KubernetesPodTrigger runs forever if non-base container in pod fails

[ndellosa95]

Under which category would you file this issue?

Providers

Apache Airflow version

3.2.0

What happened and how to reproduce it?

KubernetesPodTrigger only looks at the state of the base container when polling for state instead of the state of the pod itself. If you include an init container and that container fails, the base container stays in waiting state forever and the trigger never completes.

What you think should happen instead?

The trigger should look at the overall state of the pod and complete when the pod moves to a terminal state.

Operating System

Not Applicable

Deployment

Astronomer

Apache Airflow Provider(s)

cncf-kubernetes

Versions of Apache Airflow Providers

10.16.1

Official Helm Chart version

Not Applicable

Kubernetes Version

Not Applicable

Helm Chart configuration

No response

Docker Image customizations

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

[jroachgolf84]

@ndellosa95 - thanks for shooting this one in. I think this is a change that needs to be made. In _wait_for_pod_start, it calls out waiting for the Pod state (Loops until pod phase leaves ``PENDING`` If timeout is reached, throws error.), but the method itself returns the Container state.

I think it would make sense to create a new method called define_pod_state that implements the correct logic.

[hyungryuk]

Interesting! I'd like to take a look at this issue.

[jroachgolf84]

@hyungryuk, please feel free to open a PR for this issue!