Docs Orchestration

Scheduling and orchestration.

Lima orchestrates on Apache Airflow. This page covers the five ways a pipeline starts, how a run window and its deadline work, what happens when a check trips, how Lima tasks appear inside your own DAGs, and how enterprise schedulers and alerting fit.

Five trigger modes

ModeStarts a pipeline whenTypical use
ScheduleA time on the business calendar arrives.Nightly planning loads before the MRP run.
EventA file arrives, a webhook fires or a message lands.Partner files on SFTP; WMS events.
DependencyThe pipelines it depends on finish.Load stock after items and bills of material.
Quality gateA check passes; a failed check pauses dependents.Hold the planning load while an inbound feed is wrong.
External schedulerControl-M, AutoSys, Tidal, IBM Workload Scheduler or cron calls the job.Estates that keep an enterprise scheduler.

Run windows and deadlines

A run window has a start and a deadline on the business calendar, such as 01:00 to 05:30 before a 06:00 MRP run. The deadline clock runs while the pipeline runs. When a quality gate trips, the run pauses, the dependent runs pause, and the clock pauses with them. When the cause is fixed, the run resumes from its checkpoint and replays what it missed in order; every write carries a key, so nothing is applied twice. The report says whether the deadline was met and by how much.

Lima tasks in your Airflow DAGs

# illustrative: the operator name and parameters are in the SDK reference from lima.airflow import LimaPipelineOperator load_items = LimaPipelineOperator( task_id="load_items_to_kinaxis", pipeline="kinaxis_items_net_change", run_window={"deadline": "06:00", "calendar": "plant_1200"}, on_check_failure="pause", # pause the run and its clock; resume from checkpoint )

An estate that already runs Airflow keeps its DAGs and adds Lima tasks; Lima's own Airflow can also host the DAGs. Airflow 3 is supported; Airflow 2 reached end of life in April 2026.

Alerts and external schedulers

Alerts go to Slack, Teams, PagerDuty or email, with the pipeline, the check that tripped, the owner and the report link. An external scheduler starts a Lima job through the external-scheduler trigger and receives success, held or failed with the report link.