Pass triggering user information to DAG

[yupbank]

Description

It would be very useful if we can pass triggering user information to DAG. So that we can bring over the info to the downstream system for security auditing purpose
Use case / motivation

Expose some macro in the DAG, so user can use the trigger user info at will

Are you willing to submit a PR?

If someone can point me in the right direction
Related Issues

No

[boring-cyborg[bot]]

Thanks for opening your first issue here! Be sure to follow the issue template!

[eladkal]

In manual jobs there is a specific user who triggered it - that doesn't exist for scheduled jobs.
What value are you expecting to have for scheduled jobs?

[yupbank]

I'm looking for the manual jobs actually, currently, we can pass the info through the dag_run.conf in cli use case.
but the web UI dag_run.conf is a bit complicated. I'm wondering if it makes sense to add some current_user field in the default conf https://github.com/apache/airflow/blob/main/airflow/www/views.py#L1490 here
something like default_conf=json.dumps(dict(current_user=current_user.user.username))

it would be amazing if a default default_conf can be configured for all CLI and web UI submissions. Since airflow would perform checks on the triggering user anyway. why not pass it in to the DAG so default_conf actually have some default info

[yupbank]

e.g., i could define a BashOperator that login to a web server, in which all the credentials are managed by connections for sure. but it would also be nice to tell the web server that who triggered the job manually if that happened

[uranusjr]

Not commenting on whether this is a good idea or not, IMO a scheduled job could be categorised under either the DAG owner (since they scheduled the DAG) or None.

[yupbank]

maybe let me re-state the feature: allow the user to customize the default DAG.default_conf in the web console?

[santosh-d3vpl3x]

airflow.models.log.Log should already contain the information about trigger. Here the owner field indicates user performing actions.

You can listen on after_insert in sql alchemy to route this info via a simple airflow plugin.

[yupbank]

i'm not sure i follow, how would the plugin pass on the after_insert info to a DAG execution?

[MinusYuan]

In manual jobs there is a specific user who triggered it - that doesn't exist for scheduled jobs. What value are you expecting to have for scheduled jobs?

Hi, I want to ask which variable/what kind of method can I use to get this specific user when manual trigger happens?
Now, I also need this to know which user trigger.

[avijstein]

In case folks are still looking for a solution to this (or future folks are looking, as I was last week), more recent versions of Airflow (3.1 and higher, I believe) allow you to pull triggering_user_name from the dag_run table in context. This has solved our issue of logging manual DAG triggers.