Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Handle fail_stop / non teardown tasks in TASK SDK #44951

Open
amoghrajesh opened this issue Dec 16, 2024 · 3 comments
Open

Handle fail_stop / non teardown tasks in TASK SDK #44951

amoghrajesh opened this issue Dec 16, 2024 · 3 comments
Assignees
@amoghrajesh
Labels
area:task-execution-interface-aip72 AIP-72: Task Execution Interface (TEI) aka Task SDK area:task-sdk
Milestone
Airflow 3.0.0

Comments

@amoghrajesh
Copy link
Contributor

amoghrajesh commented Dec 16, 2024 •
edited
Loading

Body

Handle the part where we stop all remaining tasks for while marking a task as failed.
Usage:

airflow/airflow/models/taskinstance.py

Lines 3086 to 3087 in aaf29ee

if task and fail_stop:
_stop_remaining_tasks(task_instance=ti, session=session)

@amoghrajesh amoghrajesh added the area:task-execution-interface-aip72 AIP-72: Task Execution Interface (TEI) aka Task SDK label Dec 16, 2024
@amoghrajesh amoghrajesh mentioned this issue Dec 16, 2024
Merged
kaxil pushed a commit that referenced this issue Dec 18, 2024
@amoghrajesh
AIP-72: Handling failed TI state for AirflowFailException & `Airf…
cfc2157 
…lowSensorTimeout` (#44954)

related: #44414

We already have support for handling terminal states from the task execution side as well as the task SDK client side. (almost) and failed state is part of the terminal state.

This PR extends the task runner's run function to handle cases when we have to fail a task: `AirflowFailException, AirflowSensorTimeout`. It is functionally very similar to #44786

As part of failing a task, multiple other things also needs to be done like:
- Callbacks: which will eventually be converted to teardown tasks
- Retries: Handled in #44351
- unmapping TIs: #44351
- Handling task history: will be handled by #44952
- Handling downstream tasks and non teardown tasks: will be handled by #44951

### Testing performed
#### End to End with Postman

1. Run airflow with breeze and run any DAG
![image](https://github.com/user-attachments/assets/fafc89ea-4e28-4802-912b-d72bf401d94b)

2. Login to metadata DB and get the "id" for your task instance from TI table
![image](https://github.com/user-attachments/assets/75440f0f-f62a-4277-a2e6-cb78bd666dd4)

3. Send a request to `fail` your task
![image](https://github.com/user-attachments/assets/5991e944-f416-4b79-9954-15f1a6ebdd79)

Or using curl:
```
curl --location --request PATCH 'http://localhost:29091/execution/task-instances/0193cec2-f46b-7348-9c27-9869d835dc7b/state' \
--header 'Content-Type: application/json' \
--data '{
    "state": "failed",
    "end_date": "2024-10-31T12:00:00Z"
}'
```

4. Refresh back the Airflow UI to see that the task is in failed state.
![image](https://github.com/user-attachments/assets/bb866dc6-e1d6-435e-abe4-2d04c97280ad)
LefterisXefteris pushed a commit to LefterisXefteris/airflow that referenced this issue Jan 5, 2025
@amoghrajesh
AIP-72: Handling failed TI state for AirflowFailException & `Airf…
34b8c94 
…lowSensorTimeout` (apache#44954)

related: apache#44414

We already have support for handling terminal states from the task execution side as well as the task SDK client side. (almost) and failed state is part of the terminal state.

