Anyone have a suggestion for a try/finally-style op graph? I have a graph where

opA

runs some transformation and

opB

saves the state. I want

opB

to run regardless of whether

opA

ran successfully or not. However, I also want the job to show up as a failed job in dagit if

opA

fails.

this is some rough pseudocode but you might try something like this

@op(
    out=[Out("failed", is_required=False), Out("state", Any)]
)
def do_computation():
    state = # do the computation
    if computation_failed:
        yield Output(True, "failed")
    yield Output(state, "state")

@op
def save_state(state):
    # save the state
    return True

@op
def failure(failed, state_saved):
    raise Exception("failed")

@graph
def computation_graph():
    failed, state = do_computation()
    saved = save_state(state)
    failure(failed, saved)

in the

do_computation

you optionally yield the output

failed

if the computation failed. behind the scenes, dagster will only execute an op if all of its outputs have been provided, so if

failed

isn't yielded from

do_computation

then

failure

wont run. And then to make sure that

save_state

finishes before

failure

runs you have to create a dependency between them (this could be a Nothing dependency) this is based off of an example we have here https://docs.dagster.io/concepts/ops-jobs-graphs/graphs#with-conditional-branching

Thanks!