied to the MetadataImage on first construction. None falls back to Uuid::nil() (legacy single-node default). The operator sets this to the KafkaCluster UID so every broker in the same cluster shares one identifier across restarts.

dialer: Option<Arc<dyn OutboundDialer>>

Optional outbound dialer. None means: open a plain TCP socket to peers (legacy PLAINTEXT-only path). The broker injects an InterBrokerClient-backed dialer here when inter-broker TLS or SASL is configured.

handshake: Option<Arc<dyn RaftListenerHandshake>>

Optional inbound handshake hook. None keeps the legacy PLAINTEXT path. The broker injects a BrokerRaftHandshake implementation here when the controller listener should terminate TLS and/or SASL before raft frames start flowing.

shard_router: Option<Arc<dyn RaftShardRouter>>

Optional KIP-595 shard router. Metadata traffic returns None; diskless WAL shards return an encoded response body and bypass metadata dispatch.

admin_router: Option<Arc<dyn ControllerAdminRouter>>

Optional KIP-919 Admin router. The broker injects its existing handler registry here after construction, keeping controller and broker semantics on one implementation.

unstable_api_versions: UnstableApiVersions

Kafka’s internal unstable.api.versions.enable: whether the controller listener advertises and accepts a latestVersionUnstable version, and whether it advertises feature levels past the latest production ones. Kafka builds the listener’s SimpleApiVersionManager with this flag alone.

unstable_feature_versions: UnstableFeatureVersions

Kafka’s internal unstable.feature.versions.enable: the flag the controller’s own QuorumFeatures use, so it caps the feature levels this controller accepts on replay and refuses to start above.

listener_limits: ListenerLimits

The request-size, idle and connection limits of the controller listener.

max_bytes_between_snapshots: ByteSize

metadata.log.max.record.bytes.between.snapshots (default 20 MiB).

max_snapshot_interval: Time

metadata.log.max.snapshot.interval.ms (default 1 h; 0 = disabled).

snapshot_interval_records: u64

Snapshot once committed offset advances this many records past the last snapshot. The cleaning by Self::metadata_log’s retention limits then decides when the log below a snapshot goes. 0 disables this trigger.

metadata_snapshot_fetch_max: ByteSize

Maximum metadata snapshot size this follower will fetch. Deployments may lower the default 1 GiB security ceiling but cannot raise it.

metadata_log: MetadataLogConfig

How the metadata log rolls, how long it keeps the prefix a snapshot covers, and how often an idle leader appends to it.

event_queue_recorder: Arc<dyn EventQueueRecorder>

Where the engine loop reports Kafka’s controller event-queue measurements. The broker passes its metrics registry.

Implementations

impl ControllerConfig

pub fn for_tests(node_id: NodeId, log_dir: PathBuf) -> Self

Trait Implementations

impl Clone for ControllerConfig

fn clone(&self) -> ControllerConfig

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for ControllerConfig

fn fmt(&self, __derive_more_f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.