trigger_rule=TriggerRule.ONE_FAILED doesn't work properly with task_groups #30333
Description
Apache Airflow version
2.5.2 (checked on 2.5.3 also)
What happened
I'd like to setup "watcher" pattern inside of task_group, but the task always marked as "skipped".
reference to a similar issue: #30334 30334
What you think should happen instead
No response
How to reproduce
from datetime import datetime
from airflow import DAG
from airflow.decorators import task_group
from airflow.operators.python import task
from airflow.utils.trigger_rule import TriggerRule
@task(retries=0, trigger_rule=TriggerRule.ONE_FAILED)
def watcher():
pass
@task
def foo(x):
pass
@task
def failed(x):
raise Exception("Failed")
@task_group()
def process_attestation(x):
[foo(x), failed(x)] >> watcher()
with DAG(
dag_id='test',
start_date=datetime.now(),
) as dag:
# Receives sboms from database, one row per each DAG run.
# Will be changed in the future with HTTP-endpoint or something like that.
@task
def provider():
return [1, 2, 3]
process_attestation.expand(x=provider())
Operating System
Arch Linux (kernel 6.2.6)
Versions of Apache Airflow Providers
apache-airflow-providers-amazon==7.3.0
apache-airflow-providers-common-sql==1.3.4
apache-airflow-providers-ftp==3.3.1
apache-airflow-providers-google==8.11.0
apache-airflow-providers-http==4.2.0
apache-airflow-providers-imap==3.1.1
apache-airflow-providers-postgres==5.4.0
apache-airflow-providers-sqlite==3.3.1
Deployment
Docker-Compose
Deployment details
No response
Anything else
the result graph screenshot:
task instance details screenshot:
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