Struct JobCoordinator 

pub struct JobCoordinator { /* private fields */ }

Orchestrates chunk execution with retries and idempotent output capture.

The coordinator iterates over work chunks, executes each via a caller-provided function, stores outputs in a JobOutputStore, and retries transient failures according to RetryPolicy. Once all chunks are processed, the caller-provided reducer aggregates the outputs.

Chunks whose output already exists in the store are skipped (idempotent replay). This enables safe restart after partial failure.

Implementations

impl JobCoordinator

pub fn new(policy: RetryPolicy) -> Self

Create a new coordinator with the given retry policy.

pub fn run<F, R>( &mut self, chunks: &[Box<dyn WorkChunk>], store: &mut dyn JobOutputStore, execute_fn: F, reduce_fn: R, ) -> Result<JobResult, FabricError>

where F: FnMut(&[u8]) -> Result<Vec<u8>, FabricError>, R: FnOnce(&[(ChunkId, Vec<u8>)]) -> Vec<u8>,

Execute a job over the given chunks.

Parameters

• chunks: Work units to process.
• store: Output store for idempotent capture.
• execute_fn: Called with serialized chunk bytes; returns output bytes on success.
• reduce_fn: Called with all (ChunkId, output_bytes) pairs to produce the final aggregate.

Errors

Returns FabricError only on store failures. Chunk-level failures are captured in JobResult::chunk_results; the job itself succeeds as long as the store is functional.

pub fn teardown(&self, store: &mut dyn JobOutputStore)

Teardown: clear the output store.