Airflow Kubernetes Executor spams etcd when task fails because of exceeded quota error

[chenyair]

Apache Airflow version

Other Airflow 2 version (please specify below)

What happened

I am using Airflow 2.5.2 but this issue applies to all versions of Airflow.
When I'm creating a task and I don't have enough quota for the Airflow executor to create the kubernetes api returns an ApiException with the status code 403 that says Reason: Forbidden with the message: Pods ... is forbidden: exceeded quota: .... The Kubernetes executor puts the task back in the queue because the status code is not 400 or 422, from kubernetes_executor.py:

# These codes indicate something is wrong with pod definition; otherwise we assume pod
# definition is ok, and that retrying may work
if e.status in (400, 422):

The problem is that it tries excessively to run the task again and again and it spams the Kubernetes API which then makes kyverno write a lot of obejcts to etcd.

What you think should happen instead

I want to be able to control the amount of times the scheduler re-queues a job and the timeout between each time it tires to re-run the task if it was re-queued.

How to reproduce

Run an Airflow task with insufficient memory and cpu in the ACRQ

Operating System

Red Hat Enterprise Linux 8.5 (Ootpa)

Versions of Apache Airflow Providers

No response

Deployment

Other 3rd-party Helm chart

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

• I agree to follow this project's Code of Conduct

[boring-cyborg]

Thanks for opening your first issue here! Be sure to follow the issue template! If you are willing to raise PR to address this issue please do so, no need to wait for approval.

[dirrao]

Hi @chenyair,
I have seen the same behavior. I have prepared the MR to handle this issue and it will be open for review soon.
#36403