Pipelines

A pipeline is the core unit of work in Laredo. Each pipeline binds one source table to one target, with optional filters and transforms in between.

Pipeline structure

Pipeline {
    id:          "pg_main:public.config_document:indexed-memory"
    source:      SyncSource      (shared across pipelines)
    table:       TableIdentifier
    filters:     []PipelineFilter
    transforms:  []PipelineTransform
    target:      SyncTarget
    buffer:      ChangeBuffer    (bounded, per-pipeline)
    errorPolicy: ErrorPolicy
    ttlPolicy:   TtlPolicy       (optional)
}

Source sharing

Sources are instantiated once and shared. If a PostgreSQL source feeds three tables into different targets, that is one source instance, one replication stream, and three target instances. The engine demuxes changes from the source by table and dispatches to the correct pipelines.

Fan-out

A single table can fan out to multiple targets. When this happens, the engine delivers each change to all targets for that table. Source ACK advances only after all targets for that source have confirmed IsDurable().

Pipeline states

Each pipeline has an independent lifecycle:

State	Meaning
INITIALIZING	Preparing to start
BASELINING	Loading initial snapshot from source
STREAMING	Receiving and applying live changes
PAUSED	Paused by operator request
ERROR	Unrecoverable error (isolated from others)
STOPPED	Shut down

Buffer policies

Each pipeline has a bounded change buffer between the source dispatcher and the target:

Policy	Behavior
block	Backpressure to source (safe default)
drop_oldest	Ring buffer — drop oldest undelivered change
error	Mark pipeline as ERROR when full

Error policies

When a pipeline fails persistently:

Policy	Behavior
isolate	Mark pipeline ERROR, continue all others
stop_source	Stop all pipelines on this source
stop_all	Halt the entire engine

Failed changes can optionally be written to a dead letter store for later inspection and replay.