I commonly write flows in which the end step exists only to finish the graph but with no real computation (e.g. a simple pass statement). When running locally, this is not a big deal. However, when running the flow using argo workflows, the system needs to spin up a full container to run the end step, which wastes time and resources. Is there any way to avoid this?

currently - a flow needs to have a

start

and an

end

step. To avoid running the extra

end

step - you can move logic from an intermediate step to the

end

step - effectively reducing the number of total steps in your flow by 1