Airflow Core Concepts: DAGs, Tasks, Operators and Runs
The vocabulary you need to read any Airflow project: DAGs, tasks, operators, sensors, XComs, data intervals and the scheduler.
Apache AirflowBeginner2 min readUpdated 11 Oct 2026
The vocabulary
| Term | Meaning |
|---|---|
| DAG | A workflow: tasks plus their dependencies, with a schedule |
| Task | One unit of work in a DAG |
| Operator | A template for a task (PythonOperator, BashOperator, SQLExecuteQueryOperator) |
| Sensor | A task that waits for a condition (a file, a partition, another job) |
| DAG run | One execution of the DAG for one data interval |
| XCom | Small values passed between tasks (ids, paths — not datasets) |
A minimal DAG
from datetime import datetime
from airflow import DAG
from airflow.operators.python import PythonOperator
def extract(data_interval_start, data_interval_end, **_):
print(f"extract rows from {data_interval_start} to {data_interval_end}")
with DAG("orders_daily", start_date=datetime(2026, 1, 1), schedule="@daily", catchup=False) as dag:
e = PythonOperator(task_id="extract", python_callable=extract)
t = PythonOperator(task_id="transform", python_callable=lambda: None)
e >> tData intervals
A daily run for 5 January covers the interval 5 Jan 00:00 → 6 Jan 00:00, and it starts after that interval ends. Use data_interval_start and data_interval_end for what to process, never "now".
Airflow orchestrates; it does not process
Heavy work belongs in Spark, a warehouse or a container. Airflow tasks should start that work and wait for it, keeping workers light.
Finished this note?