GitHub user ahipp13 created a discussion: Airflow 3.0.6 Kubernetes Executor
Fails on First "ErrImagePull"
Hello all,
Putting this here in hopes somebody can help me understand this and let me know
if you think this is a bug or if it is intended behavior.
I am running Airflow 3.0.6 with the Kubernetes Executor. Right now sometimes
when a task pod starts up, it is unable to pull an image right away. So what
happens is the pod will give an "ErrImagePull", and then a "ImagePullBackOff",
and this loop goes until it finally is able to pull the image (usually takes
about 2 to 3 tries).
But what is happening is when Airflow sees the first "ErrImagePull", it
automatically marks the task as failed... So when it finally does get the image
and tries to run, it errors out with this error:
{"timestamp":"2025-09-24T13:01:08.094382Z","level":"warning","event":"Server
error","detail":{"detail":{"reason":"invalid_state","message":"TI was not in a
state where it could be marked as running","previous_state":"failed"}},"logg
er":"airflow.sdk.api.client"}
{"timestamp":"2025-09-24T13:01:08.104390Z","level":"info","event":"Process
exited","pid":9,"exit_code":-9,"signal_sent":"SIGKILL","logger":"supervisor"}
Is this Airflow's intended function, or is this a bug? I guess I feel as if on
an ErrImagePull it shouldn't immediately mark the task as failed as it could
pull the image in later.. But maybe I am wrong. If somebody could please get
back to me, as we have PROD jobs failing because of this, that would be greatly
appreciated.
GitHub link: https://github.com/apache/airflow/discussions/56049
----
This is an automatically sent email for [email protected].
To unsubscribe, please send an email to: [email protected]
