Hey Dagster team, how are you? Is there a way to run dynamically generated ops sequentially instead of parallel? Perhaps some version of the map function? Currently, the only way I see is to define op-level concurrency on the job - even so, I am unable to see it execute sequentially with a partitioned job in question

The

multiprocess_executor

defaults to using the number of cores you have, but you can set it to 1 if you'd like. Would this work for your needs?

@Tim Castillo this does not work for a case where we want to run ops sequentially in given graph(), but paralelly in another graph(). In other words, Dagster is missing an extremely useful feature of setting execution type (sequential or parallel) per graph, not per job.