Struct LoadBalancer 

pub struct LoadBalancer { /* private fields */ }

Top-level load balancer that manages backend groups keyed by service name.

Implementations

impl LoadBalancer

pub fn new() -> Self

Create an empty load balancer.

pub fn register( &self, service: &str, addrs: Vec<SocketAddr>, strategy: LbStrategy, )

Register (or replace) a backend group for service.

pub fn select(&self, service: &str) -> Option<Arc<Backend>>

Select a healthy backend for service, delegating to the group’s configured strategy.

pub fn update_backends(&self, service: &str, addrs: Vec<SocketAddr>)

Update the backend list for service, preserving state for unchanged addresses.

pub fn unregister(&self, service: &str)

Remove the backend group for service.

pub fn add_backend(&self, service: &str, addr: SocketAddr)

Add a single backend to an existing group.

pub fn remove_backend(&self, service: &str, addr: &SocketAddr)

Remove a single backend from an existing group.

pub fn backend_count(&self, service: &str) -> usize

Return the number of backends registered for service, or 0 if the service is not registered.

pub fn healthy_count(&self, service: &str) -> usize

Return the number of healthy backends for service, or 0 if the service is not registered.

pub fn mark_health(&self, service: &str, addr: &SocketAddr, healthy: bool)

Update the health status of a specific backend in a service group.

If healthy is true, marks the backend healthy and resets its failure counter. Otherwise marks it unhealthy and records a failure.

pub fn list_service_names(&self) -> Vec<String>

Return the list of registered service names.

pub fn group_snapshot(&self, service: &str) -> Option<BackendGroupSnapshot>

Return a snapshot of a backend group’s state for a given service.

Each entry contains the backend address, health status, active connections, and the group’s strategy.

pub fn spawn_health_checker( self: &Arc<Self>, interval: Duration, timeout: Duration, ) -> JoinHandle<()>

Spawn a background health-check task.

Every interval the task TCP-connects to every backend across all groups with a per-probe timeout. Concurrency is bounded to at most 64 simultaneous probes via a semaphore.

On success the backend is marked healthy and its failure counter is reset. On failure it is marked unhealthy and the failure counter is incremented.

Panics

The spawned task panics if the internal concurrency semaphore is unexpectedly closed.

Trait Implementations

impl Default for LoadBalancer

fn default() -> Self

Returns the “default value” for a type.