Suggestion: Adding a 'machine' argument to ./run_tower
Hi
I have been using ./run_tower (https://github.com/ESCOMP/CTSM/blob/master/tools/site_and_regional/run_tower) for a couple of months, and it is a fantastic tool; however, it would be great to include a 'machine'argument, same as create_case, to work on different machines.
This is an idea of how the suggestion could look:
./run_tower --plumber-site AT-Neu .... --machine izumi
Thank you in advance for considering this enhancement issue
hi @ijaguirre thanks for the idea. I want to make sure I understand how you would use this. There's also a workaround that might work for you that I wanted to make sure you are aware of.
When you run ./run_tower it automatically tests the machine hostname to figure out the appropriate machine to use. So if you run the command on Izumi -- it'll assume the machine is izumi. And likewise it'll assume Derecho if you are logged into Derecho.
The main utility of the "--machine" option to create_newcase are things where disks are cross mounted and you want to create a case for a machine you aren't currently logged into. For example, you are logged into casper, and you want to create a case for Derecho. That's when it could be useful.
So I wanted to check to see if the machine assumption that it already does isn't sufficient for what you want to do. So can you describe how you would use the "--machine" option for your work?
The other thing that occurred to me about this is that a generic option for using "create_newcase" command line options could be added with something like the following to change the machine and compiler
--create_newcase_option '--machine izumi --compiler intel`
Since, it's general it could be used for any of the create_newcase options, and even for options that might be added in the future, when cime is updated.
Thank you very much, @ekluzek
I understand that the script reads the machine’s hostname by default, which works great on systems like Derecho and Izumi. However, when using other machines or custom installations, this approach can be limiting.
My suggestion would be to include a --machine argument—similar to how ./create_case uses it—while keeping it optional. This way, on systems like Derecho, the script can continue identifying the host automatically. But on other machines, if the detection fails, the user can simply provide this additional argument and proceed without issues.
Thanks again for the support!