DonHaul opened a new issue, #55187:
URL: https://github.com/apache/airflow/issues/55187
### Apache Airflow version
3.0.6
### If "Other Airflow 2 version" selected, which one?
_No response_
### What happened?
I have a basic dag where a task that should be skipped, isn't.
Here its whats happening:
- Essentially if `branch1` goes to `branch2` its logs seems correct it says
`Following branch {'branch2'}` and `Skipping tasks []` which seems okay as we
dont want to skip `problematic` as it might get executed via another dependency
later
- `branch2` goes to `deadend` thus skipping dummy.
- **Even though `dummy` is skipped, `problematic` still occurs, even though
it should not**
My guess is, even though `branch1` went to `branch2` it still technically
`succeeded` as such it triggers `problematic` which has a
`NONE_FAILED_MIN_ONE_SUCCESS` Trigger Rule
<img width="1421" height="540" alt="Image"
src="https://github.com/user-attachments/assets/33f999c5-ec97-40db-a15a-451b3e9b755c";
/>
Dag code:
See in section below
A quick fix is to put a `buffer` task so the that the we see that `buffer`
actually is present on the Skipped tasks of `branch1` (see logs on the right).
Although I don't this should really be needed.
<img width="1406" height="568" alt="Image"
src="https://github.com/user-attachments/assets/b3e3030a-3306-4470-8e5e-919873e403e1";
/>
### What you think should happen instead?
`problematic_task` with `TriggerRule.NONE_FAILED_MIN_ONE_SUCCESS` should not
execute, as the only upstream successfull task is a `task.branch` that went to
another route
### How to reproduce
import the following dag. Verify that even when `deadend` is reached, the
`problematic` task is still executing
```python
import datetime
import logging
import random
from airflow.decorators import dag, task
from airflow.utils.trigger_rule import TriggerRule
logger = logging.getLogger(__name__)
@dag(
start_date=datetime.datetime(2024, 5, 5),
schedule=None,
catchup=False,
)
def hep_create_dag():
@task.branch
def branch1(**context):
value = random.choice([True, False])
if value:
return "problematic"
return "branch2"
@task
def dummy(**context,
):
print("continue")
@task
def deadend(**context):
print("deadend")
@task.branch
def branch2(**context):
if False:
return "continue"
return "deadend"
@task(trigger_rule=TriggerRule.NONE_FAILED_MIN_ONE_SUCCESS)
def problematic(**context):
print("problematic")
branch2_task = branch2()
problematic_task = problematic()
dummy_task = dummy()
deadend_task = deadend()
(
branch1()
>> [problematic_task, branch2_task]
)
branch2_task >> [ deadend_task,dummy_task,
]
(
dummy_task >> problematic_task
)
hep_create_dag()
```
### Operating System
docker in amd64 linux
### Versions of Apache Airflow Providers
default 3.0.6 installation
### Deployment
Docker-Compose
### Deployment details
_No response_
### Anything else?
_No response_
### Are you willing to submit PR?
- [ ] Yes I am willing to submit a PR!
### Code of Conduct
- [x] I agree to follow this project's [Code of
Conduct](https://github.com/apache/airflow/blob/main/CODE_OF_CONDUCT.md)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
