Struct OffloadEngineBuilder 

pub struct OffloadEngineBuilder { /* private fields */ }

Builder for OffloadEngine.

Implementations

impl OffloadEngineBuilder

pub fn new(leader: Arc<InstanceLeader>) -> Self

Create a new builder with the given instance leader.

pub fn with_runtime(self, runtime: Handle) -> Self

Set an explicit runtime handle for spawning pipeline tasks.

If not set, defaults to leader.runtime(). Use this when you need pipeline tasks to run on a specific runtime (e.g., in tests).

pub fn with_registry(self, registry: Arc<BlockRegistry>) -> Self

Set the block registry.

pub fn with_g1_manager(self, manager: Arc<BlockManager<G1>>) -> Self

Set the G1 block manager.

pub fn with_g2_manager(self, manager: Arc<BlockManager<G2>>) -> Self

Set the G2 block manager.

pub fn with_g3_manager(self, manager: Arc<BlockManager<G3>>) -> Self

Set the G3 block manager.

pub fn with_object_ops(self, object_ops: Arc<dyn ObjectBlockOps>) -> Self

Set object storage operations for G4.

G4 is object storage (S3, MinIO, etc.) and uses ObjectBlockOps instead of a BlockManager. This replaces with_g4_manager.

pub fn with_g2_physical_layout(self, layout: PhysicalLayout) -> Self

Set the G2 physical layout for object transfers.

Required when using G2→G4 pipeline. The ObjectTransferExecutor needs the physical layout to read block data for upload to object storage.

pub fn with_g1_to_g2_pipeline(self, config: PipelineConfig<G1, G2>) -> Self

Configure G1→G2 pipeline.

pub fn with_g2_to_g3_pipeline(self, config: PipelineConfig<G2, G3>) -> Self

Configure G2→G3 pipeline.

pub fn with_g2_to_g4_pipeline(self, config: ObjectPipelineConfig<G2>) -> Self

Configure G2→G4 pipeline (object storage).

Uses ObjectPipelineConfig instead of PipelineConfig since G4 is object storage, not a BlockManager destination.

For distributed setups where the leader doesn’t have physical layouts, use with_enable_remote_g4(true) instead.

pub fn with_enable_remote_g4(self, enable: bool) -> Self

Enable remote G4 offloading via workers’ ObjectBlockOps.

In distributed setups, the leader doesn’t have physical layouts (only workers do). This enables G2→G4 offloading where:

1. G1→G2 chain output is routed to a remote offload task
2. The task calls workers’ ObjectBlockOps::put_blocks() via RPC
3. Workers upload blocks from their local G2 to object storage
4. Per-block results are returned and logged

This is mutually exclusive with with_g2_to_g4_pipeline() - use one or the other.

pub fn build(self) -> Result<OffloadEngine>

Build the offload engine.