GKEStartPodOperator does not respect execution_timeout when deferrable=True

[jimmykodes]

Under which category would you file this issue?

Providers

Apache Airflow version

3.3.0

What happened and how to reproduce it?

GKEStartPodOperator(
  image="debian:bullseye",
  cmds=["bash"],
  arguments=["-c", "sleep 480"],
  deferrable=True,
  execution_timeout=timedelta(minutes=2),
  # ...
)

instead of the execution timeout causing the job to fail at the 2 minute marks, it chugs along for all 4 minutes and then succeeds.

What you think should happen instead?

Deferred pods should respect their execution timeout, similar to what was implemented in the Kubernetes provider here:
#67229

Operating System

No response

Deployment

No response

Apache Airflow Provider(s)

google

Versions of Apache Airflow Providers

apache-airflow-providers-cncf-kubernetes==10.20.0
apache-airflow-providers-google==22.2.2

Official Helm Chart version

1.18.0

Kubernetes Version

No response

Helm Chart configuration

No response

Docker Image customizations

No response

Anything else?

Every single deferrable pod with an execution timeout set does not respect it.

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.

[AyushkhatiDev]

Hi @jimmykodes I'd like to take a look at this issue and work on a fix.

I'll first investigate the existing timeout handling in the Kubernetes provider from #67229 and trace how GKEStartPodOperator handles deferral and execution_timeout. I'll also look at adding regression coverage for the deferrable case.

If you have any specific implementation considerations in mind, please let me know.