Imagine you have a DAG scheduled to run every 5 minutes:

with DAG(dag_id='sample_dag', schedule='*/5 * * * *', ...) as dag:
  ...

Once in a while, a DAG Run may take more than 5 minutes causing the next DAG Run of the same DAG to trigger in parallel, i.e., you’ll end up having two DAG runs of the same DAG running parallelly/concurrently at the same time.

But you want to ensure that only one instance of the DAG can run at any given point in time. Not more.

Use max_active_runs DAG parameter to achieve that:

with DAG(dag_id='sample_dag', schedule='*/5 * * * *', max_active_runs=1, ...) as dag:
  ...

# or when using the @dag decorator

@dag(
  schedule='*/5 * * * *',
  max_active_runs=1,
)
def sample_dag():
  ...

max_active_runs=1 will ensure that Airflow schedules any subsequent DAG Run only after the end of the previous running instance in cases of overlaps.