Struct WorkerPool 

pub struct WorkerPool { /* private fields */ }

Implementations

impl WorkerPool

pub async fn spawn( manifest_json: &[u8], worker_entry: PathBuf, n: usize, ) -> Result<Arc<Self>, WorkerError>

Spawn n workers loading manifest_path and start the watchdog.

pub async fn spawn_with_config( manifest_json: &[u8], worker_entry: PathBuf, n: usize, config_path: Option<PathBuf>, ) -> Result<Arc<Self>, WorkerError>

Like spawn, but passes config_path to each worker so its render entrypoints can reach adapters declared in mesofact.config.toml.

pub fn attach_metrics(&self, metrics: Arc<Metrics>)

Attach the shared metrics registry so respawns bump the restarting worker-pool gauge. Called once after both the pool and AppState exist.

pub async fn live_count(&self) -> usize

Live worker count — the ready worker-pool gauge, read at scrape time.

pub async fn get(&self) -> Option<Arc<WorkerClient>>

Return a live worker, round-robin across the pool. Per-worker I/O still serializes on that worker’s socket mutex (see worker_client cleanup), so spreading requests across workers is what buys real concurrency.

pub async fn drain_all(self: Arc<Self>)

Drain all workers and wait for them to exit (used during rolling reload).