execution_date format

[HumanIearning]

Apache Airflow version

2.9.2

If "Other Airflow 2 version" selected, which one?

No response

What happened?

I don't know exactly it's bug, but I think it's weird
execution date in dag task context is different from when dag is run by manually and scheduler

when triggered

Marking task as UP_FOR_RETRY. dag_id=crawling_dag, task_id=process_goods_info_html, run_id=manual__2024-06-18T10:07:49.528041+00:00, execution_date=20240618T100749, start_date=20240618T100751, end_date=20240618T100751

when scheduled

Marking task as SUCCESS. dag_id=crawling_dag, task_id=process_goods_info_html, run_id=scheduled__2024-06-17T09:54:00+00:00, execution_date=20240617T095400, start_date=20240618T095402, end_date=20240618T095402

it looks that execution_date is same
but when print them

when triggered

2024-06-18 10:07:49.528041+00:00

when scheduled

2024-06-17 10:17:00+00:00

What you think should happen instead?

both format should be same

How to reproduce

date_format = "%Y-%m-%d %H:%M:%S%z"
date = kwargs["execution_date"]
print(str(date))
date = datetime.datetime.strptime(str(date), date_format).date()

Operating System

PRETTY_NAME="Debian GNU/Linux 12 (bookworm)

Versions of Apache Airflow Providers

apache-airflow-providers-celery==3.7.2
apache-airflow-providers-common-io==1.3.2
apache-airflow-providers-common-sql==1.14.0
apache-airflow-providers-fab==1.1.1
apache-airflow-providers-ftp==3.9.1
apache-airflow-providers-http==4.11.1
apache-airflow-providers-imap==3.6.1
apache-airflow-providers-smtp==1.7.1
apache-airflow-providers-sqlite==3.8.1

Deployment

Docker-Compose

Deployment details

use docker on mac
use docker compose up

modified some env

# image: ${AIRFLOW_IMAGE_NAME:-apache/airflow:2.9.2}
build: .

add env

AIRFLOW__CORE__DEFAULT_TIMEZONE: 'Asia/Seoul'

Anything else?

No response

Are you willing to submit PR?

• Yes I am willing to submit a PR!

Code of Conduct

• I agree to follow this project's Code of Conduct

[raphaelauv]

yes , you can convert it by

from pendulum import DateTime
def cast_pendulum_date(date: DateTime) -> str:
    return date.format("YYYY-MM-DD HH:mm:ss+00:00")

DAG(
...
user_defined_macros={
            "cast_pendulum_date": cast_pendulum_date
        }
)

...


        PythonOperator(
            task_id="XXXX",
            python_callable=XXXX,
            do_xcom_push=True,
            op_kwargs={
                "date": "{{ cast_pendulum_date(logical_date) }}",
            },
        )

[HumanIearning]

thank you!

[shahar1]

It seems that @raphaelauv 's snippet could do the work if it's only a matter of formatting.
It doesn't seem like a bug, though.
Converting into a discussion.