State Management Capability

The State Management capability provides unified state persistence across all pipeline modes.

Contract

public interface StateManagementContract extends BaseProviderContract {
    void write(String key, Object state);
    Object read(String key);
    List<Object> query(Map<String, Object> filter);
    void delete(String key);
    void checkpoint(String executionId);
    void restore(String executionId, String checkpointId);
}

What State is Managed

State Type	Description
Pipeline execution state	Current status (RUNNING, COMPLETED, FAILED, etc.)
Streaming state	Checkpoints, savepoints, watermarks, offsets
Business dates	Logical processing dates
Metrics	Historical metric values
Recovery metadata	Last checkpoint, failed stage, retry count
Provider-specific state	Provider-managed operational state

State Key

All state is keyed by the composite StateKey:

record StateKey(
    String tenancy,
    String namespace,
    String dataset,
    String pipeline,
    String businessDate,
    String runId
) { }

State Transitions

State transitions are validated by the Echo service. Invalid transitions are rejected.

See Execution Modes for the full state machine diagrams.

Example Providers

Provider	Use Case
PostgreSQL	Durable state with SQL queries
Redis	Low-latency state for streaming
DynamoDB	Serverless, auto-scaling state
Apache Zookeeper	Distributed coordination state