Struct ShutdownCoordinator 

pub struct ShutdownCoordinator { /* private fields */ }

Coordinates graceful shutdown across the system.

Shutdown phases

1. Stop accepting — signal is broadcast, new requests get 503.
2. Drain in-flight — wait for the in-flight counter to reach zero.
3. Cage all gorillas — stop autonomous agents.
4. Flush logs — give tracing subscribers time to flush.
5. Close DB — release database handles.

Hooks registered via register_hook fire in registration order after the drain phase.

Implementations

impl ShutdownCoordinator

pub fn new(drain_timeout: Duration) -> Arc<Self>

Create a new shutdown coordinator with the given drain timeout.

pub fn with_default_timeout() -> Arc<Self>

Create a coordinator with the default 30-second drain timeout.

pub fn subscribe(&self) -> Receiver<bool>

Subscribe to the shutdown signal.

The returned receiver yields true once shutdown is initiated.

pub fn is_shutting_down(&self) -> bool

Returns true if shutdown has been initiated.

pub fn track_request(&self) -> bool

Increment the in-flight request counter.

Returns false if shutdown is in progress (caller should reject the request with 503).

pub fn finish_request(&self)

Decrement the in-flight request counter.

pub fn in_flight_count(&self) -> usize

Current number of in-flight requests.

pub async fn register_hook<F, Fut>(&self, hook: F)

where F: Fn() -> Fut + Send + Sync + 'static, Fut: Future<Output = ()> + Send + 'static,

Register a hook that will be called during shutdown.

Hooks fire in registration order after the drain phase completes or times out.

pub async fn initiate_shutdown(&self)

Initiate graceful shutdown.

This method is idempotent — calling it multiple times has no additional effect.

1. Broadcasts the shutdown signal (new requests will be rejected).
2. Waits for in-flight requests to drain (up to drain_timeout).
3. Runs all registered hooks in order.