quiet-artist-13956
09/09/2024, 5:51 PMfreezing-airport-6809
freezing-airport-6809
quiet-artist-13956
09/09/2024, 6:00 PMglamorous-carpet-83516
09/09/2024, 6:43 PM@workflow
def wf():
t1() >> t2()
freezing-airport-6809
quiet-artist-13956
09/09/2024, 9:02 PMfreezing-airport-6809
freezing-airport-6809
freezing-airport-6809
quiet-artist-13956
09/10/2024, 12:35 AMwhen do we do that?Many cases: 1. After a failure, user wants to start from the failed task and a input change. a. In this case, we ask our user to submit another flyte execution and let output caching to skip the succeeded tasks. (It's okay, but some task disabled cache) 2. Without a failure, user wants to start the workflow from task C (let's assume a wf with task A=>B=>C=>D) a. C might not have inputs needed, it can be a stateless task b. C might have a hardcoded inputs, let's say
hdfs_path=/user/biao/data/
c. C might have a inputs which is outputs of B:
i. In this case, we need setup the inputs dataset using some workaround, manually change the inputs of Cfreezing-airport-6809
freezing-airport-6809
freezing-airport-6809
future-monitor-58430
09/10/2024, 7:26 AMfreezing-airport-6809