Hi Outerbounds, I have a question about the behavior of

retry

when used with Step Functions. I am interested in using

retry(..., minutes_between_retries=...)

to avoid periods of likely Spot interruptions. I understand that

retry

is performed at the Batch level, not within the state machine definition. But the

time.sleep

here—is that happening within the Batch job? Stepping through the logic there, it looks like

time.sleep

happens before a Batch job submission, so I'm confused about what process is doing the sleeping, and specifically whether it's happening within a Batch job or by something orchestrating the Batch job. (It's possible I just have a poor understanding of how SFN works, or that I'm looking in the wrong place 🙂 ) Anyway, I'd be very grateful for a little help in understanding how

retry

works with SFN and AWS Batch. Thanks!

To flesh out just a bit the motivation for the above question... We've noticed less than a minute between a Batch job's Attempt i stop time and its Attempt (i + 1) start time, despite having specified an inter-retry interval of 30 minutes thusly:

python my_flow.py \
    --with retry:minutes_between_retries=30 \
    step-functions create \
    ...

This has led us to question our understanding of how `@retry`'s

minutes_between_retries

parameter is intended to work.

sorry for the delay - if you are looking for the retry control for step functions - this is the code snippet. when we integrated step functions with aws batch, there wasn't any support for

minutes_between_retries

in step functions - hence that field is ignored

I see. Thanks for the explanation!

Yes, thanks!

Feel free to disregard, but I threw up a PR adding a line to the docstring noting this fact.

thanks! we usually try to keep the decorators that are common across multiple stacks free of any references to those stacks. let me think of a way to modify the docs differently.

OK, sounds good. Thanks!