Hi, I have task A and task B, both mapped and A is the upstream task of B. Is there a way of having the successful children task in A move on to B regardless of the state of other children tasks in A?

What executor are you using?

daskexecutor

This should execute depth first. Can I see your code?

With Flow() as flow:

    # inputs is a list      Output = taskA_and_taskB.map(inputs, unmapped(prop)) @task() Def taskA_and_taskB(input, prop):     outputA = taskA(input, prop)     outputB = taskB(outputA, prop)

I am confused why A would rerun if B fails:

with Flow(...) as flow:
    a = task_a.map(inputs)
    task_b.map(a)

If your setup if like this, it will execute in a depth first way and then is B fails and you retry, the target from a should be retrieved for it. Or am I missing something?

if set up in your way, will the start of task_b be conditioning on the finish of all children task_a or on the finish of any task_a?

If A is 10 items and B is 10 items, B1 is dependent on A1 and B2 is dependent on A2, but only on DaskExecutor

got you, on a relevant topic, say if i also have 10 dask workers, 1 for each item. is there a way to know which worker the item is sent to?

I dont think so but there is a way to intentionally send them to a certain worker

ah i think i got it, so basically we could assign a label to the identical workers by for instance --resources "label=1", and send certain tasks to certain workers

I think so but I’ve never used

ok, time for some testings. thanks!

Nothing has changed around this in forever. Is your

checkpoint=False

inside the task? This is for a run with an agent right?

ah i thought adding environment variable has made local run work, it turned out it didn't

Ah did you figure it out?

ok, let me try