Struct DispatchHandle 

pub struct DispatchHandle(/* private fields */);

Dynamic-dispatch wrapper around Arc<WorkerPool> — the one concrete DispatchEngine impl today. Extract this into nzb-dispatch in Phase B.

Implementations

impl DispatchHandle

pub fn new(pool: Arc<WorkerPool>) -> Self

pub fn pool(&self) -> &Arc<WorkerPool>

Escape hatch: access the underlying pool. Intended for callers that still need pool-specific APIs not yet promoted to the trait (none today, but keeps the migration incremental).

Trait Implementations

impl DispatchEngine for DispatchHandle

fn start(&self)

Spawn workers for all currently enabled servers and start the supervisor task. Idempotent (safe to call more than once).

fn submit_job(&self, job: &NzbJob, progress_tx: Sender<ProgressUpdate>)

Register a new job and begin dispatching its unfinished articles. Progress is streamed to progress_tx as ProgressUpdates; the channel is closed when the job reaches a terminal state.

fn pause_job(&self, job_id: &str)

Pause dispatch for job_id. In-flight articles finish; no new work is popped for this job until resume_job.

fn resume_job(&self, job_id: &str)

Resume a paused job.

fn cancel_job(&self, job_id: &str)

Cancel job_id. In-flight articles may still complete but their results are discarded; no terminal progress update is emitted beyond the one triggered by cancellation.

fn abort_job(&self, job_id: &str, reason: String)

Abort job_id with a human-readable reason. Emits ProgressUpdate::JobAborted once outstanding articles drain.

fn has_job(&self, job_id: &str) -> bool

Is job_id currently known to the dispatcher?

fn reconcile_servers(&self)

Re-read the server list and adjust workers to match. Call after any mutation to the server config (add, remove, enable, disable, resize).

fn set_max_worker_idle(&self, d: Duration)

Override the idle-worker eviction threshold. Tests shrink this to make the watchdog converge in seconds; production uses the default.

fn eviction_count(&self) -> u64

Lifetime count of worker evictions performed by the heartbeat watchdog. Increases by 1 each time the supervisor reclaims a stalled worker. Useful for tests and observability.

fn shutdown<'life0, 'async_trait>( &'life0 self, ) -> Pin<Box<dyn Future<Output = ()> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait,

Gracefully shut down: stop accepting new work, signal all workers, and wait for the supervisor to exit.

fn server_stats_snapshot(&self) -> Vec<(String, ServerAttemptStats)>

Snapshot of per-server lifetime attempt counters. Used by the queue manager to emit a diagnostic breakdown alongside a job abort — distinguishes “server returned 430 for every article” (dead NZB) from “server had auth errors” (transient). Default is empty for engines that don’t track per-server stats.