retries

etl_craft.execution.retries

Persist automatic retry admission; delayed attempts use no worker slots.

RetryPolicy dataclass

Additional attempts and capped exponential delay after each failure.

delay

delay(attempt_number: int) -> float

Return the delay after this attempt, capped at one hour.

task_retry_policy

task_retry_policy(params: Mapping[str, str], config: ConnectorConfig) -> RetryPolicy

Read and validate common task retry parameters, with orchestration defaults.

refresh_retries

refresh_retries(engine: Engine, config: ConnectorConfig, pipeline_run_id: int, task_ids: Set[int]) -> tuple[dict[int, datetime], set[int]]

Queue eligible failures once; return delayed attempts and exhausted task IDs.