Incorrect await_container_completion in KubernetesPodOperator

[windmark]

Apache Airflow version

2.3.4

What happened

The await_container_completion function has a negated condition

while not self.container_is_running(pod=pod, container_name=container_name):

that causes our Airflow tasks running <1 s to never be completed, causing an infinite loop.

I see this was addressed and released in #23883, but later reverted in #24474. How come it was reverted? The thread on that revert PR with comments from @jedcunningham and @potiuk didn't really address why the fix was reverted.

What you think should happen instead

Pods finishing within 1s should be properly handled.

How to reproduce

No response

Operating System

Linux

Versions of Apache Airflow Providers

apache-airflow-providers-cncf-kubernetes==4.3.0

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

[eladkal]

It caused failure of tests and we were unable to point the source so we had to revert
#24479

https://apache-airflow.slack.com/archives/CCPRP7943/p1655306336391429
https://apache-airflow.slack.com/archives/CCPRP7943/p1655300699968739

Quoting Ash from slack:

That'll learn us for merging a bug fix without any new tests

Feel free to raise PR with tests :)

[windmark]

It caused failure of tests and we were unable to point the source so we had to revert #24479

https://apache-airflow.slack.com/archives/CCPRP7943/p1655306336391429 https://apache-airflow.slack.com/archives/CCPRP7943/p1655300699968739

Quoting Ash from slack:

That'll learn us for merging a bug fix without any new tests

Feel free to raise PR with tests :)

I see, thanks for the links. Though it's quite a severe bug if the condition incorrectly is negated when waiting for completed pods. Let me read up on the threads.