Catchup and backfill behavior in Apache Airflow - Time & Space Complexity

Time Complexity: Catchup and backfill behavior

O(n)

Understanding Time Complexity

When Airflow runs tasks for past dates, it uses catchup and backfill features.

We want to know how the number of runs grows as the date range grows.

Scenario Under Consideration

Analyze the time complexity of this Airflow DAG scheduling snippet.

from airflow import DAG
from airflow.operators.dummy_operator import DummyOperator
from datetime import datetime, timedelta

default_args = {'start_date': datetime(2023, 1, 1)}

dag = DAG('example_dag', default_args=default_args, schedule_interval='@daily', catchup=True)

task = DummyOperator(task_id='dummy_task', dag=dag)

This DAG runs daily starting from Jan 1, 2023, with catchup enabled to run missed dates.

Identify Repeating Operations

Identify the loops, recursion, array traversals that repeat.

• Primary operation: Airflow schedules one task run per missed day.
• How many times: Once for each day from start_date until today.

How Execution Grows With Input

As the number of missed days grows, the number of task runs grows linearly.

Input Size (days missed)	Approx. Operations (task runs)
10	10
100	100
1000	1000

Pattern observation: The number of task runs grows directly with the number of missed days.

Final Time Complexity

Time Complexity: O(n)

This means the work Airflow does grows in a straight line as the number of missed days increases.

Common Mistake

[X] Wrong: "Catchup runs all missed tasks instantly with no extra cost as days increase."

[OK] Correct: Each missed day adds a separate task run, so more days mean more work and time.

Interview Connect

Understanding how Airflow schedules past runs helps you explain system behavior clearly and shows you grasp task scaling.

Self-Check

"What if catchup was set to False? How would the time complexity of task runs change when starting the DAG late?"

Practice

1. What does setting catchup=True do in an Airflow DAG?

easy

A. It manually triggers DAG runs for a specific date range.

B. It disables all scheduled DAG runs.

C. It deletes all past DAG runs.

D. It runs all missed DAG runs automatically to catch up.

Solution

1. Step 1: Understand catchup parameter

   The catchup parameter controls whether Airflow runs missed DAG runs automatically.

2. Step 2: Effect of catchup=True

   When catchup=True, Airflow schedules and runs all DAG runs that were missed since the last run.

3. Final Answer:

   It runs all missed DAG runs automatically to catch up. -> Option D

4. Quick Check:

   catchup=True means run missed DAG runs [OK]

Hint: Catchup=True means Airflow fills missed runs automatically [OK]

Common Mistakes:

• Confusing catchup with manual backfill
• Thinking catchup disables scheduling
• Assuming catchup deletes runs

2. Which of the following is the correct way to disable catchup in an Airflow DAG definition?

easy

A. dag = DAG('my_dag', backfill=False)

B. dag = DAG('my_dag', catchup=True)

C. dag = DAG('my_dag', catchup=False)

D. dag = DAG('my_dag', schedule_interval=None)

Solution

1. Step 1: Identify catchup parameter usage

   The catchup parameter is used to enable or disable catchup behavior in DAGs.

2. Step 2: Disable catchup by setting catchup=False

   Setting catchup=False disables automatic running of missed DAG runs.

3. Final Answer:

   dag = DAG('my_dag', catchup=False) -> Option C

4. Quick Check:

   Disable catchup with catchup=False [OK]

Hint: Use catchup=False to stop automatic missed runs [OK]

Common Mistakes:

• Using backfill parameter instead of catchup
• Setting schedule_interval to None disables scheduling, not catchup
• Setting catchup=True enables, not disables

3. Given a DAG with start_date='2024-04-01', schedule_interval='@daily', and catchup=True, if the DAG was paused and resumed on 2024-04-05, which DAG runs will Airflow execute immediately after resuming?

medium

A. Runs for 2024-04-01 to 2024-04-05

B. Only the run for 2024-04-05

C. No runs will be executed automatically

D. Runs for 2024-04-02 to 2024-04-04 only

Solution

1. Step 1: Understand catchup=True behavior

   With catchup=True, Airflow runs all missed DAG runs from start_date to current date.

2. Step 2: Identify missed runs

   The DAG was paused from 2024-04-01 to 2024-04-04, so runs for 2024-04-01, 02, 03, 04 were missed. On resuming 2024-04-05, Airflow runs all missed plus current.

3. Final Answer:

   Runs for 2024-04-01 to 2024-04-05 -> Option A

4. Quick Check:

   Catchup runs all missed dates including current [OK]

Hint: Catchup runs all missed dates from start_date to now [OK]

Common Mistakes:

• Thinking only current date runs after resume
• Ignoring runs before resume date
• Assuming no runs run automatically after pause

4. You want to run DAG runs manually for the dates 2024-03-01 to 2024-03-05 without enabling catchup. Which command should you use?

medium

A. airflow dags backfill -s 2024-03-01 -e 2024-03-05 my_dag

B. airflow dags trigger my_dag --start-date 2024-03-01 --end-date 2024-03-05

C. airflow dags run my_dag 2024-03-01 to 2024-03-05

D. airflow dags catchup my_dag -s 2024-03-01 -e 2024-03-05

Solution

1. Step 1: Identify manual run command for date range

   The airflow dags backfill command runs DAG runs manually for a specified date range.

2. Step 2: Confirm correct syntax

   Using -s and -e flags with backfill specifies start and end dates for manual runs.

3. Final Answer:

   airflow dags backfill -s 2024-03-01 -e 2024-03-05 my_dag -> Option A

4. Quick Check:

   Backfill runs DAG manually for date range [OK]

Hint: Use airflow dags backfill with -s and -e for manual runs [OK]

Common Mistakes:

• Using trigger command with date range flags (not supported)
• Trying to use catchup command (does not exist)
• Using dags run with invalid date range syntax

5. You have a DAG with catchup=False and a start_date set to 2024-01-01. You want to run all DAG runs from 2024-01-01 to 2024-01-10 manually. What is the best approach to achieve this without enabling catchup?

hard

A. Set catchup=True and wait for Airflow to run all missed DAG runs automatically.

B. Use the backfill command for the date range 2024-01-01 to 2024-01-10.

C. Manually trigger the DAG once on 2024-01-10; Airflow will run all previous runs.

D. Change the start_date to 2024-01-10 and run the DAG normally.

Solution

1. Step 1: Understand catchup=False effect

   With catchup=False, Airflow skips running missed DAG runs automatically.

2. Step 2: Use backfill to run missed DAG runs manually

   To run missed DAG runs without enabling catchup, use the backfill command specifying the date range.

3. Final Answer:

   Use the backfill command for the date range 2024-01-01 to 2024-01-10. -> Option B

4. Quick Check:

   Backfill runs missed DAG runs manually without catchup [OK]

Hint: Backfill runs missed DAG runs manually when catchup is off [OK]

Common Mistakes:

• Enabling catchup when not desired
• Assuming one trigger runs all missed dates
• Changing start_date to skip missed runs