Ethan Veres
04/04/2022, 8:56 PMKevin Kho
04/04/2022, 9:14 PMdef registration_func(company):
with Flow(..) as flow:
company = Parameter("company", company)
...
flow.schedule = ...
flow.register()
and then you call this whenever you want to register a new function. And then if you use the default pickle-based storage that serializes the flow, this whole thing gets serialized and a copy will be uploaded (to S3 or GCS for example). This makes that file independent now of changes for any other company.
And then if the Flow inside the registration_func changes, just re-register all of them for it to take effect with some kind of loopEthan Veres
04/04/2022, 9:18 PMKevin Kho
04/04/2022, 9:24 PMpython registration_flow.py --arg1 company1 --arg2 parameter2
I don’t know the exact syntax but think you can accept arguments to this script through the command line. You can even go so far to build your own CLI, and then now you can call this from CI/CD right? I think there are a lot of possibilities and this is just one.Ethan Veres
04/04/2022, 9:26 PMKevin Kho
04/04/2022, 9:35 PMDonnchadh McAuliffe
04/05/2022, 8:24 AMKevin Kho
04/05/2022, 1:46 PMcreate_flow_run
because those are not runs created by the scheduler service. Those are runs created to “run now”. This is all for 1.0 though, not OrionDonnchadh McAuliffe
04/05/2022, 2:05 PMKevin Kho
04/05/2022, 2:08 PMEthan Veres
05/11/2022, 10:48 PMUserWarning: No result handler was specified on your Flow. Cloud features such as input caching and resuming task runs from failure may not work properly.
any ideas?Kevin Kho
05/12/2022, 5:37 AM