Struct RemoteFragmentTransport 

pub struct RemoteFragmentTransport { /* private fields */ }

Coordinator-side Arrow IPC transport for tablet fragments.

Each tablet route names an authenticated RPC carrier. Output is pulled one batch at a time, so consumer polling is the backpressure mechanism.

Implementations

impl RemoteFragmentTransport

pub fn new(default_client: Arc<dyn FragmentRpcClient>) -> Self

Creates a transport whose tablets use default_client.

pub fn with_message_limit( default_client: Arc<dyn FragmentRpcClient>, max_message_bytes: usize, ) -> Self

Creates a transport with an explicit per-call body bound.

pub fn routed() -> Self

Creates a fail-closed transport with no fallback route.

pub fn with_client( self, tablet: TabletId, client: Arc<dyn FragmentRpcClient>, ) -> Self

Routes one tablet through a specific peer carrier.

Trait Implementations

impl FragmentTransport for RemoteFragmentTransport

fn execute_fragment<'life0, 'life1, 'async_trait>( &'life0 self, query_id: QueryId, fragment: &'life1 PlanFragment, inputs: Vec<FragmentStream>, control: FragmentControl, ) -> Pin<Box<dyn Future<Output = DistributedResult<FragmentStream>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait,

Starts a fragment on its assigned worker and returns its output stream.

fn cancel_fragment( &self, query_id: QueryId, fragment_id: FragmentId, ) -> DistributedResult<()>

Best-effort cancellation of a running (or abandoned) fragment.

fn refill_top_k<'life0, 'life1, 'async_trait>( &'life0 self, query_id: QueryId, fragment: &'life1 PlanFragment, offset: usize, limit: usize, control: FragmentControl, ) -> Pin<Box<dyn Future<Output = DistributedResult<TopKRefill>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait,

Fetches the next top-k batch of a fragment’s tablet (adaptive refill). Transports without a refill binding leave the default.