consider changes to new COMPS scheduling support #2261
Labels
COMPS
Documentation
Research
Tickets that are mostly planning or researching items before implementation
Support
I ran into several issues when attempting to get a researcher's model running w/ idmtools in COMPS, mostly around trying to use the newer COMPS scheduling functionality.
scheduling=True
argument to experiment.run() (otherwise, it simply adds the asset as file_type "input", which makes it show up and seem to the user like it has done the right thing, but it doesn't actually get used for scheduling. If possible, we shouldn't require this, but if it must be required, at least there should be a note in the method documentation warning the user that they have to also add that argument. (It's also unclear to me whatsetattr(_simulation, 'scheduling', True)
does and how it's related to the other scheduling parameter)._update
argument is being used for, why it's True for regular simulations and False for TemplatedSimulations. The current behavior also prevents me from updating the command-line dynamically from the CommandTask. In my case, I had a WorkOrder.json with some of static scheduling information that I wanted to use, but needed some runtime information that changed the command-line, so can't hard-code it in a WorkOrder.json. I could create a WorkOrder string and write that to disk just so that I can read it back out again, but that seems silly, so instead, I had to change my workflow to put all the scheduling in code. Why not just say that in all cases, we use what's in WorkOrder, unless that's empty, in which case, overwrite with the CommandTask? That way the user has flexibility to use whichever method they want and the_update
argument can be removed completely.The text was updated successfully, but these errors were encountered: