Hi team, Can you help correct my understanding on something? Is

--with retry

just a way to make every step in your flow wrapped with a

retry

decorator? When your flow is running on Step Functions/Argo, it doesn't actually enforce a retry at the state machine level. It just does it at the AWS Batch/K8s level?

Correct.

--wht retry

is syntactic sugar for applying

@retry

to all the steps

Gotcha. I wouldn't mind contributing to open source to get some sort of

@retry

logic implemented at the state machine level, but would need some guidance on how to get started to facilitate that

Yeah - let's chat about that when we speak next. It is likely that we can introduce some present-by-default retries that don't alter the behavior of Metaflow execution

Okay sounds good to me. Thanks Savin! We are hacking it for now by modifying the state definition via aws cli and before we invoke

python my_flow.py step-functions trigger