fix task_instance can not get task attribute correctly when turning on sentry

[Lee-W]

Closes: #36957

As reported by @jkramer-ginkgo, we also encountered a similar issue, and I'm able to reproduce it with the steps provided. Among the suggested solutions by @jkramer-ginkgo , I think it might make more sense to get ti from kwargs instead of removing this decorator. The root cause is that

airflow/airflow/models/taskinstance.py

Line 3225 in db2b75c

cls,

cannot get task attribute

airflow/airflow/sentry.py

Line 127 in db2b75c

task = task_instance.task

. But this decorator is also used in https://github.com/apache/airflow/blob/2.8.0/airflow/models/taskinstance.py#L2280, so the original logic would still be needed. I also change _schedule_downstream_tasks arguments to be keyword-lonly so that we can ensure that we can get ti from kwargs in enrich_error.

---

^ Add meaningful description above
Read the Pull Request Guidelines for more information.
In case of fundamental code changes, an Airflow Improvement Proposal (AIP) is needed.
In case of a new dependency, check compliance with the ASF 3rd Party License Policy.
In case of backwards incompatible changes please leave a note in a newsfragment file, named {pr_number}.significant.rst or {issue_number}.significant.rst, in newsfragments.

[jkramer-ginkgo]

I'm not totally convinced this is the best fix. Per my comment I don't think @Sentry.enrich_errors makes sense on _schedule_downstream_tasks. The task is already wrapped by @Sentry.enrich_errors in _run_raw_task and the task being passed to _schedule_downstream_tasks is not being run again: it's downstream tasks will be scheduled and again wrapped by @Sentry.enrich_errors within _run_raw_task

[Lee-W]

I'm not totally convinced this is the best fix. Per my comment I don't think @Sentry.enrich_errors makes sense on _schedule_downstream_tasks. The task is already wrapped by @Sentry.enrich_errors in _run_raw_task and the task being passed to _schedule_downstream_tasks is not being run again: it's downstream tasks will be scheduled and again wrapped by @Sentry.enrich_errors within _run_raw_task

I thought we still wanted to have enriched errors when we scheduled the tasks because errors might also occur at scheduling time. But I too don't have the full context of why this was designed this way back then 🤔

[dirrao]

why task instance is mandatory for sentry error logging?

[Lee-W]

It's needed by add_tagging and add_breadcrumbs