Struct Client 

pub struct Client {
    pub endpoint: Endpoint,
    pub instance_source: Arc<Receiver<Vec<Instance>>>,
    /* private fields */
}

Fields

endpoint: Endpoint

instance_source: Arc<Receiver<Vec<Instance>>>

Implementations

impl Client

pub fn instances(&self) -> Vec<Instance>

Instances available from watching key-value store

pub fn instance_ids(&self) -> Vec<u64>

pub fn is_instance_discovered(&self, instance_id: u64) -> bool

Whether the latest discovery snapshot contains this instance, including inhibited workers.

pub fn instance_ids_avail(&self) -> Vec<u64>

pub fn instance_ids_free(&self) -> Vec<u64>

Routable instance ids excluding those currently flagged overloaded — the set used for load-aware (random / round-robin) worker selection.

pub fn routing_instance_counts(&self) -> RoutingInstanceCounts

pub fn instance_avail_watcher(&self) -> Receiver<Vec<u64>>

Get a watcher for available instance IDs

pub fn with_admitted_instances(&self, admitted_ids: Receiver<Vec<u64>>) -> Self

Create a client view whose routable instances are restricted by a caller-owned admission set.

Endpoint discovery remains the source of connection metadata and hard availability. The returned client publishes only the intersection of that endpoint membership and admitted_ids, allowing a higher-level controller to keep discovered-but-unvalidated instances out of a routing group. The view has independent overload and fault-inhibition state, just like a freshly constructed client.

pub fn with_admitted_instances_and_cancellation( &self, admitted_ids: Receiver<Vec<u64>>, cancel_token: CancellationToken, ) -> Self

Like Self::with_admitted_instances, with a lifecycle token for construction-time cancellation by an owning controller.

pub async fn wait_for_instances(&self) -> Result<Vec<Instance>>

Wait for at least one Instance to be available for this Endpoint

pub fn report_instance_down(&self, instance_id: u64)

Mark an instance as down/unavailable

pub fn set_overloaded_instances(&self, overloaded_instance_ids: &[u64]) -> bool

Replace the set of overloaded instances reported by the worker monitor. Returns true when this changes the routing snapshot.

pub fn overload_reconciliation_needed(&self) -> bool

Whether request-path backpressure changed overload state after the monitor’s most recent metric publication.

pub fn mark_overloaded_immediate(&self, instance_id: u64)

Mark an instance overloaded immediately after a worker-scoped WorkerOverloaded response. This is backpressure, not a fault, so it does not call report_instance_down. The next worker-monitor reconciliation replaces this short-lived global routing hint.

pub fn clear_overloaded_instances_for_removed( &self, removed_instance_ids: &[u64], )

pub fn overloaded_instance_ids(&self) -> Option<HashSet<u64>>

pub fn available_instance_ids(&self) -> Option<Arc<HashSet<u64>>>

Workers currently eligible for selection: discovered and not locally inhibited by Self::report_instance_down.

This hard-availability snapshot is separate from transient overload. None means this client has not discovered an instance yet. After the first discovery, Some is authoritative, including Some(empty) when the last previously discovered worker is removed.

Trait Implementations

impl Clone for Client

fn clone(&self) -> Client

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for Client

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

Formats the value using the given formatter.