This PR extends the task runner's run function to handle cases when we have to fail a task: `AirflowFailException, AirflowSensorTimeout`. It is functionally very similar to apache#44786

As part of failing a task, multiple other things also needs to be done like:
- Callbacks: which will eventually be converted to teardown tasks
- Retries: Handled in apache#44351
- unmapping TIs: apache#44351
- Handling task history: will be handled by apache#44952
- Handling downstream tasks and non teardown tasks: will be handled by apache#44951

### Testing performed
#### End to End with Postman

1. Run airflow with breeze and run any DAG
![image](https://github.com/user-attachments/assets/fafc89ea-4e28-4802-912b-d72bf401d94b)

2. Login to metadata DB and get the "id" for your task instance from TI table
![image](https://github.com/user-attachments/assets/75440f0f-f62a-4277-a2e6-cb78bd666dd4)

3. Send a request to `fail` your task
![image](https://github.com/user-attachments/assets/5991e944-f416-4b79-9954-15f1a6ebdd79)

Or using curl:
```
curl --location --request PATCH 'http://localhost:29091/execution/task-instances/0193cec2-f46b-7348-9c27-9869d835dc7b/state' \
--header 'Content-Type: application/json' \
--data '{
    "state": "failed",
    "end_date": "2024-10-31T12:00:00Z"
}'
```

4. Refresh back the Airflow UI to see that the task is in failed state.
![image](https://github.com/user-attachments/assets/bb866dc6-e1d6-435e-abe4-2d04c97280ad)
agupta01 pushed a commit to agupta01/airflow that referenced this issue Jan 6, 2025
@amoghrajesh @agupta01
AIP-72: Handling failed TI state for AirflowFailException & `Airf…
b70f1d7 
…lowSensorTimeout` (apache#44954)

related: apache#44414

We already have support for handling terminal states from the task execution side as well as the task SDK client side. (almost) and failed state is part of the terminal state.

This PR extends the task runner's run function to handle cases when we have to fail a task: `AirflowFailException, AirflowSensorTimeout`. It is functionally very similar to apache#44786

As part of failing a task, multiple other things also needs to be done like:
- Callbacks: which will eventually be converted to teardown tasks
- Retries: Handled in apache#44351
- unmapping TIs: apache#44351
- Handling task history: will be handled by apache#44952
- Handling downstream tasks and non teardown tasks: will be handled by apache#44951

### Testing performed
#### End to End with Postman

1. Run airflow with breeze and run any DAG
![image](https://github.com/user-attachments/assets/fafc89ea-4e28-4802-912b-d72bf401d94b)

2. Login to metadata DB and get the "id" for your task instance from TI table
![image](https://github.com/user-attachments/assets/75440f0f-f62a-4277-a2e6-cb78bd666dd4)

3. Send a request to `fail` your task
![image](https://github.com/user-attachments/assets/5991e944-f416-4b79-9954-15f1a6ebdd79)

Or using curl:
```
curl --location --request PATCH 'http://localhost:29091/execution/task-instances/0193cec2-f46b-7348-9c27-9869d835dc7b/state' \
--header 'Content-Type: application/json' \
--data '{
    "state": "failed",
    "end_date": "2024-10-31T12:00:00Z"
}'
```

4. Refresh back the Airflow UI to see that the task is in failed state.
![image](https://github.com/user-attachments/assets/bb866dc6-e1d6-435e-abe4-2d04c97280ad)
@amoghrajesh amoghrajesh self-assigned this Jan 8, 2025
@kaxil
Copy link
Member

kaxil commented Jan 9, 2025

Ths is still pending, right @amoghrajesh ? Or is this one already complete?

@kaxil kaxil added this to the Airflow 3.0.0 milestone Jan 17, 2025
@amoghrajesh
Copy link
Contributor Author

Yeah this is still pending and we are waiting on #45327 to get this one in.

@amoghrajesh amoghrajesh mentioned this issue Jan 20, 2025
Merged
@Lee-W
Copy link
Member

Lee-W commented Jan 21, 2025

just merged #45327

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@amoghrajesh amoghrajesh

Labels
area:task-execution-interface-aip72 AIP-72: Task Execution Interface (TEI) aka Task SDK area:task-sdk
Projects
AIP-72 - Task Execution Interface and SDK
Status: Todo
Milestone
Airflow 3.0.0
Development

No branches or pull requests
3 participants
@Lee-W @kaxil @amoghrajesh